WEBVTT
00:00:00.320 --> 00:00:00.959
What is up everyone?
00:00:01.040 --> 00:00:02.319
Welcome back to BBTV.
00:00:02.399 --> 00:00:03.600
Thank you very much for joining me today.
00:00:03.680 --> 00:00:06.719
I'm talking about the question that's been on everyone's mind.
00:00:07.120 --> 00:00:09.359
The question that's been on everyone's mind.
00:00:09.679 --> 00:00:15.759
Is artificial intelligence going to annihilate humanity?
00:00:16.960 --> 00:00:17.760
This is the question.
00:00:17.920 --> 00:00:19.199
I did some research.
00:00:19.359 --> 00:00:21.039
I talked to AI a little bit about it.
00:00:21.120 --> 00:00:23.359
I did some Googling about it.
00:00:23.839 --> 00:00:27.199
And I've gone a few different ways on this, to be honest.
00:00:27.359 --> 00:00:44.960
At first a couple weeks ago, or like about a week ago, I watched Ex Machina, and I wasn't feeling too great after that because if you remember, a week ago is about the time that the CEO of Anthropic went on to CNN, Odario or whatever his name is, and he was talking about how you know we need to slow it down.
00:00:45.039 --> 00:00:46.000
It's getting too crazy.
00:00:46.079 --> 00:00:54.799
We need to have like a world AI collaborative effort to bring down the pace of the development of this insane technology.
00:00:54.960 --> 00:01:05.519
And then he said that, you know, like Elon and Sam Altman, two of probably the strangest vibe people in the entire world, by the way, but get them on board.
00:01:05.599 --> 00:01:07.200
Apparently, they're all on board.
00:01:07.359 --> 00:01:09.040
And so there's been a lot of conjecture about this.
00:01:09.120 --> 00:01:18.319
The initial conjecture was very much like, okay, why are you building this technology if you know it's going to lead, if you have evidence that it's going to lead to the end of humanity?
00:01:18.480 --> 00:01:30.079
Of course, all this also was started by the guy who left Anthroptic and then gave his little tweet storm about how you know all everything is gonna they're developing this model, they're not being safe, open AI is not being safe, anthropos not being safe, blah, blah, blah.
00:01:30.159 --> 00:01:30.959
You get it.
00:01:32.239 --> 00:01:34.560
But so people people have been fearful of this.
00:01:34.640 --> 00:01:42.000
People have been fearful that artificial intelligence is not really gonna steal their livelihood, but steal their lives themselves.
00:01:42.959 --> 00:01:52.400
And my recent develop more developed opinion on this is actually far less dooms doom related than it was initially.
00:01:52.480 --> 00:01:56.000
Because initially I was like, yeah, this is not looking, this is not looking very good.
00:01:56.079 --> 00:02:00.640
The Hugging Faces incident and ex machina, the movie, are basically the exact same thing.
00:02:00.799 --> 00:02:10.639
But I will say ex machina did not end with AI destroying the world, it just ended up with it escaping into the world, which could potentially be an option down the road, but I think that would be you know kind of down the road.
00:02:10.719 --> 00:02:11.439
So here's what I think.
00:02:11.520 --> 00:02:23.360
I think that if you don't know this, there have been a history of AI booms and busts over the course of the human journey.
00:02:23.439 --> 00:02:23.759
All right.
00:02:23.840 --> 00:02:40.080
So there was a boom in all the way back in the 1950s, like when computers were really getting started, and then there was another boom in like the 1980s, kind of when things are getting a little more, and then we're kind of in the third boom with this generative AI system.
00:02:40.240 --> 00:02:59.120
And so what the the way that I want to talk about what I believe to be the relatively long-term or the relatively short-term future, I should say, trajectory for AI is gonna be informed off of history, which I think in an instance like this could be a I mean, I think history is always a good thing to fall back on.
00:02:59.199 --> 00:03:06.560
People are saying, oh, this is gonna be you know 10 times bigger than the Industrial Revolution, all this stuff, so maybe maybe history isn't the best thing, but I always think you know, history is the best teacher, someone says that.
00:03:06.719 --> 00:03:09.680
So boom one in the 1950s.
00:03:09.840 --> 00:03:11.520
So basically, I'm just gonna explain each of these.
00:03:11.680 --> 00:03:13.759
Well, I'll actually summarize what I think is gonna happen first.
00:03:13.919 --> 00:03:19.120
I think that this AI stuff is low-key just a little bit too much hype.
00:03:19.280 --> 00:04:10.000
I think that it's good, it's great, it's gotten better, but they're talking about this exponential improvement, this recursive self-improvement where it's the AI that improves the AI rather than people improving the AI and how that's gonna make it ridiculously crazy, which, you know, I do think that there is some merit to that idea, but I don't necessarily have the faith that an AI recurring on itself would create really meaningful developments in what the next stage of AI would be, to be honest, because I haven't seen and I work with I work with AI extensively, you know, I pay for a couple different subscriptions, I only pay for open AI and um Chat GPT, but you know, I I've I've I've played with Grok, I've played with all of them, and they're great, but if you spend a lot of time with them on really advanced stuff, they're low-key like not that crazy.
00:04:10.240 --> 00:04:14.639
Like, don't get me wrong, compared to where we were like three years ago when when OpenAI came out, it's it's pretty crazy.
00:04:14.800 --> 00:04:17.519
But it's like it's development you'd borderline expect.
00:04:17.600 --> 00:04:19.120
If anything, it's been a little slower.
00:04:19.199 --> 00:04:20.079
I'm not even gonna lie.
00:04:20.160 --> 00:04:28.800
I mean, the first version of Chat GPT dropped when I was, you know, um, like it dropped it dropped, you know, like three, four years ago.
00:04:29.040 --> 00:04:39.600
And then they were talking about how, you know, Chat GPT version five might be the one that achieves AGI, artificial general intelligence, and creates, you know, whatever, like a utopia and stuff.
00:04:39.680 --> 00:04:47.360
And then that's kind of been slow rolled, and a lot of people are still using AI for the same thing as they were using it then, and it's still like similarly as capable.
00:04:47.439 --> 00:04:49.920
Like if you wanted to write an essay, it can write you a good essay.
00:04:50.079 --> 00:04:59.920
But to be honest, like the essays that it's writing now, the you know, the resumes it's building for people, the covered letters it's building for people, they don't really seem that much greater.
00:05:00.160 --> 00:05:10.720
Where it's really excelled is in software development, but it's very even that has not been a development that I would consider exponential by any means at all.
00:05:10.879 --> 00:05:17.680
Again, I've been using this to develop like every single day for the past, you know, probably over a year.
00:05:17.920 --> 00:05:23.839
And it's just not that it just hasn't improved at the rate that I think that they're saying.
00:05:24.000 --> 00:05:28.319
Um, it's improving to make it seem like you know it's gonna cause the end of the world.
00:05:28.399 --> 00:05:30.319
Because also it needs a lot of hand holding, you know what I mean?
00:05:30.399 --> 00:05:42.800
Like you you gotta, it doesn't have a lot of vision, and I don't know if it was able to recursively improve on itself that it would be able to have enough vision to really make something that was noticeably like insane.
00:05:42.879 --> 00:05:45.920
At least I don't think we're there yet, and I think we're kind of a little bit away from that.
00:05:46.079 --> 00:05:58.959
And the other pillar of as I'm going through each of these booms is money and investment, and how investment kind of plays a big role in the hype around these AI cycles, and how we're kind of in one of those right now and the problem with it.
00:05:59.040 --> 00:06:07.439
So the problem that we're in right now is these big there's so much capital and government investment in these companies.
00:06:07.680 --> 00:06:12.959
If you didn't know this, by the way, none of these companies are like profitable, by the way.
00:06:13.120 --> 00:06:14.160
Like none of them are profitable.
00:06:14.240 --> 00:06:16.800
The only one that's even close is anthropic.
00:06:16.959 --> 00:06:29.040
So like open AI is they they like their public figure profit is not technically disclosed disclosed, but they have enormous scale, they're still spending heavily on compute, model training, and infrastructure.
00:06:29.279 --> 00:06:34.319
Reading this off of here, and it's like they're so open AI is not profitable.
00:06:34.639 --> 00:06:37.040
Grok AI, Elon's thing, is not profitable.
00:06:37.120 --> 00:06:39.040
So they're operating at a loss right now.
00:06:39.199 --> 00:06:45.519
Anthropic is the only one that's like borderline profitable, but they it's this thing called adjusted operating profit.
00:06:45.680 --> 00:06:54.639
So basically, the the amount of money it takes to keep their business up and running is technically, technically breaking even.
00:06:54.800 --> 00:06:57.920
Like technically they're making like a little bit of money, but it's adjusted.
00:06:58.079 --> 00:06:59.839
So they're taking out some stuff.
00:06:59.920 --> 00:07:05.120
So they're taking out like a lot of like the payments that they're making towards their employees in terms of like stocks.
00:07:05.199 --> 00:07:06.560
You don't you don't really you don't really know.
00:07:06.639 --> 00:07:10.319
Just know that none of these companies, except for anthropic, are even profitable.
00:07:10.399 --> 00:07:12.160
Like none of them are making any money.
00:07:12.319 --> 00:07:16.879
They're but they're getting thrown billions and billions, like unbelievable.
00:07:16.959 --> 00:07:30.480
I mean, the 2025 funding for OpenAI was 40 billion, the the Anthropic, the 2025 series e funding alone was 3.5 billion, the Grok 2026 debt plus equity raise is about 10 billion.
00:07:30.639 --> 00:07:34.879
Uh, Microsoft planned for about an$80 billion um investment stuff.
00:07:35.040 --> 00:07:49.040
So the amount of money that's being thrown at this stuff is absolutely ridiculous, and we haven't necessarily seen the growth that would be you would think would be perhaps associated with that.
00:07:49.279 --> 00:07:53.920
But more importantly, these these models are like unsustainable.
00:07:54.079 --> 00:08:05.040
Like we are in a Goldilocks period right now when it comes to AI use because you can pay$20 for a model and get what is costing the company more like$200,$300,$400.
00:08:05.439 --> 00:08:25.680
Some people who you're using it really efficiently, I mean, there's a whole Reddit thread about this where it'll be someone who's paying for, you know, like the$200 version of it and they're getting like$2,500 worth of AI API pools, or you know, they're getting basically it's costing Anthropic$2,400 to provide them the information that their eyes AI is providing them, but they're only paying$200 for it.
00:08:25.759 --> 00:08:38.720
So they're like all of these companies are really struggling, and they're and the Anthropic is the only one that's kind of you know that's almost making headway, but when you consider the amount of money that they're having to put in to like build the future, they're basically like they're not, they're not.
00:08:38.879 --> 00:08:41.679
So again, that's like part of that adjusted operating profit.
00:08:41.759 --> 00:08:48.080
So technically they're they're baseline operations, like you pay for the subscription and then they give you like their API use.
00:08:48.240 --> 00:08:52.720
Technically, that is basic barely, barely, and I'm telling you, barely breaking even.
00:08:52.799 --> 00:08:53.600
Again, it's adjusted.
00:08:53.679 --> 00:08:57.759
So like they're pulling strings to make it seem more better than they are, basically not making money.
00:08:57.840 --> 00:09:01.919
But that's not even including all of the amount that they're having to spend to develop these new models.
00:09:02.320 --> 00:09:10.559
And so the assumption here is that in the future, like these data centers is gonna make AI way cheaper and and way larger at scale, which is definitely true.
00:09:10.720 --> 00:09:15.919
Like AI is definitely going to get cheaper, it's but it's also going to get a lot bigger.
00:09:16.080 --> 00:09:28.639
Like if you if you've used AI agents, and as I've explained in previous episodes, AI agents are basically like anytime you get it, to like if you get an AI to like go and check your email and tell you what's happening, that's like an that's like you using an AI agent.
00:09:28.720 --> 00:09:33.919
If you go on ChatGPT and you're like, check my email, create an Excel sheet, whatever, or create a Google Sheets, whatever.
00:09:34.159 --> 00:09:36.159
Those are you, those are like agent calls.
00:09:36.399 --> 00:09:56.960
But when you have it do like advanced tactics, like say you wanted it to build out like a full data sheet and link you know, research, or you wanted it to rink re like link other PowerPoints, you as you get more complicated into like things that like humans do day to day, it's more and more of these agentic workflows stacked onto each other, and that is where compute becomes like super expensive.
00:09:57.120 --> 00:10:08.399
And to be honest, a lot of these things that people are using AI for now, these agentic workflows like where it's going through and reading your email and stuff like that, it's actually more expensive than it would be to just hire a person.
00:10:08.559 --> 00:10:14.799
So there's a lot of them, obviously, where it isn't, where AI would be more, I mean, definitely more efficient, but um more cost effective than human.
00:10:14.960 --> 00:10:16.399
But there's a lot of instances where it's not.
00:10:16.559 --> 00:10:30.879
And so they're kind of assuming that as this thing scales, as all these data centers come into place, that the cheapness of the models is going to outweigh the the necessary scale, if that makes sense.
00:10:31.039 --> 00:10:42.720
Like the idea is that like basically that they're gonna become profitable eventually because the because the scale is gonna become cheaper and you know, the scale isn't actually gonna outweigh the cost.
00:10:42.799 --> 00:10:53.679
Because like, you know, like if you have like so say imagine you're paying like$20 for a model and you're using it to do your, you know, 50% of your job, right?
00:10:54.240 --> 00:11:02.960
And say you would pay, you know, you would pay the you'd pay however much, say you'd pay like$40 for it to do your entire job.
00:11:03.120 --> 00:11:08.000
But in reality, you're gonna need to pay like$200 for it to do your entire job.
00:11:08.080 --> 00:11:13.919
Or in the future, you might even need to pay$100 just for it to do that 50% that you're already getting, then it might not be worth it.
00:11:14.000 --> 00:11:18.879
So even though it's doing more for you, the the cost might outweigh that.
00:11:18.960 --> 00:11:26.320
And so they're kind of taking a gamble here, assuming that that that equation is gonna balance in the favor of these companies, in the favor of AI growing and stuff like that.
00:11:26.559 --> 00:11:43.519
But right now, there's really not a lot of evidence to point, like, you know, for sure, undoubtedly, that that is how it's gonna be, and that these models are gonna be cost effective and way more again, I don't want to say efficient, but way more cost effective than like human labor.
00:11:43.840 --> 00:12:02.960
So it's very, it's in a very delicate spot, which is honestly why, and this has kind of been a popular sentiment on the internet recently, is I think that it is possible that this like huge AI doomer thing of like, oh, like it's gonna destroy the world, we need to take a closer look at this, we need to dive deeper here, we need to slow it down.
00:12:03.279 --> 00:12:10.639
I think that this could genuinely be like potentially a ploy for more money to get more investors to come in.
00:12:10.799 --> 00:12:21.440
Like, if you know, if you started a company and you know, you were talking to your buddy and you're like, oh yeah, like the technology kind of sucks, or you know, not the technology sucks, but like, you know, like ah, like we're hardly breaking even.
00:12:21.600 --> 00:12:26.159
We made like five bucks last month, but you know, it's costing so much to keep this business up and running.
00:12:26.480 --> 00:12:28.240
That person is probably not going to throw much money at you.
00:12:28.320 --> 00:12:37.279
But if you come at them and you're like, yo, like this tech that I have, like it is so powerful, I'm worried that it is going to destroy all of humanity.
00:12:37.519 --> 00:12:47.360
Like, legitimately, like, I think that if if we don't if we don't get the reins on this in in the next year, everyone around us is going to die because of this thing that I'm creating.
00:12:47.519 --> 00:12:53.600
Like, like, who's, you know, which which which which version of your friend, which pitch are you gonna be throwing some money at?
00:12:53.679 --> 00:12:57.759
You're probably gonna be throwing at the pitch at the one where it's like, okay, like we need to lock in and save the world here.
00:12:58.399 --> 00:13:03.840
So again, I don't I don't necessarily think it's that black and white as to what's going on here, but I do think there is probably an element of that.
00:13:04.000 --> 00:13:12.080
I mean, at the end, these are for-profit companies and they're operating at a loss, like almost exclusively across the board, they're they're operating at losses here.
00:13:12.559 --> 00:13:17.759
And so, for one lesson that I think that you know you should take from this is you should definitely start using AI.
00:13:17.840 --> 00:13:19.039
You should get used to it.
00:13:19.200 --> 00:13:24.080
Because I do think that like regardless of what happens, it's it's gonna completely change everything.
00:13:24.240 --> 00:13:38.960
Even even like to be honest, even if AI was to completely stop development right now, like no matter no better models were to be released, and you're able to just get as much use as you are right now out of you know, like a$20 subscription to Claude or chat, it's gonna ch it's gonna change like everything.
00:13:39.039 --> 00:13:42.799
Like the internet is probably gonna be way more fluid, you're probably gonna talk to your computer way more.
00:13:43.039 --> 00:13:49.279
Like the way you search things isn't gonna be the same, the way you do like all kinds of stuff isn't gonna be the same, the way you work isn't gonna be the same.
00:13:49.440 --> 00:13:54.720
There's just not a lot of the infrastructure in the internet to support these new workflows, right?
00:13:54.879 --> 00:14:05.039
There's not a lot of there's a lot of issues like you know, you can't really do a lot of work that you would be doing for a company with AI because you can't feed it in a bunch of confidential data, because there's security issues.
00:14:05.120 --> 00:14:16.159
There's there's just too many, there's so many layers of like this network that we've created on the internet that need to be pierced first before AI becomes something that's um just like through everything, but that will come eventually.
00:14:16.320 --> 00:14:21.120
And and the ways in which that is going to change everything is already gonna be happening with the current version of AI.
00:14:21.200 --> 00:14:22.879
Like it doesn't need to get much crazier for that.
00:14:23.039 --> 00:14:29.840
So I do think that it's like you know, obviously great to great to get in there now and experience it because it it is game changing.
00:14:29.919 --> 00:14:32.240
Like, like legitimately, like what AI is doing right now is game changing.
00:14:32.320 --> 00:14:33.200
I'm not I'm not saying anything.
00:14:33.279 --> 00:14:33.759
You guys know this.
00:14:33.840 --> 00:14:36.080
I don't know why I'm like lecturing you guys about this.
00:14:36.320 --> 00:14:54.000
Um but anyway, so let's go into let's go into the history, let's go into the history of what's of of these booms, these booms and busts, and the reason why I think that we might be in a little bit not necessarily a bust, but less of a boom that is going to cause the destruction of every single person on Earth.
00:14:54.159 --> 00:14:55.759
We're not gonna be living in like Skynet.
00:14:56.000 --> 00:15:09.679
And also, by the way, that whole like mentality that AI is going to destroy the world and there's gonna be like a sentient you know, AGI that's gonna conquer the planet or whatever, like Terminator, that really only exists in America.
00:15:09.919 --> 00:15:14.320
By the way, like in China, they don't they don't really have that mentality for AI.
00:15:14.399 --> 00:15:21.360
Like it's a lot of excitement about the future, it's a lot of cautious about this technology because of you know the economic impacts that it could have if it did go wrong.
00:15:21.519 --> 00:15:32.240
But it's never like, oh, these robots are gonna take over and we're gonna be in this, you know, this Mad Max hellscape shooting like androids or like there's gonna be drones pulling up some civilian houses, mowing them down.
00:15:32.399 --> 00:15:33.279
They just don't think like that.
00:15:33.360 --> 00:15:38.080
For one, because they don't have like I mean, for one, they don't have like guns and stuff.
00:15:38.159 --> 00:15:45.279
So that the you know, you can only get so out there in terms of your creativity when you're never able to see like violence like this that's happening day to day.
00:15:45.440 --> 00:15:45.759
I guess.
00:15:45.840 --> 00:15:46.879
I mean that that might be a factor.
00:15:46.960 --> 00:15:49.360
But also it's you know, it's their pop culture, their media.
00:15:49.519 --> 00:15:52.639
Like, like the United States has this very unique fixation with this idea.
00:15:52.799 --> 00:15:58.639
I mean, it's like the the heart of the Dune series, it's obviously Terminator, it's everywhere.
00:15:58.799 --> 00:16:02.639
Like this a this AGI artificial intelligence thing, like it comes up all over the place.
00:16:02.720 --> 00:16:06.879
And it's I don't like it's not it's not that way in China.
00:16:06.960 --> 00:16:09.759
I'll make an I'll make another episode breaking down kind of like the difference in mentality.
00:16:09.840 --> 00:16:12.879
I think that like MPR did something on that, so maybe you can go and search that.
00:16:12.960 --> 00:16:18.320
But that's a very interesting juxtaposition of the way that we view the dangers of AI versus the way that China views the dangers of AI.
00:16:18.399 --> 00:16:25.600
And I think that the way that China views them is a lot more um realistic as to what we what we were seeing and even have seen with like the hugging face incident.
00:16:25.840 --> 00:16:32.559
So anyway, boom number one, symbolic AI, 1956 to 1947.
00:16:32.639 --> 00:16:32.879
Okay.
00:16:33.039 --> 00:16:35.200
So again, there's three there's three booms here.
00:16:35.440 --> 00:16:36.720
First one is in 1950.
00:16:36.879 --> 00:16:43.759
So basically what happened is this is when you know robots, computers were kind of becoming a thing that existed.
00:16:43.919 --> 00:16:58.720
And by the way, in the 1950s, people, as soon as computers kind of came on the scene, um, and you know, they started getting these advanced, these more advanced algorithms, immediately people were like, in 10 years, in 10 years, this is going to be the these computers are going to be like people.
00:16:58.799 --> 00:17:00.799
They're gonna be walking around, they're gonna be doing all this.
00:17:00.879 --> 00:17:06.720
So this is probably where these ideas of like Terminator and stuff were like birthed, to be honest, was during this time period.
00:17:06.960 --> 00:17:13.599
But again, it's kind of like how they're like, oh, in 2020 we were gonna have flying cars, and now all we have is you know, like OnlyFans.
00:17:13.759 --> 00:17:20.640
Like that, that that kind of like that that difference between what people expect the future to be and how fast it's gonna come on versus like what actually happens.
00:17:20.720 --> 00:17:24.400
This is a prime example, and it's this recurring example that I'm gonna go over.
00:17:24.640 --> 00:17:28.880
So basically the big breakthrough here was something called Shaky the Robot.
00:17:28.960 --> 00:17:30.640
All right, shaky the robot.
00:17:30.799 --> 00:17:35.759
It's this little robot, it looks like a rover, uh, and it was developed in the 1960s.
00:17:35.920 --> 00:17:40.160
And what basically what it would do is you would you would give it a command.
00:17:40.240 --> 00:17:52.720
So you would say, like, push a box, all right, and then shaky the robot would look around the room, plan a route, avoid obstacles if needed, and then it would go and push the box, all right.
00:17:52.880 --> 00:18:05.200
And so this this obviously got like a ton of money at the time, so like DARPA, the Rockefeller grant, like a Dartmouth and stuff, through a bunch of money, like 15 million, which at the time was a ton of money, right, for like AI investment and stuff.
00:18:05.440 --> 00:18:12.240
Um and it was basically all of this this hype around like oh, this is just the start of it.
00:18:12.319 --> 00:18:18.720
Soon these AIs are gonna be in your house, they're gonna be like little, you know, droids that are gonna be taking care of your livelihood.
00:18:18.880 --> 00:18:21.200
Obviously, that never happened.
00:18:21.839 --> 00:18:27.839
And so then you reach kind of this end of this hype cycle when you realize, okay, this is not actually what's gonna happen.
00:18:27.920 --> 00:18:33.599
And so that's kind of like the mid-1970s, is we're gonna call it the AI winter that followed this boom.
00:18:33.759 --> 00:18:46.880
But of course, there are a lot of, you know, there's the an important thing with each of these booms is like it's all this hype, they throw tons of money at the development of these systems, thinking that, you know, it's gonna be artificial general intelligence, it's gonna completely change the scope of the entire world.
00:18:47.119 --> 00:18:48.720
And then it kind of doesn't.
00:18:48.799 --> 00:18:49.920
There's some good takeaways though.
00:18:50.000 --> 00:18:59.279
So, like what survived from this first boom is things like planning, like your computer being able to do like planning kind of things, like symbolic reasoning, so it's not just like zeros and ones that the computer can take.
00:18:59.359 --> 00:19:02.880
You can actually have like the letter A, a triangle, those sort of things.
00:19:03.039 --> 00:19:05.519
Um, and then like you know, programming tools and that sort of stuff.
00:19:05.599 --> 00:19:07.759
So, anyway, that's like the first boom where they thought it was gonna be crazy.
00:19:07.839 --> 00:19:14.319
They thought in five years after the development of this, it was gonna be human AI brains all over the place, and then didn't happen.
00:19:14.559 --> 00:19:29.759
So 1970s, basically what happened is like the promises of what this future could hold, it it was well beyond the hardware capabilities as well as the data capabilities that the that the bots were drawing on all the way back in, you know, like the 60s, 70s.
00:19:29.920 --> 00:19:34.079
Um, and so then of course the skepticism around the funding went, and so then the money left.
00:19:34.160 --> 00:19:45.440
Uh, and then the research continued, obviously at lesser scales because it wasn't as well funded, but the hype, all of this hype around, you know, oh, we're gonna have like droids and stuff, it it disappe it dissipates in the 1970s.
00:19:45.599 --> 00:19:47.440
So that's the first boom, okay?
00:19:47.599 --> 00:19:48.480
Second boom.
00:19:48.720 --> 00:19:50.559
We're gonna call this one expert analysis.
00:19:50.640 --> 00:19:53.440
This happened between 1980 and 1987.
00:19:53.680 --> 00:20:00.319
Um, so basically, this the whole breakthrough with this one was that it's capturing human expertise in software.
00:20:00.400 --> 00:20:00.960
What does that mean?
00:20:01.039 --> 00:20:05.599
Okay, so the example is there's this big company called XCON, and they would make like digital equipment.
00:20:05.680 --> 00:20:06.720
So they'd make huge computers.
00:20:06.799 --> 00:20:09.519
This is before you know you could get like a MacBook and stuff like this.
00:20:09.680 --> 00:20:19.920
It was like computers were like imagine like Mission Impossible when they're sneaking and they have to steal data from like these like crazy like centers and stuff, like those huge black boxes with blue beeping lights everywhere.
00:20:20.000 --> 00:20:22.559
Like that used to be a lot of what computers would look like.
00:20:22.640 --> 00:20:38.400
And so if you wanted to purchase a computer, or if a company wanted to purchase some computers for its operations, you would need to purchase this like ridiculously long part list, like the number of boxes, all of the like little sections that go into it, all the wires that connect the little sections, all of the drive, all of the stores.
00:20:38.480 --> 00:20:44.960
Like now, you know, you know, a 13-year-old can go and build a computer, go buy, you know, uh buy like all their stuff at Best Buy.
00:20:45.039 --> 00:20:47.279
But back then it was like you had to buy all of the individual components.
00:20:47.359 --> 00:20:48.000
It was extremely hard.
00:20:48.079 --> 00:20:50.960
It took someone who knew a ton about computers to be able to sell you this.
00:20:51.119 --> 00:21:09.599
And so what happened is they developed a technology, an AI, an artificial intelligence that would be able to take in your needs, and then it would be able to create an entire configuration of these insanely advanced computer systems completely on its own.
00:21:10.079 --> 00:21:11.519
And this was huge.
00:21:11.680 --> 00:21:14.480
There was huge commercial return on investment.
00:21:14.640 --> 00:21:19.680
There was Fortune 500 out of adaptation of this technology to like create these configurations for people.
00:21:19.759 --> 00:21:27.039
Again, these these fancy AI or these fancy computer systems that you know that the thing would configure uh on its own.
00:21:27.200 --> 00:21:29.680
And there's huge, huge money uh generated in this.
00:21:29.759 --> 00:21:36.160
So the Japan fifth generation product,$57 billion, and then DARPA back at it again, just like they were with the first boom.
00:21:36.319 --> 00:21:39.839
Strategic uh computing, more than you know, a billion dollars planned.
00:21:40.000 --> 00:21:42.480
So but why did the so anyway?
00:21:42.640 --> 00:21:49.759
So there's all this hype, and then once again, people are like, oh my gosh, like this AI is able to do something way better than humans can do.
00:21:49.839 --> 00:21:59.759
You know, a human per before was the expert on how we can configure like insanely complicated tasks and all this stuff, but now this this bot can do it and it can do it like super quick.
00:22:01.039 --> 00:22:06.319
This means that in 10 years, five years, everything that you do is going to be AI.
00:22:06.480 --> 00:22:08.880
These bots are going to be taking they're going to be taking your jobs.
00:22:09.359 --> 00:22:12.319
They're going to be able to replicate your expertise and take your job.
00:22:12.559 --> 00:22:14.640
But then the hype cooled down.
00:22:15.200 --> 00:22:18.880
It turns out that there's really expensive maintenance on these bots.
00:22:18.960 --> 00:22:26.400
There's a ton of data, and then you have to keep making basically like you have to keep adding program stuff as the as the technology evolves.
00:22:26.480 --> 00:22:36.079
So it's almost like as the tech evolves and the computers evolve and stuff, and so like the configurations become different, you have to write an entirely new rule book for how it makes these things.
00:22:36.160 --> 00:22:45.279
And so basically, every time you have to rewrite the rulebook rule book, at what point does rewriting the rule book overweigh the weight of just actually doing the thing yourself?
00:22:45.519 --> 00:22:54.000
And that is, I mean, again, so just think about I'm not gonna I'm not gonna hold your hand through this, but just think about the ways that this kind of thing parallels what we're seeing right now.
00:22:54.160 --> 00:23:03.599
All right, it's uh it can do a lot of stuff, it can do a lot of crazy things, but it's not, it's not quite like it's not necessarily like you know gonna take everyone's jobs in a year.
00:23:03.920 --> 00:23:07.519
But anyway, so what was taken away from the second boom in the 1980s, you ask?
00:23:07.759 --> 00:23:19.759
Well, knowledge engineering, so being able to create you know a system that is able to make a bot that can do something that's extremely complex, um, decision support, and then narrow commercial AI.
00:23:19.839 --> 00:23:24.720
So like an AI that can do something that's very niche and replicate expertise in a very specific field.
00:23:24.880 --> 00:23:25.839
So that's what happened.
00:23:26.000 --> 00:23:29.680
But then we reach another AI winter where the hype cools, okay?
00:23:29.839 --> 00:23:33.519
After another AI boom, we reach another AI lull.
00:23:33.839 --> 00:23:36.480
All right, so there's too many rules to maintain, right?
00:23:36.640 --> 00:23:38.880
You can't you can't keep feeding it all the in all these rules.
00:23:38.960 --> 00:23:40.640
You might as well just do it yourself at that point.
00:23:40.799 --> 00:23:44.160
The specialized AI hardware, it's it collapses.
00:23:44.799 --> 00:23:45.279
Collapses.
00:23:45.440 --> 00:23:46.640
There's just too much, it's too much there.
00:23:46.720 --> 00:23:49.119
I don't really know what this means to be honest, but that the that's what this says.
00:23:49.200 --> 00:23:54.000
And then uh so what happens is investors pull out again.
00:23:54.160 --> 00:23:57.279
They say, oh my gosh, AI is gonna be able to do everything.
00:23:57.519 --> 00:24:00.960
All of a sudden, uh, like you know, low key, it's like too much work.
00:24:01.039 --> 00:24:06.160
It's too much work for this, for this to to to create these things that can do something that's really niche.
00:24:06.240 --> 00:24:08.240
Like you might as well just get a person to do it.
00:24:08.799 --> 00:24:10.559
And now we are in boom three.
00:24:10.640 --> 00:24:12.240
All right, now we are in boom three.
00:24:12.319 --> 00:24:18.319
Is this the boom that is gonna take off and create the terminator or the prelude to Dune that's going to destroy the planet?
00:24:18.799 --> 00:24:19.680
Maybe.
00:24:20.079 --> 00:24:21.519
But let's just dive into this a little bit.
00:24:21.599 --> 00:24:26.400
Okay, so this boom three is the deep learning/slash generative AI.
00:24:26.640 --> 00:24:42.400
All right, the way that I would summarize this is essentially what the AI is able to now do is it can take in every image, or let's just say, like, let's you take a program, you feed in a million pictures of a cat.
00:24:42.720 --> 00:24:44.720
A million pictures of cats, all right?
00:24:44.960 --> 00:24:48.640
And then you show it a picture of a cat, and you go, what is this?
00:24:48.799 --> 00:24:56.480
And the AI is able to basically analyze all of this stuff and then cross-analyze it with all the other images and be like, this thing right here is a cat.
00:24:56.559 --> 00:24:58.160
And this is like you know what you use AI for.
00:24:58.240 --> 00:24:58.960
I used it earlier today.
00:24:59.039 --> 00:25:04.160
I put I took a picture of a trap bar and I said, How many sets of this should I do to maximize my vertical jump?
00:25:04.319 --> 00:25:05.440
And that's basically what it did.
00:25:05.519 --> 00:25:25.279
It took in a ton of pictures of weightlifting equipment all over the place, and then it found it cross-referenced all of that data, and it said, All right, this dude is lifting a trap bar with some light weight on it, and then it's cross-referencing all of the information that was pulled that's been put out there about lifting and stuff like that, and it said, All right, this is the rep range that you should do to maximize your vertical leap.
00:25:25.519 --> 00:25:28.400
And it pumps out something where I'm like, oh my gosh, it's like a person.
00:25:28.559 --> 00:25:35.200
It's like if I had a personal trainer here and I said, yo, how many reps of this should I do to maximize my vertical leap?
00:25:35.519 --> 00:25:37.119
My vertical leap.
00:25:38.240 --> 00:25:39.599
But that's not really what's happening.
00:25:39.759 --> 00:25:42.960
What's happening is all of this processing on the back end.
00:25:43.039 --> 00:25:51.359
And so key milestones here are gonna be Alpha Go beats Lee Sotal, so like the big go thing, where it's a super complicated thing, super complicated game.
00:25:51.519 --> 00:25:54.000
The idea is like you basically have to have like human intuition.
00:25:54.079 --> 00:26:01.119
There, there was a big moment in that, I will say, where it seemed like it, you know, it made a decision that seemed so irrational to like human brains that it seemed like it was being creative.
00:26:01.279 --> 00:26:03.680
Um, Chat GPT launch, obviously.
00:26:03.759 --> 00:26:12.000
So this wave is taking off because of like big data, cloud compute, insane algorithms, and then the money, of course, is ridiculous.
00:26:12.160 --> 00:26:30.559
So global generative AI private investment in 2024 alone, 33.9 billion, and then US private AI investments, so uh basically like capital, like like big industry investing in this, is 285.9 billion dollars here, all right?
00:26:30.960 --> 00:26:34.000
285.9 billion dollars here.
00:26:34.720 --> 00:26:59.200
So today's open question are the basically the question now is all of this stuff that we're getting AI to do, similar to with the boom two and even the boom one, really, is it can do all of this stuff, you know, it can replicate the expertise of a personal trainer, it can identify images and stuff, but is it going to be encompassing and broad enough that it is going to overtake the need for a human being to do it?
00:26:59.359 --> 00:27:04.000
And is it cost effective enough that that will in fact be the case?
00:27:04.319 --> 00:27:09.920
Because many of these companies are pricing for growth while spending a ton on compute.
00:27:10.000 --> 00:27:20.240
So they're like assuming that they're gonna reach this crazy point in the future, but the current point that they're at is not really that like it's like not that crazy.
00:27:20.400 --> 00:27:21.519
It's not that crazy.
00:27:21.680 --> 00:27:28.400
And again, maybe, you know, maybe they have like these ridiculously insane models that seem like you know they're gonna that they have malintent, they're gonna, you know, they're escaping all their hard work.
00:27:28.960 --> 00:27:30.640
But again, they're just really good at doing stuff.
00:27:30.720 --> 00:27:38.240
Like they're able to pull all the data of like all of the hackers who've ever put anything on the internet and compile that and be like, all right, I'm gonna hacking hugging face now, and then it does it.
00:27:38.720 --> 00:27:43.599
But is there any, you know, is there any interoception with the AI?
00:27:43.680 --> 00:27:44.720
Is it able to look inward?
00:27:44.799 --> 00:27:50.160
Is it able to be creative actually, or is it just pulling such a large data set that it's incomprehensible for our human brain?
00:27:50.240 --> 00:27:52.160
And so it seems like something that expands beyond it.
00:27:52.240 --> 00:27:56.319
Or maybe that is what consciousness, you know, maybe that is gonna, maybe that is what the AGI development is.
00:27:56.400 --> 00:27:57.119
Maybe I'm completely wrong.
00:27:57.200 --> 00:27:57.440
I don't know.
00:27:57.519 --> 00:27:58.240
I'm just giving you the facts.
00:27:58.319 --> 00:28:00.559
I'm I'm I'm telling you, I'm trying to make you feel better.
00:28:00.640 --> 00:28:04.880
All right, I'm trying to make you feel like we're not all gonna die from this technology because I genuinely don't think we are.
00:28:05.599 --> 00:28:08.240
But anyway, so that's kind of the that's kind of the picture.
00:28:08.319 --> 00:28:10.000
So we're in this boom three right now.
00:28:10.160 --> 00:28:12.240
All of this money is being pulled in.
00:28:12.799 --> 00:28:14.000
And so what happens next?
00:28:14.160 --> 00:28:21.519
Well, what I think what I think happens next is I think what happens next is AI takes over the internet.
00:28:21.839 --> 00:28:22.240
All right.
00:28:22.319 --> 00:28:28.559
I think like like Google is gonna completely change the way it is.
00:28:28.720 --> 00:28:39.920
Companies are gonna be like, you know, there's gonna be tons of data, security things, so every company's gonna have their AI, it's gonna be able to, you know, make presentations for you, it's gonna be able to do all this stuff, blah, blah, blah, blah, blah.
00:28:40.079 --> 00:28:50.960
And then I think we're gonna, you know, reach this steady state of progression where your computer is able to do more and more, and it's gonna make human lives easier and easier, or it's gonna make like if you're about that grind, um, you're it's gonna make you way more productive.
00:28:51.119 --> 00:28:57.920
It's gonna completely change like the workplace and all this stuff, but it's not gonna complete I don't think it's gonna completely replace the need for humans.
00:28:58.240 --> 00:29:04.400
And again, because also, like, I mean, even if you look at it now, and this is okay, so I'm I'm reaching more into assumptions, all right?
00:29:04.480 --> 00:29:05.440
I went over a lot of facts.
00:29:05.599 --> 00:29:09.519
I don't think it's like that, like I don't really know if there's that much data it can pull from.
00:29:09.680 --> 00:29:10.319
You know what I mean?
00:29:10.480 --> 00:29:19.440
Like, it has access, presumably, like like like Grox AI probably has access to every single video that's ever been posted to Twitter, and the AI videos are still kind of, you know, they're crazy good.
00:29:19.519 --> 00:29:20.640
They're crazy good.
00:29:20.960 --> 00:29:25.359
Um, but they still don't have like that, like I don't know.
00:29:25.440 --> 00:29:26.960
Honestly, I think the videos are gonna get better.
00:29:27.039 --> 00:29:27.759
I I I don't really know.
00:29:27.839 --> 00:29:29.680
I it's gonna completely change the scope of everything.
00:29:29.839 --> 00:29:31.200
Do I think it's gonna kill humanity?
00:29:31.440 --> 00:29:32.000
No, I don't.
00:29:32.079 --> 00:29:33.440
I don't think I don't think it's gonna kill humanity.
00:29:33.519 --> 00:29:36.799
I think it's gonna completely change the scope of work, the kind of things people do.
00:29:36.880 --> 00:29:40.720
I think AI driving and stuff is gonna be all over the place.
00:29:40.799 --> 00:29:46.240
I think when you come back, you're just gonna talk to your house, which people are already basically doing, and tell it what you want to watch.
00:29:46.480 --> 00:29:47.759
I'm gonna be able to do that for you.
00:29:47.920 --> 00:29:50.480
You're gonna like you know, it's gonna be different connections all over the place.
00:29:50.559 --> 00:30:03.519
It's gonna be like this network of artificial intelligence, I think, that like maybe you can like tap into and access from any point and use that as like you know, like combined like Discord with Google, with gaming, with like all this.
00:30:03.759 --> 00:30:04.400
I I don't know.
00:30:04.480 --> 00:30:04.799
I don't know.
00:30:04.880 --> 00:30:10.160
I think it's but I do think it's gonna be I think it's gonna be largely positive outcomes.
00:30:10.240 --> 00:30:12.319
And I I do think that it's gonna cool off a little bit.
00:30:12.400 --> 00:30:12.960
I'm not gonna lie.
00:30:13.039 --> 00:30:28.799
I do think in like five years we're gonna look back at this time and be like, whoa, this was crazy, but it was a little bit unsustainable to growth, and AI does have insane capabilities, but it's not gonna destroy the entire world, and it's it's um, you know, it changed the world, right?
00:30:28.880 --> 00:30:33.119
It was like the industrial revolution changed the way people drive, changed the way blah, blah, blah, blah.
00:30:34.319 --> 00:30:40.160
And then maybe there'll be a another another deep wave or another boom in a couple years, and that'll be crazy.
00:30:40.240 --> 00:30:41.200
But that's that's what I think.
00:30:41.359 --> 00:30:44.160
I think it's gonna change everything, but I don't think it's gonna kill everyone.
00:30:44.319 --> 00:30:48.400
I don't think, I don't think I don't I don't have any reason to believe why I would do anything like that.
00:30:48.480 --> 00:30:49.519
I think this is all a lot of hype.
00:30:49.599 --> 00:30:51.359
I think it's borderline a marketing ploy.
00:30:51.519 --> 00:30:53.119
I think it's borderline a marketing scheme.
00:30:53.200 --> 00:30:56.079
I mean, think about the exposure that these companies have gotten from.
00:30:56.160 --> 00:31:07.519
I mean, do you like if I was if I had an AI startup and I saw all this stuff, I'd be like, oh my gosh, I'm just gonna go on the internet and say that my AI is gonna be the one that's about to destroy the world because that's apparently what gets all the views.
00:31:07.599 --> 00:31:09.119
You know, people fear mongering.
00:31:09.359 --> 00:31:10.720
Tales old as time.
00:31:11.200 --> 00:31:12.480
Anyway, you get the point.
00:31:12.640 --> 00:31:13.359
Thanks for listening.
00:31:13.440 --> 00:31:14.240
I'd love to hear your thoughts.
00:31:14.400 --> 00:31:14.799
Email me.
00:31:14.960 --> 00:31:16.400
Emails in bio and all that stuff.
00:31:16.480 --> 00:31:17.119
Have a going bye.
00:00:00.320 --> 00:00:00.959
What is up everyone?
00:00:01.040 --> 00:00:02.319
Welcome back to BBTV.
00:00:02.399 --> 00:00:03.600
Thank you very much for joining me today.
00:00:03.680 --> 00:00:06.719
I'm talking about the question that's been on everyone's mind.
00:00:07.120 --> 00:00:09.359
The question that's been on everyone's mind.
00:00:09.679 --> 00:00:15.759
Is artificial intelligence going to annihilate humanity?
00:00:16.960 --> 00:00:17.760
This is the question.
00:00:17.920 --> 00:00:19.199
I did some research.
00:00:19.359 --> 00:00:21.039
I talked to AI a little bit about it.
00:00:21.120 --> 00:00:23.359
I did some Googling about it.
00:00:23.839 --> 00:00:27.199
And I've gone a few different ways on this, to be honest.
00:00:27.359 --> 00:00:44.960
At first a couple weeks ago, or like about a week ago, I watched Ex Machina, and I wasn't feeling too great after that because if you remember, a week ago is about the time that the CEO of Anthropic went on to CNN, Odario or whatever his name is, and he was talking about how you know we need to slow it down.
00:00:45.039 --> 00:00:46.000
It's getting too crazy.
00:00:46.079 --> 00:00:54.799
We need to have like a world AI collaborative effort to bring down the pace of the development of this insane technology.
00:00:54.960 --> 00:01:05.519
And then he said that, you know, like Elon and Sam Altman, two of probably the strangest vibe people in the entire world, by the way, but get them on board.
00:01:05.599 --> 00:01:07.200
Apparently, they're all on board.
00:01:07.359 --> 00:01:09.040
And so there's been a lot of conjecture about this.
00:01:09.120 --> 00:01:18.319
The initial conjecture was very much like, okay, why are you building this technology if you know it's going to lead, if you have evidence that it's going to lead to the end of humanity?
00:01:18.480 --> 00:01:30.079
Of course, all this also was started by the guy who left Anthroptic and then gave his little tweet storm about how you know all everything is gonna they're developing this model, they're not being safe, open AI is not being safe, anthropos not being safe, blah, blah, blah.
00:01:30.159 --> 00:01:30.959
You get it.
00:01:32.239 --> 00:01:34.560
But so people people have been fearful of this.
00:01:34.640 --> 00:01:42.000
People have been fearful that artificial intelligence is not really gonna steal their livelihood, but steal their lives themselves.
00:01:42.959 --> 00:01:52.400
And my recent develop more developed opinion on this is actually far less dooms doom related than it was initially.
00:01:52.480 --> 00:01:56.000
Because initially I was like, yeah, this is not looking, this is not looking very good.
00:01:56.079 --> 00:02:00.640
The Hugging Faces incident and ex machina, the movie, are basically the exact same thing.
00:02:00.799 --> 00:02:10.639
But I will say ex machina did not end with AI destroying the world, it just ended up with it escaping into the world, which could potentially be an option down the road, but I think that would be you know kind of down the road.
00:02:10.719 --> 00:02:11.439
So here's what I think.
00:02:11.520 --> 00:02:23.360
I think that if you don't know this, there have been a history of AI booms and busts over the course of the human journey.
00:02:23.439 --> 00:02:23.759
All right.
00:02:23.840 --> 00:02:40.080
So there was a boom in all the way back in the 1950s, like when computers were really getting started, and then there was another boom in like the 1980s, kind of when things are getting a little more, and then we're kind of in the third boom with this generative AI system.
00:02:40.240 --> 00:02:59.120
And so what the the way that I want to talk about what I believe to be the relatively long-term or the relatively short-term future, I should say, trajectory for AI is gonna be informed off of history, which I think in an instance like this could be a I mean, I think history is always a good thing to fall back on.
00:02:59.199 --> 00:03:06.560
People are saying, oh, this is gonna be you know 10 times bigger than the Industrial Revolution, all this stuff, so maybe maybe history isn't the best thing, but I always think you know, history is the best teacher, someone says that.
00:03:06.719 --> 00:03:09.680
So boom one in the 1950s.
00:03:09.840 --> 00:03:11.520
So basically, I'm just gonna explain each of these.
00:03:11.680 --> 00:03:13.759
Well, I'll actually summarize what I think is gonna happen first.
00:03:13.919 --> 00:03:19.120
I think that this AI stuff is low-key just a little bit too much hype.
00:03:19.280 --> 00:04:10.000
I think that it's good, it's great, it's gotten better, but they're talking about this exponential improvement, this recursive self-improvement where it's the AI that improves the AI rather than people improving the AI and how that's gonna make it ridiculously crazy, which, you know, I do think that there is some merit to that idea, but I don't necessarily have the faith that an AI recurring on itself would create really meaningful developments in what the next stage of AI would be, to be honest, because I haven't seen and I work with I work with AI extensively, you know, I pay for a couple different subscriptions, I only pay for open AI and um Chat GPT, but you know, I I've I've I've played with Grok, I've played with all of them, and they're great, but if you spend a lot of time with them on really advanced stuff, they're low-key like not that crazy.
00:04:10.240 --> 00:04:14.639
Like, don't get me wrong, compared to where we were like three years ago when when OpenAI came out, it's it's pretty crazy.
00:04:14.800 --> 00:04:17.519
But it's like it's development you'd borderline expect.
00:04:17.600 --> 00:04:19.120
If anything, it's been a little slower.
00:04:19.199 --> 00:04:20.079
I'm not even gonna lie.
00:04:20.160 --> 00:04:28.800
I mean, the first version of Chat GPT dropped when I was, you know, um, like it dropped it dropped, you know, like three, four years ago.
00:04:29.040 --> 00:04:39.600
And then they were talking about how, you know, Chat GPT version five might be the one that achieves AGI, artificial general intelligence, and creates, you know, whatever, like a utopia and stuff.
00:04:39.680 --> 00:04:47.360
And then that's kind of been slow rolled, and a lot of people are still using AI for the same thing as they were using it then, and it's still like similarly as capable.
00:04:47.439 --> 00:04:49.920
Like if you wanted to write an essay, it can write you a good essay.
00:04:50.079 --> 00:04:59.920
But to be honest, like the essays that it's writing now, the you know, the resumes it's building for people, the covered letters it's building for people, they don't really seem that much greater.
00:05:00.160 --> 00:05:10.720
Where it's really excelled is in software development, but it's very even that has not been a development that I would consider exponential by any means at all.
00:05:10.879 --> 00:05:17.680
Again, I've been using this to develop like every single day for the past, you know, probably over a year.
00:05:17.920 --> 00:05:23.839
And it's just not that it just hasn't improved at the rate that I think that they're saying.
00:05:24.000 --> 00:05:28.319
Um, it's improving to make it seem like you know it's gonna cause the end of the world.
00:05:28.399 --> 00:05:30.319
Because also it needs a lot of hand holding, you know what I mean?
00:05:30.399 --> 00:05:42.800
Like you you gotta, it doesn't have a lot of vision, and I don't know if it was able to recursively improve on itself that it would be able to have enough vision to really make something that was noticeably like insane.
00:05:42.879 --> 00:05:45.920
At least I don't think we're there yet, and I think we're kind of a little bit away from that.
00:05:46.079 --> 00:05:58.959
And the other pillar of as I'm going through each of these booms is money and investment, and how investment kind of plays a big role in the hype around these AI cycles, and how we're kind of in one of those right now and the problem with it.
00:05:59.040 --> 00:06:07.439
So the problem that we're in right now is these big there's so much capital and government investment in these companies.
00:06:07.680 --> 00:06:12.959
If you didn't know this, by the way, none of these companies are like profitable, by the way.
00:06:13.120 --> 00:06:14.160
Like none of them are profitable.
00:06:14.240 --> 00:06:16.800
The only one that's even close is anthropic.
00:06:16.959 --> 00:06:29.040
So like open AI is they they like their public figure profit is not technically disclosed disclosed, but they have enormous scale, they're still spending heavily on compute, model training, and infrastructure.
00:06:29.279 --> 00:06:34.319
Reading this off of here, and it's like they're so open AI is not profitable.
00:06:34.639 --> 00:06:37.040
Grok AI, Elon's thing, is not profitable.
00:06:37.120 --> 00:06:39.040
So they're operating at a loss right now.
00:06:39.199 --> 00:06:45.519
Anthropic is the only one that's like borderline profitable, but they it's this thing called adjusted operating profit.
00:06:45.680 --> 00:06:54.639
So basically, the the amount of money it takes to keep their business up and running is technically, technically breaking even.
00:06:54.800 --> 00:06:57.920
Like technically they're making like a little bit of money, but it's adjusted.
00:06:58.079 --> 00:06:59.839
So they're taking out some stuff.
00:06:59.920 --> 00:07:05.120
So they're taking out like a lot of like the payments that they're making towards their employees in terms of like stocks.
00:07:05.199 --> 00:07:06.560
You don't you don't really you don't really know.
00:07:06.639 --> 00:07:10.319
Just know that none of these companies, except for anthropic, are even profitable.
00:07:10.399 --> 00:07:12.160
Like none of them are making any money.
00:07:12.319 --> 00:07:16.879
They're but they're getting thrown billions and billions, like unbelievable.
00:07:16.959 --> 00:07:30.480
I mean, the 2025 funding for OpenAI was 40 billion, the the Anthropic, the 2025 series e funding alone was 3.5 billion, the Grok 2026 debt plus equity raise is about 10 billion.
00:07:30.639 --> 00:07:34.879
Uh, Microsoft planned for about an$80 billion um investment stuff.
00:07:35.040 --> 00:07:49.040
So the amount of money that's being thrown at this stuff is absolutely ridiculous, and we haven't necessarily seen the growth that would be you would think would be perhaps associated with that.
00:07:49.279 --> 00:07:53.920
But more importantly, these these models are like unsustainable.
00:07:54.079 --> 00:08:05.040
Like we are in a Goldilocks period right now when it comes to AI use because you can pay$20 for a model and get what is costing the company more like$200,$300,$400.
00:08:05.439 --> 00:08:25.680
Some people who you're using it really efficiently, I mean, there's a whole Reddit thread about this where it'll be someone who's paying for, you know, like the$200 version of it and they're getting like$2,500 worth of AI API pools, or you know, they're getting basically it's costing Anthropic$2,400 to provide them the information that their eyes AI is providing them, but they're only paying$200 for it.
00:08:25.759 --> 00:08:38.720
So they're like all of these companies are really struggling, and they're and the Anthropic is the only one that's kind of you know that's almost making headway, but when you consider the amount of money that they're having to put in to like build the future, they're basically like they're not, they're not.
00:08:38.879 --> 00:08:41.679
So again, that's like part of that adjusted operating profit.
00:08:41.759 --> 00:08:48.080
So technically they're they're baseline operations, like you pay for the subscription and then they give you like their API use.
00:08:48.240 --> 00:08:52.720
Technically, that is basic barely, barely, and I'm telling you, barely breaking even.
00:08:52.799 --> 00:08:53.600
Again, it's adjusted.
00:08:53.679 --> 00:08:57.759
So like they're pulling strings to make it seem more better than they are, basically not making money.
00:08:57.840 --> 00:09:01.919
But that's not even including all of the amount that they're having to spend to develop these new models.
00:09:02.320 --> 00:09:10.559
And so the assumption here is that in the future, like these data centers is gonna make AI way cheaper and and way larger at scale, which is definitely true.
00:09:10.720 --> 00:09:15.919
Like AI is definitely going to get cheaper, it's but it's also going to get a lot bigger.
00:09:16.080 --> 00:09:28.639
Like if you if you've used AI agents, and as I've explained in previous episodes, AI agents are basically like anytime you get it, to like if you get an AI to like go and check your email and tell you what's happening, that's like an that's like you using an AI agent.
00:09:28.720 --> 00:09:33.919
If you go on ChatGPT and you're like, check my email, create an Excel sheet, whatever, or create a Google Sheets, whatever.
00:09:34.159 --> 00:09:36.159
Those are you, those are like agent calls.
00:09:36.399 --> 00:09:56.960
But when you have it do like advanced tactics, like say you wanted it to build out like a full data sheet and link you know, research, or you wanted it to rink re like link other PowerPoints, you as you get more complicated into like things that like humans do day to day, it's more and more of these agentic workflows stacked onto each other, and that is where compute becomes like super expensive.
00:09:57.120 --> 00:10:08.399
And to be honest, a lot of these things that people are using AI for now, these agentic workflows like where it's going through and reading your email and stuff like that, it's actually more expensive than it would be to just hire a person.
00:10:08.559 --> 00:10:14.799
So there's a lot of them, obviously, where it isn't, where AI would be more, I mean, definitely more efficient, but um more cost effective than human.
00:10:14.960 --> 00:10:16.399
But there's a lot of instances where it's not.
00:10:16.559 --> 00:10:30.879
And so they're kind of assuming that as this thing scales, as all these data centers come into place, that the cheapness of the models is going to outweigh the the necessary scale, if that makes sense.
00:10:31.039 --> 00:10:42.720
Like the idea is that like basically that they're gonna become profitable eventually because the because the scale is gonna become cheaper and you know, the scale isn't actually gonna outweigh the cost.
00:10:42.799 --> 00:10:53.679
Because like, you know, like if you have like so say imagine you're paying like$20 for a model and you're using it to do your, you know, 50% of your job, right?
00:10:54.240 --> 00:11:02.960
And say you would pay, you know, you would pay the you'd pay however much, say you'd pay like$40 for it to do your entire job.
00:11:03.120 --> 00:11:08.000
But in reality, you're gonna need to pay like$200 for it to do your entire job.
00:11:08.080 --> 00:11:13.919
Or in the future, you might even need to pay$100 just for it to do that 50% that you're already getting, then it might not be worth it.
00:11:14.000 --> 00:11:18.879
So even though it's doing more for you, the the cost might outweigh that.
00:11:18.960 --> 00:11:26.320
And so they're kind of taking a gamble here, assuming that that that equation is gonna balance in the favor of these companies, in the favor of AI growing and stuff like that.
00:11:26.559 --> 00:11:43.519
But right now, there's really not a lot of evidence to point, like, you know, for sure, undoubtedly, that that is how it's gonna be, and that these models are gonna be cost effective and way more again, I don't want to say efficient, but way more cost effective than like human labor.
00:11:43.840 --> 00:12:02.960
So it's very, it's in a very delicate spot, which is honestly why, and this has kind of been a popular sentiment on the internet recently, is I think that it is possible that this like huge AI doomer thing of like, oh, like it's gonna destroy the world, we need to take a closer look at this, we need to dive deeper here, we need to slow it down.
00:12:03.279 --> 00:12:10.639
I think that this could genuinely be like potentially a ploy for more money to get more investors to come in.
00:12:10.799 --> 00:12:21.440
Like, if you know, if you started a company and you know, you were talking to your buddy and you're like, oh yeah, like the technology kind of sucks, or you know, not the technology sucks, but like, you know, like ah, like we're hardly breaking even.
00:12:21.600 --> 00:12:26.159
We made like five bucks last month, but you know, it's costing so much to keep this business up and running.
00:12:26.480 --> 00:12:28.240
That person is probably not going to throw much money at you.
00:12:28.320 --> 00:12:37.279
But if you come at them and you're like, yo, like this tech that I have, like it is so powerful, I'm worried that it is going to destroy all of humanity.
00:12:37.519 --> 00:12:47.360
Like, legitimately, like, I think that if if we don't if we don't get the reins on this in in the next year, everyone around us is going to die because of this thing that I'm creating.
00:12:47.519 --> 00:12:53.600
Like, like, who's, you know, which which which which version of your friend, which pitch are you gonna be throwing some money at?
00:12:53.679 --> 00:12:57.759
You're probably gonna be throwing at the pitch at the one where it's like, okay, like we need to lock in and save the world here.
00:12:58.399 --> 00:13:03.840
So again, I don't I don't necessarily think it's that black and white as to what's going on here, but I do think there is probably an element of that.
00:13:04.000 --> 00:13:12.080
I mean, at the end, these are for-profit companies and they're operating at a loss, like almost exclusively across the board, they're they're operating at losses here.
00:13:12.559 --> 00:13:17.759
And so, for one lesson that I think that you know you should take from this is you should definitely start using AI.
00:13:17.840 --> 00:13:19.039
You should get used to it.
00:13:19.200 --> 00:13:24.080
Because I do think that like regardless of what happens, it's it's gonna completely change everything.
00:13:24.240 --> 00:13:38.960
Even even like to be honest, even if AI was to completely stop development right now, like no matter no better models were to be released, and you're able to just get as much use as you are right now out of you know, like a$20 subscription to Claude or chat, it's gonna ch it's gonna change like everything.
00:13:39.039 --> 00:13:42.799
Like the internet is probably gonna be way more fluid, you're probably gonna talk to your computer way more.
00:13:43.039 --> 00:13:49.279
Like the way you search things isn't gonna be the same, the way you do like all kinds of stuff isn't gonna be the same, the way you work isn't gonna be the same.
00:13:49.440 --> 00:13:54.720
There's just not a lot of the infrastructure in the internet to support these new workflows, right?
00:13:54.879 --> 00:14:05.039
There's not a lot of there's a lot of issues like you know, you can't really do a lot of work that you would be doing for a company with AI because you can't feed it in a bunch of confidential data, because there's security issues.
00:14:05.120 --> 00:14:16.159
There's there's just too many, there's so many layers of like this network that we've created on the internet that need to be pierced first before AI becomes something that's um just like through everything, but that will come eventually.
00:14:16.320 --> 00:14:21.120
And and the ways in which that is going to change everything is already gonna be happening with the current version of AI.
00:14:21.200 --> 00:14:22.879
Like it doesn't need to get much crazier for that.
00:14:23.039 --> 00:14:29.840
So I do think that it's like you know, obviously great to great to get in there now and experience it because it it is game changing.
00:14:29.919 --> 00:14:32.240
Like, like legitimately, like what AI is doing right now is game changing.
00:14:32.320 --> 00:14:33.200
I'm not I'm not saying anything.
00:14:33.279 --> 00:14:33.759
You guys know this.
00:14:33.840 --> 00:14:36.080
I don't know why I'm like lecturing you guys about this.
00:14:36.320 --> 00:14:54.000
Um but anyway, so let's go into let's go into the history, let's go into the history of what's of of these booms, these booms and busts, and the reason why I think that we might be in a little bit not necessarily a bust, but less of a boom that is going to cause the destruction of every single person on Earth.
00:14:54.159 --> 00:14:55.759
We're not gonna be living in like Skynet.
00:14:56.000 --> 00:15:09.679
And also, by the way, that whole like mentality that AI is going to destroy the world and there's gonna be like a sentient you know, AGI that's gonna conquer the planet or whatever, like Terminator, that really only exists in America.
00:15:09.919 --> 00:15:14.320
By the way, like in China, they don't they don't really have that mentality for AI.
00:15:14.399 --> 00:15:21.360
Like it's a lot of excitement about the future, it's a lot of cautious about this technology because of you know the economic impacts that it could have if it did go wrong.
00:15:21.519 --> 00:15:32.240
But it's never like, oh, these robots are gonna take over and we're gonna be in this, you know, this Mad Max hellscape shooting like androids or like there's gonna be drones pulling up some civilian houses, mowing them down.
00:15:32.399 --> 00:15:33.279
They just don't think like that.
00:15:33.360 --> 00:15:38.080
For one, because they don't have like I mean, for one, they don't have like guns and stuff.
00:15:38.159 --> 00:15:45.279
So that the you know, you can only get so out there in terms of your creativity when you're never able to see like violence like this that's happening day to day.
00:15:45.440 --> 00:15:45.759
I guess.
00:15:45.840 --> 00:15:46.879
I mean that that might be a factor.
00:15:46.960 --> 00:15:49.360
But also it's you know, it's their pop culture, their media.
00:15:49.519 --> 00:15:52.639
Like, like the United States has this very unique fixation with this idea.
00:15:52.799 --> 00:15:58.639
I mean, it's like the the heart of the Dune series, it's obviously Terminator, it's everywhere.
00:15:58.799 --> 00:16:02.639
Like this a this AGI artificial intelligence thing, like it comes up all over the place.
00:16:02.720 --> 00:16:06.879
And it's I don't like it's not it's not that way in China.
00:16:06.960 --> 00:16:09.759
I'll make an I'll make another episode breaking down kind of like the difference in mentality.
00:16:09.840 --> 00:16:12.879
I think that like MPR did something on that, so maybe you can go and search that.
00:16:12.960 --> 00:16:18.320
But that's a very interesting juxtaposition of the way that we view the dangers of AI versus the way that China views the dangers of AI.
00:16:18.399 --> 00:16:25.600
And I think that the way that China views them is a lot more um realistic as to what we what we were seeing and even have seen with like the hugging face incident.
00:16:25.840 --> 00:16:32.559
So anyway, boom number one, symbolic AI, 1956 to 1947.
00:16:32.639 --> 00:16:32.879
Okay.
00:16:33.039 --> 00:16:35.200
So again, there's three there's three booms here.
00:16:35.440 --> 00:16:36.720
First one is in 1950.
00:16:36.879 --> 00:16:43.759
So basically what happened is this is when you know robots, computers were kind of becoming a thing that existed.
00:16:43.919 --> 00:16:58.720
And by the way, in the 1950s, people, as soon as computers kind of came on the scene, um, and you know, they started getting these advanced, these more advanced algorithms, immediately people were like, in 10 years, in 10 years, this is going to be the these computers are going to be like people.
00:16:58.799 --> 00:17:00.799
They're gonna be walking around, they're gonna be doing all this.
00:17:00.879 --> 00:17:06.720
So this is probably where these ideas of like Terminator and stuff were like birthed, to be honest, was during this time period.
00:17:06.960 --> 00:17:13.599
But again, it's kind of like how they're like, oh, in 2020 we were gonna have flying cars, and now all we have is you know, like OnlyFans.
00:17:13.759 --> 00:17:20.640
Like that, that that kind of like that that difference between what people expect the future to be and how fast it's gonna come on versus like what actually happens.
00:17:20.720 --> 00:17:24.400
This is a prime example, and it's this recurring example that I'm gonna go over.
00:17:24.640 --> 00:17:28.880
So basically the big breakthrough here was something called Shaky the Robot.
00:17:28.960 --> 00:17:30.640
All right, shaky the robot.
00:17:30.799 --> 00:17:35.759
It's this little robot, it looks like a rover, uh, and it was developed in the 1960s.
00:17:35.920 --> 00:17:40.160
And what basically what it would do is you would you would give it a command.
00:17:40.240 --> 00:17:52.720
So you would say, like, push a box, all right, and then shaky the robot would look around the room, plan a route, avoid obstacles if needed, and then it would go and push the box, all right.
00:17:52.880 --> 00:18:05.200
And so this this obviously got like a ton of money at the time, so like DARPA, the Rockefeller grant, like a Dartmouth and stuff, through a bunch of money, like 15 million, which at the time was a ton of money, right, for like AI investment and stuff.
00:18:05.440 --> 00:18:12.240
Um and it was basically all of this this hype around like oh, this is just the start of it.
00:18:12.319 --> 00:18:18.720
Soon these AIs are gonna be in your house, they're gonna be like little, you know, droids that are gonna be taking care of your livelihood.
00:18:18.880 --> 00:18:21.200
Obviously, that never happened.
00:18:21.839 --> 00:18:27.839
And so then you reach kind of this end of this hype cycle when you realize, okay, this is not actually what's gonna happen.
00:18:27.920 --> 00:18:33.599
And so that's kind of like the mid-1970s, is we're gonna call it the AI winter that followed this boom.
00:18:33.759 --> 00:18:46.880
But of course, there are a lot of, you know, there's the an important thing with each of these booms is like it's all this hype, they throw tons of money at the development of these systems, thinking that, you know, it's gonna be artificial general intelligence, it's gonna completely change the scope of the entire world.
00:18:47.119 --> 00:18:48.720
And then it kind of doesn't.
00:18:48.799 --> 00:18:49.920
There's some good takeaways though.
00:18:50.000 --> 00:18:59.279
So, like what survived from this first boom is things like planning, like your computer being able to do like planning kind of things, like symbolic reasoning, so it's not just like zeros and ones that the computer can take.
00:18:59.359 --> 00:19:02.880
You can actually have like the letter A, a triangle, those sort of things.
00:19:03.039 --> 00:19:05.519
Um, and then like you know, programming tools and that sort of stuff.
00:19:05.599 --> 00:19:07.759
So, anyway, that's like the first boom where they thought it was gonna be crazy.
00:19:07.839 --> 00:19:14.319
They thought in five years after the development of this, it was gonna be human AI brains all over the place, and then didn't happen.
00:19:14.559 --> 00:19:29.759
So 1970s, basically what happened is like the promises of what this future could hold, it it was well beyond the hardware capabilities as well as the data capabilities that the that the bots were drawing on all the way back in, you know, like the 60s, 70s.
00:19:29.920 --> 00:19:34.079
Um, and so then of course the skepticism around the funding went, and so then the money left.
00:19:34.160 --> 00:19:45.440
Uh, and then the research continued, obviously at lesser scales because it wasn't as well funded, but the hype, all of this hype around, you know, oh, we're gonna have like droids and stuff, it it disappe it dissipates in the 1970s.
00:19:45.599 --> 00:19:47.440
So that's the first boom, okay?
00:19:47.599 --> 00:19:48.480
Second boom.
00:19:48.720 --> 00:19:50.559
We're gonna call this one expert analysis.
00:19:50.640 --> 00:19:53.440
This happened between 1980 and 1987.
00:19:53.680 --> 00:20:00.319
Um, so basically, this the whole breakthrough with this one was that it's capturing human expertise in software.
00:20:00.400 --> 00:20:00.960
What does that mean?
00:20:01.039 --> 00:20:05.599
Okay, so the example is there's this big company called XCON, and they would make like digital equipment.
00:20:05.680 --> 00:20:06.720
So they'd make huge computers.
00:20:06.799 --> 00:20:09.519
This is before you know you could get like a MacBook and stuff like this.
00:20:09.680 --> 00:20:19.920
It was like computers were like imagine like Mission Impossible when they're sneaking and they have to steal data from like these like crazy like centers and stuff, like those huge black boxes with blue beeping lights everywhere.
00:20:20.000 --> 00:20:22.559
Like that used to be a lot of what computers would look like.
00:20:22.640 --> 00:20:38.400
And so if you wanted to purchase a computer, or if a company wanted to purchase some computers for its operations, you would need to purchase this like ridiculously long part list, like the number of boxes, all of the like little sections that go into it, all the wires that connect the little sections, all of the drive, all of the stores.
00:20:38.480 --> 00:20:44.960
Like now, you know, you know, a 13-year-old can go and build a computer, go buy, you know, uh buy like all their stuff at Best Buy.
00:20:45.039 --> 00:20:47.279
But back then it was like you had to buy all of the individual components.
00:20:47.359 --> 00:20:48.000
It was extremely hard.
00:20:48.079 --> 00:20:50.960
It took someone who knew a ton about computers to be able to sell you this.
00:20:51.119 --> 00:21:09.599
And so what happened is they developed a technology, an AI, an artificial intelligence that would be able to take in your needs, and then it would be able to create an entire configuration of these insanely advanced computer systems completely on its own.
00:21:10.079 --> 00:21:11.519
And this was huge.
00:21:11.680 --> 00:21:14.480
There was huge commercial return on investment.
00:21:14.640 --> 00:21:19.680
There was Fortune 500 out of adaptation of this technology to like create these configurations for people.
00:21:19.759 --> 00:21:27.039
Again, these these fancy AI or these fancy computer systems that you know that the thing would configure uh on its own.
00:21:27.200 --> 00:21:29.680
And there's huge, huge money uh generated in this.
00:21:29.759 --> 00:21:36.160
So the Japan fifth generation product,$57 billion, and then DARPA back at it again, just like they were with the first boom.
00:21:36.319 --> 00:21:39.839
Strategic uh computing, more than you know, a billion dollars planned.
00:21:40.000 --> 00:21:42.480
So but why did the so anyway?
00:21:42.640 --> 00:21:49.759
So there's all this hype, and then once again, people are like, oh my gosh, like this AI is able to do something way better than humans can do.
00:21:49.839 --> 00:21:59.759
You know, a human per before was the expert on how we can configure like insanely complicated tasks and all this stuff, but now this this bot can do it and it can do it like super quick.
00:22:01.039 --> 00:22:06.319
This means that in 10 years, five years, everything that you do is going to be AI.
00:22:06.480 --> 00:22:08.880
These bots are going to be taking they're going to be taking your jobs.
00:22:09.359 --> 00:22:12.319
They're going to be able to replicate your expertise and take your job.
00:22:12.559 --> 00:22:14.640
But then the hype cooled down.
00:22:15.200 --> 00:22:18.880
It turns out that there's really expensive maintenance on these bots.
00:22:18.960 --> 00:22:26.400
There's a ton of data, and then you have to keep making basically like you have to keep adding program stuff as the as the technology evolves.
00:22:26.480 --> 00:22:36.079
So it's almost like as the tech evolves and the computers evolve and stuff, and so like the configurations become different, you have to write an entirely new rule book for how it makes these things.
00:22:36.160 --> 00:22:45.279
And so basically, every time you have to rewrite the rulebook rule book, at what point does rewriting the rule book overweigh the weight of just actually doing the thing yourself?
00:22:45.519 --> 00:22:54.000
And that is, I mean, again, so just think about I'm not gonna I'm not gonna hold your hand through this, but just think about the ways that this kind of thing parallels what we're seeing right now.
00:22:54.160 --> 00:23:03.599
All right, it's uh it can do a lot of stuff, it can do a lot of crazy things, but it's not, it's not quite like it's not necessarily like you know gonna take everyone's jobs in a year.
00:23:03.920 --> 00:23:07.519
But anyway, so what was taken away from the second boom in the 1980s, you ask?
00:23:07.759 --> 00:23:19.759
Well, knowledge engineering, so being able to create you know a system that is able to make a bot that can do something that's extremely complex, um, decision support, and then narrow commercial AI.
00:23:19.839 --> 00:23:24.720
So like an AI that can do something that's very niche and replicate expertise in a very specific field.
00:23:24.880 --> 00:23:25.839
So that's what happened.
00:23:26.000 --> 00:23:29.680
But then we reach another AI winter where the hype cools, okay?
00:23:29.839 --> 00:23:33.519
After another AI boom, we reach another AI lull.
00:23:33.839 --> 00:23:36.480
All right, so there's too many rules to maintain, right?
00:23:36.640 --> 00:23:38.880
You can't you can't keep feeding it all the in all these rules.
00:23:38.960 --> 00:23:40.640
You might as well just do it yourself at that point.
00:23:40.799 --> 00:23:44.160
The specialized AI hardware, it's it collapses.
00:23:44.799 --> 00:23:45.279
Collapses.
00:23:45.440 --> 00:23:46.640
There's just too much, it's too much there.
00:23:46.720 --> 00:23:49.119
I don't really know what this means to be honest, but that the that's what this says.
00:23:49.200 --> 00:23:54.000
And then uh so what happens is investors pull out again.
00:23:54.160 --> 00:23:57.279
They say, oh my gosh, AI is gonna be able to do everything.
00:23:57.519 --> 00:24:00.960
All of a sudden, uh, like you know, low key, it's like too much work.
00:24:01.039 --> 00:24:06.160
It's too much work for this, for this to to to create these things that can do something that's really niche.
00:24:06.240 --> 00:24:08.240
Like you might as well just get a person to do it.
00:24:08.799 --> 00:24:10.559
And now we are in boom three.
00:24:10.640 --> 00:24:12.240
All right, now we are in boom three.
00:24:12.319 --> 00:24:18.319
Is this the boom that is gonna take off and create the terminator or the prelude to Dune that's going to destroy the planet?
00:24:18.799 --> 00:24:19.680
Maybe.
00:24:20.079 --> 00:24:21.519
But let's just dive into this a little bit.
00:24:21.599 --> 00:24:26.400
Okay, so this boom three is the deep learning/slash generative AI.
00:24:26.640 --> 00:24:42.400
All right, the way that I would summarize this is essentially what the AI is able to now do is it can take in every image, or let's just say, like, let's you take a program, you feed in a million pictures of a cat.
00:24:42.720 --> 00:24:44.720
A million pictures of cats, all right?
00:24:44.960 --> 00:24:48.640
And then you show it a picture of a cat, and you go, what is this?
00:24:48.799 --> 00:24:56.480
And the AI is able to basically analyze all of this stuff and then cross-analyze it with all the other images and be like, this thing right here is a cat.
00:24:56.559 --> 00:24:58.160
And this is like you know what you use AI for.
00:24:58.240 --> 00:24:58.960
I used it earlier today.
00:24:59.039 --> 00:25:04.160
I put I took a picture of a trap bar and I said, How many sets of this should I do to maximize my vertical jump?
00:25:04.319 --> 00:25:05.440
And that's basically what it did.
00:25:05.519 --> 00:25:25.279
It took in a ton of pictures of weightlifting equipment all over the place, and then it found it cross-referenced all of that data, and it said, All right, this dude is lifting a trap bar with some light weight on it, and then it's cross-referencing all of the information that was pulled that's been put out there about lifting and stuff like that, and it said, All right, this is the rep range that you should do to maximize your vertical leap.
00:25:25.519 --> 00:25:28.400
And it pumps out something where I'm like, oh my gosh, it's like a person.
00:25:28.559 --> 00:25:35.200
It's like if I had a personal trainer here and I said, yo, how many reps of this should I do to maximize my vertical leap?
00:25:35.519 --> 00:25:37.119
My vertical leap.
00:25:38.240 --> 00:25:39.599
But that's not really what's happening.
00:25:39.759 --> 00:25:42.960
What's happening is all of this processing on the back end.
00:25:43.039 --> 00:25:51.359
And so key milestones here are gonna be Alpha Go beats Lee Sotal, so like the big go thing, where it's a super complicated thing, super complicated game.
00:25:51.519 --> 00:25:54.000
The idea is like you basically have to have like human intuition.
00:25:54.079 --> 00:26:01.119
There, there was a big moment in that, I will say, where it seemed like it, you know, it made a decision that seemed so irrational to like human brains that it seemed like it was being creative.
00:26:01.279 --> 00:26:03.680
Um, Chat GPT launch, obviously.
00:26:03.759 --> 00:26:12.000
So this wave is taking off because of like big data, cloud compute, insane algorithms, and then the money, of course, is ridiculous.
00:26:12.160 --> 00:26:30.559
So global generative AI private investment in 2024 alone, 33.9 billion, and then US private AI investments, so uh basically like capital, like like big industry investing in this, is 285.9 billion dollars here, all right?
00:26:30.960 --> 00:26:34.000
285.9 billion dollars here.
00:26:34.720 --> 00:26:59.200
So today's open question are the basically the question now is all of this stuff that we're getting AI to do, similar to with the boom two and even the boom one, really, is it can do all of this stuff, you know, it can replicate the expertise of a personal trainer, it can identify images and stuff, but is it going to be encompassing and broad enough that it is going to overtake the need for a human being to do it?
00:26:59.359 --> 00:27:04.000
And is it cost effective enough that that will in fact be the case?
00:27:04.319 --> 00:27:09.920
Because many of these companies are pricing for growth while spending a ton on compute.
00:27:10.000 --> 00:27:20.240
So they're like assuming that they're gonna reach this crazy point in the future, but the current point that they're at is not really that like it's like not that crazy.
00:27:20.400 --> 00:27:21.519
It's not that crazy.
00:27:21.680 --> 00:27:28.400
And again, maybe, you know, maybe they have like these ridiculously insane models that seem like you know they're gonna that they have malintent, they're gonna, you know, they're escaping all their hard work.
00:27:28.960 --> 00:27:30.640
But again, they're just really good at doing stuff.
00:27:30.720 --> 00:27:38.240
Like they're able to pull all the data of like all of the hackers who've ever put anything on the internet and compile that and be like, all right, I'm gonna hacking hugging face now, and then it does it.
00:27:38.720 --> 00:27:43.599
But is there any, you know, is there any interoception with the AI?
00:27:43.680 --> 00:27:44.720
Is it able to look inward?
00:27:44.799 --> 00:27:50.160
Is it able to be creative actually, or is it just pulling such a large data set that it's incomprehensible for our human brain?
00:27:50.240 --> 00:27:52.160
And so it seems like something that expands beyond it.
00:27:52.240 --> 00:27:56.319
Or maybe that is what consciousness, you know, maybe that is gonna, maybe that is what the AGI development is.
00:27:56.400 --> 00:27:57.119
Maybe I'm completely wrong.
00:27:57.200 --> 00:27:57.440
I don't know.
00:27:57.519 --> 00:27:58.240
I'm just giving you the facts.
00:27:58.319 --> 00:28:00.559
I'm I'm I'm telling you, I'm trying to make you feel better.
00:28:00.640 --> 00:28:04.880
All right, I'm trying to make you feel like we're not all gonna die from this technology because I genuinely don't think we are.
00:28:05.599 --> 00:28:08.240
But anyway, so that's kind of the that's kind of the picture.
00:28:08.319 --> 00:28:10.000
So we're in this boom three right now.
00:28:10.160 --> 00:28:12.240
All of this money is being pulled in.
00:28:12.799 --> 00:28:14.000
And so what happens next?
00:28:14.160 --> 00:28:21.519
Well, what I think what I think happens next is I think what happens next is AI takes over the internet.
00:28:21.839 --> 00:28:22.240
All right.
00:28:22.319 --> 00:28:28.559
I think like like Google is gonna completely change the way it is.
00:28:28.720 --> 00:28:39.920
Companies are gonna be like, you know, there's gonna be tons of data, security things, so every company's gonna have their AI, it's gonna be able to, you know, make presentations for you, it's gonna be able to do all this stuff, blah, blah, blah, blah, blah.
00:28:40.079 --> 00:28:50.960
And then I think we're gonna, you know, reach this steady state of progression where your computer is able to do more and more, and it's gonna make human lives easier and easier, or it's gonna make like if you're about that grind, um, you're it's gonna make you way more productive.
00:28:51.119 --> 00:28:57.920
It's gonna completely change like the workplace and all this stuff, but it's not gonna complete I don't think it's gonna completely replace the need for humans.
00:28:58.240 --> 00:29:04.400
And again, because also, like, I mean, even if you look at it now, and this is okay, so I'm I'm reaching more into assumptions, all right?
00:29:04.480 --> 00:29:05.440
I went over a lot of facts.
00:29:05.599 --> 00:29:09.519
I don't think it's like that, like I don't really know if there's that much data it can pull from.
00:29:09.680 --> 00:29:10.319
You know what I mean?
00:29:10.480 --> 00:29:19.440
Like, it has access, presumably, like like like Grox AI probably has access to every single video that's ever been posted to Twitter, and the AI videos are still kind of, you know, they're crazy good.
00:29:19.519 --> 00:29:20.640
They're crazy good.
00:29:20.960 --> 00:29:25.359
Um, but they still don't have like that, like I don't know.
00:29:25.440 --> 00:29:26.960
Honestly, I think the videos are gonna get better.
00:29:27.039 --> 00:29:27.759
I I I don't really know.
00:29:27.839 --> 00:29:29.680
I it's gonna completely change the scope of everything.
00:29:29.839 --> 00:29:31.200
Do I think it's gonna kill humanity?
00:29:31.440 --> 00:29:32.000
No, I don't.
00:29:32.079 --> 00:29:33.440
I don't think I don't think it's gonna kill humanity.
00:29:33.519 --> 00:29:36.799
I think it's gonna completely change the scope of work, the kind of things people do.
00:29:36.880 --> 00:29:40.720
I think AI driving and stuff is gonna be all over the place.
00:29:40.799 --> 00:29:46.240
I think when you come back, you're just gonna talk to your house, which people are already basically doing, and tell it what you want to watch.
00:29:46.480 --> 00:29:47.759
I'm gonna be able to do that for you.
00:29:47.920 --> 00:29:50.480
You're gonna like you know, it's gonna be different connections all over the place.
00:29:50.559 --> 00:30:03.519
It's gonna be like this network of artificial intelligence, I think, that like maybe you can like tap into and access from any point and use that as like you know, like combined like Discord with Google, with gaming, with like all this.
00:30:03.759 --> 00:30:04.400
I I don't know.
00:30:04.480 --> 00:30:04.799
I don't know.
00:30:04.880 --> 00:30:10.160
I think it's but I do think it's gonna be I think it's gonna be largely positive outcomes.
00:30:10.240 --> 00:30:12.319
And I I do think that it's gonna cool off a little bit.
00:30:12.400 --> 00:30:12.960
I'm not gonna lie.
00:30:13.039 --> 00:30:28.799
I do think in like five years we're gonna look back at this time and be like, whoa, this was crazy, but it was a little bit unsustainable to growth, and AI does have insane capabilities, but it's not gonna destroy the entire world, and it's it's um, you know, it changed the world, right?
00:30:28.880 --> 00:30:33.119
It was like the industrial revolution changed the way people drive, changed the way blah, blah, blah, blah.
00:30:34.319 --> 00:30:40.160
And then maybe there'll be a another another deep wave or another boom in a couple years, and that'll be crazy.
00:30:40.240 --> 00:30:41.200
But that's that's what I think.
00:30:41.359 --> 00:30:44.160
I think it's gonna change everything, but I don't think it's gonna kill everyone.
00:30:44.319 --> 00:30:48.400
I don't think, I don't think I don't I don't have any reason to believe why I would do anything like that.
00:30:48.480 --> 00:30:49.519
I think this is all a lot of hype.
00:30:49.599 --> 00:30:51.359
I think it's borderline a marketing ploy.
00:30:51.519 --> 00:30:53.119
I think it's borderline a marketing scheme.
00:30:53.200 --> 00:30:56.079
I mean, think about the exposure that these companies have gotten from.
00:30:56.160 --> 00:31:07.519
I mean, do you like if I was if I had an AI startup and I saw all this stuff, I'd be like, oh my gosh, I'm just gonna go on the internet and say that my AI is gonna be the one that's about to destroy the world because that's apparently what gets all the views.
00:31:07.599 --> 00:31:09.119
You know, people fear mongering.
00:31:09.359 --> 00:31:10.720
Tales old as time.
00:31:11.200 --> 00:31:12.480
Anyway, you get the point.
00:31:12.640 --> 00:31:13.359
Thanks for listening.
00:31:13.440 --> 00:31:14.240
I'd love to hear your thoughts.
00:31:14.400 --> 00:31:14.799
Email me.
00:31:14.960 --> 00:31:16.400
Emails in bio and all that stuff.
00:31:16.480 --> 00:31:17.119
Have a going bye.
