Will AI Replace Your Job? What the Research Actually Says [MM280]

WEBVTT
00:00:00.000 --> 00:00:08.580
So if you're anxious that you're not technical enough or you're starting too late, the data says your anxiety is pointed at exactly the wrong target.
00:00:17.579 --> 00:00:19.219
AI won't replace you.
00:00:19.339 --> 00:00:21.239
The people who know how to use it will.
00:00:21.699 --> 00:00:22.800
This is everywhere.
00:00:22.960 --> 00:00:23.660
It's on posters.
00:00:23.899 --> 00:00:24.980
It's on LinkedIn posts.
00:00:25.420 --> 00:00:27.800
Nobody can even agree who said this first.
00:00:27.920 --> 00:00:32.960
And normally, when a line turns into that much of a cliche, it stops meaning anything.
00:00:33.520 --> 00:00:35.240
So I went and actually checked this out.
00:00:35.299 --> 00:00:36.479
I did some real research.
00:00:37.100 --> 00:00:38.359
I read real economists.
00:00:38.359 --> 00:00:44.439
And what they had to say, I looked at the kind of data you don't actually see attached to a line that's overused like this.
00:00:44.820 --> 00:00:48.740
And it turns out to be true, but not for the reasons you'd guessed.
00:00:48.939 --> 00:00:52.179
And it's better news than the quote even promises.
00:00:52.700 --> 00:00:57.560
So first of all, nobody can even agree on who actually said that AI won't replace you.
00:00:57.920 --> 00:01:01.140
The people who know how to use AI will.
00:01:01.140 --> 00:01:10.099
The earliest cleanly dated version that I could cite was from Santiago, who's a computer science and a machine learning engineer.
00:01:10.379 --> 00:01:17.620
He posted that AI will not replace you, that a person using AI will on X in January of 2023.
00:01:18.299 --> 00:01:26.319
Andrew Ng, he's a Stanford professor, of course, Google Brain co-founder, has separately said close to the same thing a couple of times.
00:01:26.439 --> 00:01:28.540
I don't think it really matters who said it first.
00:01:28.739 --> 00:01:31.760
What matters is whether or not it's true and why.
00:01:31.920 --> 00:01:34.260
And that's the story of this video.
00:01:35.519 --> 00:01:46.040
So to kind of put this in perspective, I thought I'd tell a quick story that explains more about AI and jobs than most of these hot takes do that you hear about on the internet.
00:01:46.299 --> 00:01:57.319
When automatic teller machines showed up on the scene, ATMs as we call them here in the United States, everybody assumed that bank tellers were finished.
00:01:57.480 --> 00:01:59.760
I mean, this, I was around for that.
00:02:00.000 --> 00:02:10.879
But instead, teller jobs roughly doubled over the next 40 years because cheaper branches meant banks opened way more branches.
00:02:11.180 --> 00:02:12.860
You can see this in my neighborhood.
00:02:13.060 --> 00:02:18.599
There are three or four Chase bank branches and many bank branches of other varieties.
00:02:18.599 --> 00:02:26.280
In 1985, there were about 60,000 ATMs and 485,000 bank tellers.
00:02:26.639 --> 00:02:33.300
By the mid-2000s, teller employment had grown to roughly half a million, so that's up.
00:02:33.719 --> 00:02:39.780
And employment doubled from a quarter of a million to almost half a million over 45 years.
00:02:40.280 --> 00:02:43.240
So ATMs didn't really have an impact on employment.
00:02:43.800 --> 00:02:46.300
That's the version that everyone tells.
00:02:46.300 --> 00:02:49.360
But here's the part that almost nobody includes.
00:02:49.939 --> 00:02:53.780
Teller jobs have been crashing since around 2010.
00:02:54.560 --> 00:03:00.900
Not because of ATMs, but because of mobile banking, which didn't just automate a task.
00:03:01.020 --> 00:03:04.919
It removed the reason to walk into the branch at all.
00:03:05.259 --> 00:03:14.340
And the Bureau of Labor Statistics is currently predicting a 13% decline in bank tellers by 2035.
00:03:14.340 --> 00:03:16.580
So that's the real lesson.
00:03:17.159 --> 00:03:20.860
Automating a task inside your job, you're usually fine.
00:03:21.240 --> 00:03:27.900
Something that removes the real reason anyone needs you at all, that's the one to watch out for.
00:03:27.900 --> 00:03:34.979
And here's another real personal example that I don't have great statistics on, but it's an interesting example.
00:03:35.439 --> 00:03:42.039
My son was a high school baseball player, and we traveled all around Texas going to high school baseball tournaments.
00:03:42.580 --> 00:03:44.879
And oftentimes we were eating drive-thru food.
00:03:45.020 --> 00:03:47.400
And one time we pulled up to a Panda Express.
00:03:48.180 --> 00:03:52.319
He loved that orange chicken, and there's a lot of calories in there, which is great.
00:03:52.319 --> 00:03:58.099
If you're burning calories, throwing 90-mile-an-hour fastballs every day, you need those calories.
00:03:58.500 --> 00:04:01.860
And Panda Express delivers in the calorie department.
00:04:02.319 --> 00:04:08.620
When we rolled up to the drive-thru, we noticed right away that AI was running the drive-thru.
00:04:09.000 --> 00:04:10.560
It was voice recognition.
00:04:11.120 --> 00:04:13.080
It was an automated screen.
00:04:13.560 --> 00:04:19.920
And we were able to give our order to the AI, and there were a lot of great benefits for us from that.
00:04:19.920 --> 00:04:29.100
And when it did make mistakes, they were easily corrected because the system had been designed so that I could see my order on the screen as I delivered it.
00:04:29.120 --> 00:04:32.459
So it was a very good user experience.
00:04:32.459 --> 00:04:44.319
And as a company experience, I imagine it was also pretty cool because the company didn't need to worry about whether or not that 17-year-old kid was going to show up at that drive-thru.
00:04:44.620 --> 00:04:47.959
Does that mean that there weren't still people working at the Panda Express?
00:04:48.620 --> 00:04:49.100
Well, no.
00:04:49.300 --> 00:04:51.620
When we got to the drive-thru, there was a person there.
00:04:51.860 --> 00:04:54.100
There were people in there cooking the food and so on.
00:04:54.339 --> 00:04:56.360
So that job wasn't eliminated.
00:04:56.560 --> 00:05:01.819
It was shifted and changed, but there were still people working at the Panda Express.
00:05:02.120 --> 00:05:08.100
Were there fewer people, and did that get passed down to the customer as a better price?
00:05:08.180 --> 00:05:09.000
I'm not sure.
00:05:09.379 --> 00:05:11.879
Again, it's kind of like the ATM example.
00:05:12.620 --> 00:05:18.920
Just because I automated the taking of the order, I didn't stop the reason for going to Panda Express.
00:05:19.360 --> 00:05:25.420
The reason for going to Panda Express was calories needed to throw fastballs, and that still existed.
00:05:25.959 --> 00:05:46.740
And I really firmly do believe that the kids who were leaning into that and helping the company implement those systems and learning how to run those things and run that business, those are the ones that got promoted to shift leaders, assistant managers, and so forth, because they knew how to use the AI to help the management team run the business.
00:05:47.519 --> 00:05:57.139
So yeah, this is my opinion, but I also went and checked the actual research on AI and jobs across four completely different fields.
00:05:57.300 --> 00:06:02.160
Customer support, consulting, writing, and software engineering.
00:06:02.379 --> 00:06:03.819
Same results every single time.
00:06:04.279 --> 00:06:10.180
The biggest gains from AI don't go to the people who are already the best at their job.
00:06:10.180 --> 00:06:13.500
They go to the people who are the furthest behind.
00:06:13.639 --> 00:06:20.279
In one study, novices improved roughly twice to four times as much as experts did.
00:06:20.720 --> 00:06:29.420
So if you're anxious that you're not technical enough or you're starting too late, the data says your anxiety is pointed at exactly the wrong target.
00:06:30.199 --> 00:06:38.019
Now, here's an interesting thing to watch out for as you decide whether or not AI is going to replace you and your job.
00:06:38.199 --> 00:06:43.980
In a previous video, I talked about how you should treat AI as a really smart intern.
00:06:43.980 --> 00:06:48.899
In a BCG consulting study, they took two sets of employees.
00:06:49.040 --> 00:06:55.180
One set of employees had AI available to them, and the other set of employees did not.
00:06:55.579 --> 00:07:13.199
And the study found that the users with AI did 19 points worse than the people without AI, here's the catch, on a task that was deliberately designed to do things that AI was not good at.
00:07:13.199 --> 00:07:19.060
The AI-assisted group confidently trusted the wrong output.
00:07:19.600 --> 00:07:22.160
So this is the same lesson as that video.
00:07:22.300 --> 00:07:23.800
It's the Ronald Reagan lesson.
00:07:24.180 --> 00:07:25.439
Trust but verify.
00:07:26.439 --> 00:07:30.120
The maxim is, dovii, no provii.
00:07:30.519 --> 00:07:32.120
Trust but verify.
00:07:33.579 --> 00:07:49.480
As you're working to avoid being replaced by AI and trying to become the person who knows how to use AI, make sure that you're verifying AI so that you're a net benefit to the company and producing more work that's actually valuable.
00:07:50.540 --> 00:07:55.939
So if you want to become this person that won't get replaced by AI, what are your moves?
00:07:55.939 --> 00:08:00.300
Here are three concrete, helpful moves that you can make today.
00:08:00.800 --> 00:08:06.060
Number one, pick one task this week that you would normally avoid or that you normally do badly.
00:08:06.220 --> 00:08:08.899
Remember, don't pick the thing you're already good at.
00:08:09.279 --> 00:08:13.420
That's where AI can help you the least and where your payoff is the smallest.
00:08:13.939 --> 00:08:16.879
Use the intern framework in my previous video.
00:08:17.040 --> 00:08:18.560
I'll link to it down in the comments.
00:08:18.980 --> 00:08:24.240
And give AI real context on that specific task and learn how to use it.
00:08:24.579 --> 00:08:27.819
Become a person who knows how to use AI.
00:08:27.819 --> 00:08:32.519
And don't wait to become technical enough before you get started.
00:08:33.120 --> 00:08:36.039
The data says that that instinct is backwards.
00:08:36.799 --> 00:08:40.480
Novices gain the most from AI, not experts.
00:08:40.659 --> 00:08:45.320
If you're not sure how to do the thing you want to do, ask AI how to do it.
00:08:46.159 --> 00:08:49.000
Here's an honest complication, and I think it's really good news.
00:08:49.240 --> 00:09:02.039
A rigorous study using real government pay data found no measurable wage bump yet for people using AI, and that's two years in to this AI era that we're in now.
00:09:02.039 --> 00:09:16.059
But at the same time, exactly the same time, employers are already paying 56% to 62% more for people with AI skills in job postings that are out there right now.
00:09:16.320 --> 00:09:26.700
So the market is already pricing the AI capability stuff in, even before the research has even caught up with this to prove why.
00:09:27.100 --> 00:09:28.840
That's not a reason to wait.
00:09:29.159 --> 00:09:35.700
That's the description of a window that's still open for you to start leveling up your AI skills right now, today.
00:09:45.399 --> 00:09:49.620
Hey, thanks for listening to another episode of The Mark Mason Show.
00:09:49.840 --> 00:09:58.200
For more content like this, hit subscribe or head on over to markmason.com, hook up with us on social media, or leave feedback.
00:09:58.879 --> 00:10:00.960
Until next time, ciao.