WEBVTT
00:01:01.340 --> 00:01:04.299
The following is a site to practice production.
00:01:14.460 --> 00:01:22.780
Hi there and welcome to our very first inaugural podcast of Beyond the Bot, AI and Mental Health.
00:01:22.939 --> 00:01:24.460
I'm Dan Florel.
00:01:24.540 --> 00:01:34.460
I'm a professor over at Eastern Kentucky University and CEO of MindSci, in addition to uh the vice president of Psych to Practice.
00:01:34.620 --> 00:01:36.379
And with me today is Adam.
00:01:36.539 --> 00:01:38.300
Adam, why don't you introduce yourself?
00:01:38.859 --> 00:01:43.740
Adam Lockwood, associate professor of school psychology at Kent State University.
00:01:43.819 --> 00:01:49.180
And I have a consulting business called Lockwood Educational Educational and Psychology.
00:01:49.500 --> 00:01:51.340
It's good to get your own play here, Adam.
00:01:51.659 --> 00:01:55.819
Lockwood Educational and Psychological Psychological Consulting.
00:01:56.300 --> 00:02:04.219
So for all of you out there, one of the first lessons to learn is don't make your own business a huge mouthful to say when you're first getting there.
00:02:04.539 --> 00:02:05.900
So what is the focus?
00:02:06.060 --> 00:02:19.099
As you can tell, the focus of our podcast, and this is an initial episode, is to take about 30 minutes of your day every couple weeks to chat about AI and its influence in mental health.
00:02:19.500 --> 00:02:22.219
It's just going to be the two of us, Adam and I.
00:02:22.460 --> 00:02:38.140
We've had quite a few years of experience in AI, as much as anybody can really talk about it when it first burst on the scene in 2020 with the initial release of Chat GPT, though there was, of course, things going behind the scenes for quite a bit longer than that.
00:02:38.300 --> 00:02:51.180
And both of us have spent a fair amount of time and do quite a few trainings looking at the ethics of AI use and also how it might be applied within mental health settings, specifically more school psychology related.
00:02:51.419 --> 00:03:06.860
But we're going to start off with our inaugural issue looking at the promises and perils of AI agents, which, if you're just understanding initially what AI is, this is a whole other level.
00:03:07.020 --> 00:03:14.060
And I'm going to defer to Adam because Adam has many of his little AI bot minions working for him.
00:03:14.299 --> 00:03:27.099
And I know about them, uh, but am slightly intimidated by uh these little independent, autonomous things going on and skittering about and doing things uh that I'm not quite sure what they're doing.
00:03:27.259 --> 00:03:36.139
So, Adam, I'm going to be somewhat of a learner here, and you can kind of tout some of your expertise as far as how we might be able to use it within mental health.
00:03:36.219 --> 00:03:41.579
And of course, we also named it perils, so that there will be some drawbacks, I'm sure.
00:03:41.819 --> 00:03:43.659
But uh turn it over to you.
00:03:44.299 --> 00:03:44.859
Sure.
00:03:45.019 --> 00:03:50.060
I think first it's a good idea to differentiate between maybe a chatbot and an agent.
00:03:50.379 --> 00:03:50.859
Yeah.
00:03:51.099 --> 00:03:57.019
So I mean, like chatbots are basically uh ones that everybody's familiar with, right?
00:03:57.099 --> 00:04:02.620
You get your little phone out, you open up the app, and you start talking like it's a real person.
00:04:02.939 --> 00:04:09.019
And it goes ahead and responds and has all its little tendencies, like making you feel like you're a brilliant person in the world.
00:04:09.180 --> 00:04:21.340
And then agents, of course, are where you're going to go ahead and set it missions, and it will accomplish those missions any way it sees fit to get the end result that you are specifying for it.
00:04:21.500 --> 00:04:28.220
And I imagine the more guardrails and specifics, probably the better return you get on those agents.
00:04:28.379 --> 00:04:32.699
But that's just the off the top of my head, kind of what the difference between the two would be.
00:04:32.939 --> 00:04:39.259
Yeah, well, first of all, I like you said, you talk to the chatbot, because at this point I don't type.
00:04:39.500 --> 00:04:40.699
Do you type still?
00:04:41.180 --> 00:04:42.699
I'm still a typer, you know.
00:04:42.860 --> 00:04:44.699
I'm even a clumsy texter.
00:04:44.780 --> 00:04:45.500
You know what it is?
00:04:45.579 --> 00:04:54.220
I just, especially when I'm out and about, I'm a little uh conscientious about talking to my machine uh out there.
00:04:54.379 --> 00:04:56.460
So I haven't really embraced that yet.
00:04:56.620 --> 00:04:59.980
I need to, because that's actually the way that I usually communicate.
00:05:00.139 --> 00:05:05.740
Like when I know I have decent ideas, I talk them through, but I usually find colleagues to do that with.
00:05:05.819 --> 00:05:11.340
And I guess uh, you know, if I don't find colleagues, this will be a nice little option here.
00:05:11.819 --> 00:05:18.540
But in the back of my mind, playing that sometimes it gets a little too enthusiastic about my own ideas and doesn't push back enough.
00:05:18.780 --> 00:05:22.460
So I take it though, you're definitely into the verbal descriptions.
00:05:22.540 --> 00:05:28.620
And I think the couple times we've hung out, you know, online and to talk about things, I see how you work with that.
00:05:28.860 --> 00:05:36.860
Any advice on it, or is it just so natural that you really don't have to think about learning yet another skill and ways of framing things?
00:05:36.939 --> 00:05:39.020
Kind of I'm thinking like prompting, right?
00:05:39.180 --> 00:05:45.020
Prompting, if the more specific and the more context-driven, usually the better results you get.
00:05:45.260 --> 00:05:47.340
Well, I don't think people need to learn how to talk.
00:05:47.500 --> 00:05:49.420
Most people know how to talk already.
00:05:50.140 --> 00:05:52.700
So just talk, ramble.
00:05:52.860 --> 00:06:04.460
The truth is the earlier, the early research that's coming out is you get better output if you talk, because it's through this kind of rambling what you say, maybe not.
00:06:04.700 --> 00:06:05.820
Well, let me think.
00:06:06.220 --> 00:06:16.460
It gives the AI some insight into kind of the inner workings that you you have going on that you normally would self-censor by typing.
00:06:16.939 --> 00:06:20.060
And so by doing that, you actually get better output.
00:06:20.300 --> 00:06:21.900
Not only that, it's just faster.
00:06:21.980 --> 00:06:25.180
Um, and it you know can produce carpal tunnel.
00:06:25.660 --> 00:06:32.140
Yeah, well, you know, I think I know why I have now that we've been discussing it, why I have this mental block about it.
00:06:32.379 --> 00:06:47.500
It's because way back when I got into dictation software a bit, and where you go stop, and you know, for a period, and you had to be pretty precise because if you put in your ums and everything, it messed up what it was transcribing.
00:06:47.660 --> 00:07:06.060
And this obviously, this technology doesn't do that, but I have this prior technology mental block that I think is causing some of the issues with me doing the interaction with AI because that previous learning is telling me like, no, you have to be precise, otherwise, it's gonna mess it up.
00:07:06.540 --> 00:07:08.780
Don't worry about being precise.
00:07:09.180 --> 00:07:09.820
Gotcha.
00:07:09.900 --> 00:07:11.100
I'm checking that off.
00:07:11.340 --> 00:07:18.700
What I'll say about agents, how I use them, is they're essentially chatbots that have arms and legs.
00:07:18.860 --> 00:07:36.860
And so, as normally, if you're using ChatGPT, for example, you might be like, Well, I want to create a blog and it will give you information, and you go back and forth, and then you take that information and you might paste it into your blog.
00:07:37.100 --> 00:07:39.500
In this case, it works autonomously.
00:07:39.580 --> 00:07:45.420
So I might have that back and forth and say, okay, now great, go to my website, update my language.
00:07:45.740 --> 00:07:49.420
It just takes the human out of the loop so much.
00:07:49.580 --> 00:07:52.379
They they they refer to it as the sandwich or the burger.
00:07:52.540 --> 00:07:57.980
You have human on the front end, so we'll have a back and forth with AI, figuring out what exactly we want to do.
00:07:58.140 --> 00:08:04.860
Once we have a plan, you send it out to go do it, and then at the end you check and you you act as a human in the loop.
00:08:04.939 --> 00:08:10.460
But all that cutting and pacing, going back and forth, is eliminated.
00:08:10.700 --> 00:08:19.340
All of the busy work, all of the kind of boring stuff that you don't want to do can be outsourced at this point.
00:08:20.300 --> 00:08:37.179
So when you say that, what comes to mind with me is that you are giving access to this AI, to things that require your passwords and you know, your different basically personal information that I think would give some people pause.
00:08:37.340 --> 00:08:44.460
I mean, is this all retained that you I mean, have you looked up and see like how is this retained within Chat GPT?
00:08:44.700 --> 00:08:58.059
Um, you know, like what safeguards are there that it doesn't just go gonzo one day and all of a sudden you have 50 appointments with a massage therapist, you know, in California or something.
00:08:58.299 --> 00:09:04.299
Well, you got to be careful because the truth is agents have gone and deleted files that were very important.
00:09:04.539 --> 00:09:07.100
So you need to make sure that you sandbox.
00:09:07.340 --> 00:09:13.340
So the first thing you do is you start out and give it access to an email account that is um burning.
00:09:15.019 --> 00:09:15.659
Yeah.
00:09:15.899 --> 00:09:16.220
Yeah.
00:09:16.460 --> 00:09:20.220
So I'll give you a use of one of my, I'll give you an example of a use case.
00:09:20.460 --> 00:09:25.340
I get a lot of Google Scholar alerts that are just AI articles.
00:09:25.500 --> 00:09:27.580
I get six emails a day.
00:09:27.899 --> 00:09:33.740
And within those, about every day it works out to about 48 different articles.
00:09:34.060 --> 00:09:37.180
I have an email account that's just a burner account.
00:09:37.259 --> 00:09:43.340
I use it for nothing but that, and I let AI have complete read access to it.
00:09:43.500 --> 00:09:47.259
It can't send anything out from there, but it can read those.
00:09:47.420 --> 00:09:49.899
I definitely wouldn't want my bank statements going there.
00:09:50.060 --> 00:09:54.379
I certainly wouldn't want patient or student information to go in there.
00:09:54.779 --> 00:09:57.580
But that's that's one way.
00:09:57.740 --> 00:10:03.740
And then I give it access to folders, but those folders are sandboxed.
00:10:03.820 --> 00:10:13.820
So it has its own folder within documents that it can go, write files, edit files, delete files, but it cannot leave that folder.
00:10:14.220 --> 00:10:21.420
So when you talk about that folder, is that like for me, I have Dropbox, and let's say I wanted to give you access to a folder.
00:10:21.659 --> 00:10:29.820
I could request and say, hey, give this access, but that doesn't mean you have carte blanche in my whole Dropbox, you know, account.
00:10:29.980 --> 00:10:32.060
You just have the access to one folder.
00:10:32.220 --> 00:10:41.899
And I guess, you know, this is where I need to do a little more research, but like to get the agents going, is there a particular agent button for our common like Google Gemini?
00:10:42.379 --> 00:10:44.620
I know you've been using Chet or Claude.
00:10:44.779 --> 00:10:49.740
I mean, is there something that's labeled agent on the side and you click on that and start setting things up?
00:10:49.899 --> 00:10:54.620
I mean, where would people get a better idea of how to even try this out?
00:10:54.860 --> 00:10:56.620
Because I I wouldn't even know.
00:10:56.860 --> 00:11:00.139
You know, I guess I could ask Chet, like, how do I use your agent?
00:11:00.220 --> 00:11:05.500
But yeah, I guess I should say that I'm using agent as a fairly loose term.
00:11:05.659 --> 00:11:08.220
I think this is more like agentic AI.
00:11:08.300 --> 00:11:15.500
That means it has some autonomy and can work on its own, can go to achieve goals with feedback from me.
00:11:16.060 --> 00:11:19.899
The two there, and there's a wide spectrum.
00:11:20.060 --> 00:11:55.550
You have like open cloud where people are running their own open weight or open source.
00:11:55.629 --> 00:12:04.750
It's just models that they're running on their own machine, or they might be using indirect access to other models.
00:12:05.389 --> 00:12:15.550
But the easiest entry point would be either codecs, which is open AI, so it's essentially Chat GPT, or Claude Anthropic Cowork.
00:12:16.110 --> 00:12:16.190
Okay.
00:12:16.750 --> 00:12:35.389
But I know mostly Codex because if you've ever used CLOD, you know that the the use limits are pretty stringent with tokens, just meaning they give you a set amount of use of the product, and when you use too much, it will say, hey, you're done for three hours.
00:12:35.550 --> 00:12:39.389
Actually, Codex will do that too, but it's much more generous.
00:12:40.030 --> 00:12:52.670
And you say this, is this where people would need paid accounts regardless, and they still have the token limits, or is this the free accounts that you uh run against the token limits more typically?
00:12:53.389 --> 00:12:56.110
You should have a paid account, just full stop.
00:12:56.509 --> 00:13:04.190
You get access to better models, it's$20 a month, which is an extremely good deal.
00:13:04.430 --> 00:13:09.870
Right now, this whole industry is being subsidized by the companies.
00:13:10.030 --> 00:13:10.269
Right.
00:13:10.509 --> 00:13:15.790
So with a$20 account, you can get hundreds of dollars worth of compute use.
00:13:16.030 --> 00:13:20.670
With a$200 account, I've heard estimates as high as like$8,000.
00:13:21.150 --> 00:13:27.070
So they're losing a lot of money right now because it partially because they want to get your use.
00:13:27.230 --> 00:13:28.990
They want to get you as a customer.
00:13:29.310 --> 00:13:31.629
But maybe that$20, it's worth it.
00:13:31.790 --> 00:13:34.670
I don't know if the free accounts of these have access.
00:13:34.830 --> 00:13:35.550
I doubt it.
00:13:35.710 --> 00:13:37.790
I don't know because I have paid accounts.
00:13:37.950 --> 00:13:42.830
Um, I could Google it or I could ask AI right now, but pay for the$20 for one of them.
00:13:42.910 --> 00:13:45.710
And and my guess would be ChatGPT.
00:13:46.030 --> 00:13:53.790
You have to, of course, remember that it's not HIPAA or FERPA compliant unless you have some sort of enterprise account.
00:13:54.030 --> 00:14:04.190
And you sure of that through someone at your company or organization or district, but or ChatGPT for teachers, you can have a FERPA.
00:14:04.990 --> 00:14:05.230
Right.
00:14:05.389 --> 00:14:16.750
For the like Google, Google Gemini, uh, if you have an enterprise through that, you should be having some FERPA compliance in regards to any of the things you use within the school's account.
00:14:16.990 --> 00:14:22.190
Just because you work for a school and you set up a free account or something, that doesn't apply, right?
00:14:22.269 --> 00:14:24.830
It has to be within that email account.
00:14:25.150 --> 00:14:44.750
So just pivoting a little bit, how would someone use an agent or agentic AI within kind of mental health or within the school kind of setting, knowing that institutions may be a little reluctant to have these agentics running around doing autonomous things?
00:14:45.470 --> 00:14:49.870
I wouldn't start by using it for healthcare or for school services.
00:14:49.950 --> 00:14:56.990
I would start by finding other repetitive time sucks in your life that you you want to automate.
00:14:57.310 --> 00:15:04.350
And so in your case, I know it was uh really looking up research, which resonates with me because I am a research hound.
00:15:04.430 --> 00:15:15.389
I mean, I'm always downloading articles and putting them away and you know, throwing them through Notebook LM, uh, which I think you know could also be used for mental health, obviously.
00:15:15.550 --> 00:15:22.750
So think about all of the areas, even though Adam and I are very techie oriented and use it a lot for AI.
00:15:22.990 --> 00:15:34.910
You can imagine uh doing it for, say, reading comprehension interventions and coming up with the latest ways of tracking that or the novel ways to use curriculum-based measurement in a particular age group.
00:15:35.070 --> 00:15:59.070
You know, whatever setting that you're working in, one of the things that I think we have challenges with as practitioners is that continuing ed aspect and where we're advancing our knowledge to be the most cutting edge, hopefully, in whatever area that we are providing services in, because that's going to help our students or our clients.
00:15:59.389 --> 00:16:02.190
And it doesn't need to be where you're spending a lot of money.
00:16:02.269 --> 00:16:18.350
For that 20 bucks uh for one of these, you know, big clawed Chat GPT, you can get and be more well informed from a scientific journal kind of standpoint than I think you could ever when being in practice before.
00:16:18.509 --> 00:16:27.950
Like you don't have to be affiliated with a university anymore because Google Scholar and Google Scholar Labs has access to a tremendous range of journals.
00:16:28.110 --> 00:16:34.509
It's not comprehensive, but it's a heck of a lot more than, say, 10 years ago, most practitioners have.
00:16:34.590 --> 00:16:50.590
And I think if you use agentic AI to help boil it down, you give it some structure, set up a dummy email like Adam was talking about, I think it really can get you up to speed on stuff that will have a real impact on the way that you practice.
00:16:51.070 --> 00:16:57.550
When you say, I mean, that seems like a perfect use case because nobody's really worried about a journal article somehow getting loose.
00:16:57.629 --> 00:16:59.629
Um, you know, it's open source anyway.
00:16:59.710 --> 00:17:00.750
It already has it.
00:17:00.830 --> 00:17:03.150
The AI already has that information.
00:17:03.389 --> 00:17:06.670
And yet we can easily apply it to our clientele.
00:17:07.470 --> 00:17:07.869
Yeah.
00:17:08.029 --> 00:17:13.549
And I think it's not going to be too long before we're going to have HIPAA compliant agents.
00:17:13.710 --> 00:17:19.389
So getting an understanding of how this all works will just help you to be ahead of the curve.
00:17:24.349 --> 00:17:32.589
Right, as opposed to trying to go from crawl to run right away, which can be very messy if you've ever seen a child that tries to go from crawl to run.
00:17:32.910 --> 00:17:35.469
It's just a disaster, face plants all around.
00:17:35.629 --> 00:17:39.149
We don't want that, the equivalent when we're using AI.
00:17:39.389 --> 00:17:43.869
And I think already, you know, we've got a sense there's a wide range out there.
00:17:44.029 --> 00:17:52.349
And one of the things Adam and I are going to run into is we're going to get a little too tech wonky sometimes because we've been living in this space for so long.
00:17:52.509 --> 00:17:57.789
And some of the shortcuts and things that he tells me and I tell him makes perfect sense to us.
00:17:58.029 --> 00:18:04.989
But as listeners, you know, it may, you may have a little bit more time to get up to speed.
00:18:05.069 --> 00:18:12.909
So we'll try to, I'll try to put the brakes on every now and then so we can explain a little further detail about what exactly we mean with some of this stuff.
00:18:13.069 --> 00:18:17.869
Because we didn't really pick the easiest topic to introduce people to our podcast.
00:18:18.029 --> 00:18:28.349
I think this is probably one of the most cutting-edge aspects from a general public, anyway, that most general public could utilize are these agentic AIs.
00:18:28.509 --> 00:18:32.589
I'm not talking about the programmers and all that, you know, that's really high in the weeds.
00:18:32.829 --> 00:18:40.909
But uh, would you agree, or would you say there's some other stuff out there that people need to be kind of aware of, even if they haven't quite used them yet?
00:18:41.149 --> 00:18:45.069
Well, I was thinking when you said that, yeah, maybe this isn't the best topic for the first one.
00:18:45.469 --> 00:18:47.309
But so yeah, I don't know.
00:18:47.469 --> 00:18:56.669
Uh I can think of a couple of use cases for one for school psychologists and one for uh for people in private practice right off the bat.
00:18:56.909 --> 00:18:57.469
Yeah.
00:18:57.709 --> 00:19:03.389
So one thing that school psychologists should be doing, and they don't do as much because they don't have time, is program evaluation.
00:19:04.909 --> 00:19:12.349
AI, create a survey with me, upload the survey to Google Forms, whatever.
00:19:12.749 --> 00:19:17.309
All essentially with no work on my side.
00:19:17.869 --> 00:19:18.589
Very little.
00:19:18.749 --> 00:19:23.389
I mean, let's say you want to be careful, you don't want to give access to your Google Forms.
00:19:23.709 --> 00:19:28.109
I can understand that, but it could still generate the document that you could upload.
00:19:28.349 --> 00:19:30.429
But at this point, go ahead.
00:19:30.909 --> 00:19:37.949
I was gonna say with the Google forms, would that be something that a say school psychologist could do within the enterprise suite?
00:19:38.109 --> 00:19:54.429
Like, does Google, because that's not one you've mentioned, does Google have some agent equivalency that we talked about with both Claude and ChatGPT that people in the school might be able to utilize um for doing something like that uh in a survey?
00:19:54.989 --> 00:19:59.789
It's funny, Google has um kind of fallen behind in 2026.
00:20:00.669 --> 00:20:05.789
I thought ChatGPT won the race for 23, 24.
00:20:05.949 --> 00:20:08.669
I thought 2025 was Google's year.
00:20:08.909 --> 00:20:17.149
I think that the first part of 2026 was Cloud, and now I think ChatGPT if you're looking at just the big models.
00:20:17.869 --> 00:20:26.509
But yeah, there are agent platforms for um Gemini, but I'm not seeing their use very often.
00:20:26.669 --> 00:20:32.909
I think it's I think it's probably more within enterprises where they're being built.
00:20:40.749 --> 00:21:01.949
Well, I was gonna say what I've been hearing lately is, and these are broad stereotypes, but the most consumer-focused AI, and by that I mean like general public, is more Gemini, like a lot of tools are coming out for it, that they're being very like that's the primary, what they're chasing.
00:21:02.189 --> 00:21:16.269
Whereas Claude is chasing more a business environment, and chat is starting to move a little bit more towards that business versus consumers, you know, general consumers or the public.
00:21:16.509 --> 00:21:24.189
And you can see that, like, you know, they had their little video creation snippet thing on ChatGPT that they decided to jettison.
00:21:24.829 --> 00:21:33.229
You know, I also have seen it in ChatGPT decided to go ahead and lessen where you could find learn mode.
00:21:33.389 --> 00:21:37.229
It made a big deal about how you could learn tutoring, you know, through it.
00:21:37.309 --> 00:21:40.669
And now it's still there, but I noticed they've buried it.
00:21:40.749 --> 00:21:46.589
It's not even on like the more button and then the more button, and you have to like actually ask, like, what happened to it?
00:21:46.669 --> 00:21:49.149
And they're like, well, we didn't want to feature it anymore.
00:21:49.229 --> 00:21:50.589
It still has that capability.
00:21:50.749 --> 00:22:10.589
Well, what it tells me is that they're de-emphasizing the usage for, say, a student population or a more public population, they're shifting to that more programming and like this maybe agentic view compared to say Gemini, where you can still feel see like learn mode is pretty prominently displayed.
00:22:10.829 --> 00:22:13.389
Would you agree with that kind of characterization?
00:22:13.549 --> 00:22:15.789
Or is it I'm sure there's some subtleties there?
00:22:16.029 --> 00:22:41.549
Would you say it's a decent characterization that Gemini has continued a more consumer slash student focus with things like learn mode, where chat is now pivoting more to business, and Claude has always been fairly more business oriented and has not really made as much effort to be more public consumer kind of something you would use in schools, unless you get the higher use.
00:22:42.509 --> 00:22:43.069
That's fair.
00:22:43.149 --> 00:22:43.389
Yeah.
00:22:43.469 --> 00:22:48.669
So Jim and I is very much more personal use, and the other two are more enterprise.
00:22:48.829 --> 00:23:03.149
And with that, you know, I did remember that about a month ago, Jim and I came out with Spark, which is their agent, but I don't think of it as much of an agent because it's more like help me do shopping, help me, you know, give me my daily brief.
00:23:03.309 --> 00:23:21.469
Like once again, kind of that consumer, yeah, you know, individual wants to gab something to help their life out a little bit, but this is not like a business oriented, give me a spreadsheet that's going to plan out, you know, our sales growth for the next quarter kind of thing.
00:23:21.629 --> 00:23:26.109
And I mean, that makes sense because we know that Gemini is embedded in the schools.
00:23:26.189 --> 00:23:31.549
They've Google Tools, the Google ecosystem has been offered in the public schools for free.
00:23:32.189 --> 00:23:34.109
For well over a decade now.
00:23:34.269 --> 00:23:48.509
And as they roll out their Gemini, you know, the idea is, and it's a good business idea, is that we want to go ahead and have all of our people learning and getting used to our AI so that they just continue using it.
00:23:48.669 --> 00:23:56.669
I mean, to me, that's the only explanation that anybody uses Facebook anymore, is because they grew up with it and that's what they're used to.
00:23:56.829 --> 00:24:00.269
Is it the superior social media network that's the best?
00:24:00.989 --> 00:24:02.029
Yeah, exactly.
00:24:02.269 --> 00:24:05.229
Laughable, but yet, you know, millions of people still use it.
00:24:05.389 --> 00:24:09.149
A lot of people in school psychology, you know, still use it.
00:24:09.229 --> 00:24:14.349
Um, and so I guess that's their approach, but is it the best product for what you're trying to get at?
00:24:14.509 --> 00:24:25.629
And I think with the agentic AIs, that it's kind of the theme for our episode today, uh you the answer would be yes, uh, you know, agents for Claude and Chat GPT are going to be better.
00:24:25.869 --> 00:24:29.709
But, you know, you have to, and I I always go back to this, Adam.
00:24:29.789 --> 00:24:31.709
You're like, why are you using this?
00:24:31.949 --> 00:24:36.509
It's a cool tool, but is there utility to learning this at this point?
00:24:36.669 --> 00:24:42.029
And if not, maybe you just wait a while until a use becomes apparent.
00:24:42.109 --> 00:24:50.989
Uh, hopefully, though, our little chat, we've given you a couple uses that you can try out things that might be quite helpful, especially if you're listening to this.
00:24:51.149 --> 00:24:52.989
We're recording this in the summer months.
00:24:53.069 --> 00:25:07.149
And the summer for me is always a time to have a little bit of fun and trying out new stuff when I'm not so pressed with all of the daily requirements of being in academia or being in the schools or you know, otherwise.
00:25:07.309 --> 00:25:09.709
So this is a good time to take advantage of that.
00:25:09.949 --> 00:25:11.549
Adam, you have any other suggestions?
00:25:11.709 --> 00:25:14.189
You said you had two agentic agent ideas.
00:25:14.349 --> 00:25:23.389
One was using the program evaluation and Google surveys or something like that, and kind of going that route, but you also had one for maybe people in private practice.
00:25:23.629 --> 00:25:33.229
If I had a private practice and I was interested in driving business my way, I would have, if you have a website, you get all sorts of analytics.
00:25:33.549 --> 00:25:38.189
And those analytics never get looked at because it's just who has the time for that?
00:25:38.429 --> 00:25:41.709
I can point AI at my website.
00:25:41.789 --> 00:25:50.189
It can look at the clicks, which which pages get the most interaction, what I can do to improve the language.
00:25:50.589 --> 00:25:53.949
And this is these are things I used to do, but I would cut and paste and cut and paste.
00:25:54.109 --> 00:25:56.509
Now I can just say, go read my website.
00:25:56.749 --> 00:26:02.589
Go find me five other private practices in my area that have similar websites.
00:26:03.549 --> 00:26:05.309
Doing a market analysis.
00:26:05.469 --> 00:26:09.549
What can I do to put myself out ahead?
00:26:09.629 --> 00:26:25.389
I mean, if you're at all interested in business in that sort of way, it can be a way to increase visibility, which would be really good for someone who's either got inside private private practice or new to practice, or just wants to increase their visibility.
00:26:25.869 --> 00:26:28.189
So I do it in two ways with my website.
00:26:28.349 --> 00:26:30.429
I give it access to all my analytics.
00:26:30.669 --> 00:26:36.669
Google also gives me Google analytics that tell me uh where traffic's coming from.
00:26:36.909 --> 00:26:48.589
And then I also say, okay, now work as a marketing agent, go look at the other other companies that might be competitors.
00:26:48.749 --> 00:26:51.309
There's not a whole lot, luckily, right now.
00:26:52.669 --> 00:26:55.309
And just give me ideas for how to improve.
00:26:55.629 --> 00:27:01.229
And you know, a half an hour or an hour later, I'll get some feedback and I'll be like, oh, that's a good idea.
00:27:01.389 --> 00:27:09.789
No, that's not, or like dig in more here, and it's just back and forth where it's essentially like having an intern that I set loose to do research on the internet.
00:27:10.029 --> 00:27:15.869
So anything you want to find on the internet, you can have it go and do that.
00:27:16.429 --> 00:27:48.189
Well, it sounds like even if you were going to say go into a salary negotiation with an agency or even your school district, though I know school districts aren't necessarily always into bargaining, but you could certainly go, you know, look up the last six months of help wanted in the area, you know, see if you can find any published salaries regarding these types of positions, what is the average, you know, and that would be a really good way of using it because I can only imagine the amount of hours that you might spend otherwise trying to ascertain.
00:27:48.269 --> 00:27:58.509
And I think that's one of the things because in you know, our fields, whether it's school psychology or even the broader kind of clinical, there are not enough of us out there.
00:27:58.669 --> 00:28:11.869
And so we are kind of in the bargaining, you know, we we have a little bit of a leg up because we can provide a skill that is not in high, uh well, is in high demand, but not a lot of very low supply.
00:28:12.029 --> 00:28:21.149
And so how can you bargain effectively where you're not just, you know, I want a million dollars a year, you know, that's obviously laughable, but you don't want to downplay either.
00:28:21.309 --> 00:28:30.669
And I think in fields uh with psychology, traditionally we don't really do a good job advocating for our own salaries, perks, and things like that.
00:28:30.829 --> 00:28:37.149
I find that in my private practice I have to keep encouraging my employees and things of like, hey, no, you're worth it.
00:28:37.309 --> 00:28:40.669
Like, you know, this is this is what you need to be arguing for.
00:28:40.829 --> 00:28:45.709
Like if you're working for a district, yes, they have a salary scale, but guess what?
00:28:45.869 --> 00:28:49.629
They make exceptions when they can't find anybody and have to have them.
00:28:49.789 --> 00:28:57.469
You know, so don't continually kind of shoot yourself in the foot because you need to look out for yourself in addition to the students.
00:28:57.549 --> 00:29:05.789
And one of the ways is that you get properly compensated for you know what is a hard job, especially when you're in season uh in school.
00:29:05.949 --> 00:29:26.749
So I keep thinking, as we chat longer, Adam, I keep thinking of other uses that aren't necessarily technically using personal information, which I think from what you've been telling me um in our conversation here, maybe that is not a starting point or even anything you should do at this point with a gentic kind of AI processes.
00:29:27.949 --> 00:29:33.869
I would stay away from giving it access to anything that you um want to make it out of that.
00:29:34.029 --> 00:29:36.589
You don't want fed into AI.
00:29:36.669 --> 00:29:38.989
So no PII, no PHI.
00:29:39.229 --> 00:29:39.469
Yeah.
00:29:39.789 --> 00:29:40.109
Yeah.
00:29:40.349 --> 00:29:41.469
I I would agree.
00:29:41.549 --> 00:29:47.949
Um, and just a quick PII, PHI, it's either personal identifiable information or personal health information.
00:29:48.269 --> 00:29:49.469
Protected health information.
00:29:49.789 --> 00:29:51.309
Protected, you know, I don't know why.
00:29:51.389 --> 00:29:55.469
It's because PII and PHI, they're basically very similar.
00:29:55.629 --> 00:29:58.429
And you're right, it's protected health information.
00:29:58.749 --> 00:30:08.429
But, you know, we just need to be very aware, it and that's any identifiable characteristics of, of course, any clients, students that you might be working with.
00:30:08.749 --> 00:30:13.149
So all of these what even if it's redacted.
00:30:13.709 --> 00:30:14.189
Right.
00:30:15.469 --> 00:30:21.469
And remember, AI, of all things, is really good at summarizing and finding things, right?
00:30:21.549 --> 00:30:26.189
And boiling down lots of information, condensing it to a high probability.
00:30:26.429 --> 00:30:28.909
By definition, like that's how the whole system works.
00:30:29.069 --> 00:30:36.509
So can you imagine any better kind of detective in trying to figure out who you've tried to blind some sort of information from?
00:30:36.829 --> 00:30:40.989
AI is going to be able to do that, probably most effectively of all its functions.
00:30:41.149 --> 00:30:43.789
And so just don't use any of it.
00:30:43.869 --> 00:30:46.829
And I would just steer clear if students keep it broad.
00:30:47.069 --> 00:31:01.149
If you're going to use these agentic AIs, I think as we go on with our future podcasts, uh, we will get into more use cases of AI where you can use PII, and that's within particular systems that have been set up for that purpose.
00:31:01.229 --> 00:31:14.669
But these general purpose usage, including especially agentics, I would say, you know, keep it broad, uh, can still be very, very helpful without, you know, even skirting the issues of legality and ethics.
00:31:14.989 --> 00:31:15.709
Agreed.
00:31:16.029 --> 00:31:16.749
All right, Adam.
00:31:16.829 --> 00:31:24.989
So believe it or not, we've survived our first half hour, and hopefully so have our listeners, as we kind of get into this whole uh back and forth.
00:31:25.149 --> 00:31:36.349
But uh, you want to leave our listeners with anything uh at the end of kind of a takeaway for them to maybe try in the next couple weeks before they come back and hear our next discussion?
00:31:36.669 --> 00:31:38.109
I think just have fun.
00:31:38.509 --> 00:31:39.469
Be curious.
00:31:40.029 --> 00:31:41.709
That's in in general in life.
00:31:41.869 --> 00:31:44.189
Having fun being curious is a good thing.
00:31:44.269 --> 00:32:02.269
But for AI, I mean, if you can if you can embrace the idea that this is all new and this is all going to change, and this is something that can be fun, then it's gonna be a lot better than feeling like I have to stay on top of this to stay relevant.
00:32:02.429 --> 00:32:06.589
Um that's that's just not a good, that's not a good way to to feel.
00:32:06.749 --> 00:32:08.109
It's not a good approach.
00:32:08.349 --> 00:32:26.909
So if you just think of this as like, wow, this is a really cool, interesting thing that I can learn about that will increase efficiency and can help improve the services I provide, you're gonna, you're gonna have a lot more, you're just gonna be in a better headspace and you're gonna learn more and save time.
00:32:27.789 --> 00:32:30.749
Yeah, I totally agree about the curiosity aspect.
00:32:30.909 --> 00:32:35.549
I mean, that's why I've liked technology over all these years and reviewed so much.
00:32:35.789 --> 00:32:41.629
It's not just AI, you know, it's how could this be utilized, how could this enhance what I can do?
00:32:41.789 --> 00:32:45.149
How can this enhance what other practitioners can do?
00:32:45.309 --> 00:32:54.509
You know, I think most of us got into it, right, to do well by our clients and students, and we want to be able to spend most of our time doing that.
00:32:54.669 --> 00:33:10.509
We don't want to spend it filling out forms, we don't want to spend it doing a lot of these kind of background things that these agencies now have a promise to maybe alleviate so that we can really focus on the core of why majority of us got into the field.
00:33:10.749 --> 00:33:17.389
So uh, but once again, not because you have to, because that's just where you see the promise.
00:33:17.469 --> 00:33:22.109
And some of these are gonna fall flat on their face and just be utter failures.
00:33:22.269 --> 00:33:27.709
I'm still struggling with this comic strip idea that I have uh that just hasn't quite come to fruition.
00:33:28.029 --> 00:33:30.669
But I was curious and I'll get it there.
00:33:30.829 --> 00:33:34.989
Maybe I'm just waiting for the technology to make it easier on me with these AIs.
00:33:35.069 --> 00:33:38.189
Like if something doesn't work, give it two months, return to it.
00:33:38.269 --> 00:33:44.589
You might find all of a sudden there's a button that'll do it, you know, in a tenth of the speed it would have taken two months ago.
00:33:44.669 --> 00:33:47.469
Um, you know, the field is just changing so quickly.
00:33:47.549 --> 00:33:51.869
And so, you know, being curious like that, I think is just a requirement.
00:33:52.029 --> 00:33:53.629
Otherwise, you're gonna get buried.
00:33:53.789 --> 00:33:59.149
This is not behavior analysis where you learn it once and it won't change in 50 years, kind of thing.
00:33:59.309 --> 00:34:01.709
You know, this is something that's changing every week.
00:34:01.949 --> 00:34:05.949
I know when I do my presentation, I looked back at what I was doing a year ago.
00:34:06.109 --> 00:34:11.630
I have like five slides left out of a hundred that I haven't had to change or update.
00:34:11.710 --> 00:34:21.389
And Adam, you probably have gone through three cycles of you know, full cycles of slide presentations that you just don't even use because they're no longer relevant.
00:34:21.710 --> 00:34:23.630
Yeah, it's it's changing pretty rapidly.
00:34:23.789 --> 00:34:33.469
And I will say I met with someone in OpenAI last week who does the chet for educators, and that was my biggest ask because he was like, What would you like to see?
00:34:33.710 --> 00:34:40.110
I said, I'd like to see a FERPA compliant a way to use codecs in a FERPA compliant environment.
00:34:40.430 --> 00:34:42.670
And he took notes.
00:34:42.910 --> 00:34:44.110
I don't know what that means.
00:34:44.990 --> 00:34:48.829
Notes are good, notes are not dependent on what's in here anymore, you know.
00:34:49.150 --> 00:34:59.390
But I'm hoping, you know, that that one of these frontier AI companies will make this happen sooner than later because it's gotta happen.
00:35:00.350 --> 00:35:01.550
Yeah, I agree.
00:35:01.950 --> 00:35:02.829
Imagine it not.
00:35:03.630 --> 00:35:06.670
All right, well, I think we're gonna go ahead and wrap up then.
00:35:06.829 --> 00:35:11.630
Uh, thanks for listening to our initial Beyond the Box AI and mental health.
00:35:11.789 --> 00:35:19.309
I'm Dan Florell with Adam Lockwood, and we hope that you enjoyed it enough to come back in a couple of weeks.
00:35:19.390 --> 00:35:25.789
Uh, this is part of the Cite the Practice Podcast Network, and see you next time.
00:01:01.340 --> 00:01:04.299
The following is a site to practice production.
00:01:14.460 --> 00:01:22.780
Hi there and welcome to our very first inaugural podcast of Beyond the Bot, AI and Mental Health.
00:01:22.939 --> 00:01:24.460
I'm Dan Florel.
00:01:24.540 --> 00:01:34.460
I'm a professor over at Eastern Kentucky University and CEO of MindSci, in addition to uh the vice president of Psych to Practice.
00:01:34.620 --> 00:01:36.379
And with me today is Adam.
00:01:36.539 --> 00:01:38.300
Adam, why don't you introduce yourself?
00:01:38.859 --> 00:01:43.740
Adam Lockwood, associate professor of school psychology at Kent State University.
00:01:43.819 --> 00:01:49.180
And I have a consulting business called Lockwood Educational Educational and Psychology.
00:01:49.500 --> 00:01:51.340
It's good to get your own play here, Adam.
00:01:51.659 --> 00:01:55.819
Lockwood Educational and Psychological Psychological Consulting.
00:01:56.300 --> 00:02:04.219
So for all of you out there, one of the first lessons to learn is don't make your own business a huge mouthful to say when you're first getting there.
00:02:04.539 --> 00:02:05.900
So what is the focus?
00:02:06.060 --> 00:02:19.099
As you can tell, the focus of our podcast, and this is an initial episode, is to take about 30 minutes of your day every couple weeks to chat about AI and its influence in mental health.
00:02:19.500 --> 00:02:22.219
It's just going to be the two of us, Adam and I.
00:02:22.460 --> 00:02:38.140
We've had quite a few years of experience in AI, as much as anybody can really talk about it when it first burst on the scene in 2020 with the initial release of Chat GPT, though there was, of course, things going behind the scenes for quite a bit longer than that.
00:02:38.300 --> 00:02:51.180
And both of us have spent a fair amount of time and do quite a few trainings looking at the ethics of AI use and also how it might be applied within mental health settings, specifically more school psychology related.
00:02:51.419 --> 00:03:06.860
But we're going to start off with our inaugural issue looking at the promises and perils of AI agents, which, if you're just understanding initially what AI is, this is a whole other level.
00:03:07.020 --> 00:03:14.060
And I'm going to defer to Adam because Adam has many of his little AI bot minions working for him.
00:03:14.299 --> 00:03:27.099
And I know about them, uh, but am slightly intimidated by uh these little independent, autonomous things going on and skittering about and doing things uh that I'm not quite sure what they're doing.
00:03:27.259 --> 00:03:36.139
So, Adam, I'm going to be somewhat of a learner here, and you can kind of tout some of your expertise as far as how we might be able to use it within mental health.
00:03:36.219 --> 00:03:41.579
And of course, we also named it perils, so that there will be some drawbacks, I'm sure.
00:03:41.819 --> 00:03:43.659
But uh turn it over to you.
00:03:44.299 --> 00:03:44.859
Sure.
00:03:45.019 --> 00:03:50.060
I think first it's a good idea to differentiate between maybe a chatbot and an agent.
00:03:50.379 --> 00:03:50.859
Yeah.
00:03:51.099 --> 00:03:57.019
So I mean, like chatbots are basically uh ones that everybody's familiar with, right?
00:03:57.099 --> 00:04:02.620
You get your little phone out, you open up the app, and you start talking like it's a real person.
00:04:02.939 --> 00:04:09.019
And it goes ahead and responds and has all its little tendencies, like making you feel like you're a brilliant person in the world.
00:04:09.180 --> 00:04:21.340
And then agents, of course, are where you're going to go ahead and set it missions, and it will accomplish those missions any way it sees fit to get the end result that you are specifying for it.
00:04:21.500 --> 00:04:28.220
And I imagine the more guardrails and specifics, probably the better return you get on those agents.
00:04:28.379 --> 00:04:32.699
But that's just the off the top of my head, kind of what the difference between the two would be.
00:04:32.939 --> 00:04:39.259
Yeah, well, first of all, I like you said, you talk to the chatbot, because at this point I don't type.
00:04:39.500 --> 00:04:40.699
Do you type still?
00:04:41.180 --> 00:04:42.699
I'm still a typer, you know.
00:04:42.860 --> 00:04:44.699
I'm even a clumsy texter.
00:04:44.780 --> 00:04:45.500
You know what it is?
00:04:45.579 --> 00:04:54.220
I just, especially when I'm out and about, I'm a little uh conscientious about talking to my machine uh out there.
00:04:54.379 --> 00:04:56.460
So I haven't really embraced that yet.
00:04:56.620 --> 00:04:59.980
I need to, because that's actually the way that I usually communicate.
00:05:00.139 --> 00:05:05.740
Like when I know I have decent ideas, I talk them through, but I usually find colleagues to do that with.
00:05:05.819 --> 00:05:11.340
And I guess uh, you know, if I don't find colleagues, this will be a nice little option here.
00:05:11.819 --> 00:05:18.540
But in the back of my mind, playing that sometimes it gets a little too enthusiastic about my own ideas and doesn't push back enough.
00:05:18.780 --> 00:05:22.460
So I take it though, you're definitely into the verbal descriptions.
00:05:22.540 --> 00:05:28.620
And I think the couple times we've hung out, you know, online and to talk about things, I see how you work with that.
00:05:28.860 --> 00:05:36.860
Any advice on it, or is it just so natural that you really don't have to think about learning yet another skill and ways of framing things?
00:05:36.939 --> 00:05:39.020
Kind of I'm thinking like prompting, right?
00:05:39.180 --> 00:05:45.020
Prompting, if the more specific and the more context-driven, usually the better results you get.
00:05:45.260 --> 00:05:47.340
Well, I don't think people need to learn how to talk.
00:05:47.500 --> 00:05:49.420
Most people know how to talk already.
00:05:50.140 --> 00:05:52.700
So just talk, ramble.
00:05:52.860 --> 00:06:04.460
The truth is the earlier, the early research that's coming out is you get better output if you talk, because it's through this kind of rambling what you say, maybe not.
00:06:04.700 --> 00:06:05.820
Well, let me think.
00:06:06.220 --> 00:06:16.460
It gives the AI some insight into kind of the inner workings that you you have going on that you normally would self-censor by typing.
00:06:16.939 --> 00:06:20.060
And so by doing that, you actually get better output.
00:06:20.300 --> 00:06:21.900
Not only that, it's just faster.
00:06:21.980 --> 00:06:25.180
Um, and it you know can produce carpal tunnel.
00:06:25.660 --> 00:06:32.140
Yeah, well, you know, I think I know why I have now that we've been discussing it, why I have this mental block about it.
00:06:32.379 --> 00:06:47.500
It's because way back when I got into dictation software a bit, and where you go stop, and you know, for a period, and you had to be pretty precise because if you put in your ums and everything, it messed up what it was transcribing.
00:06:47.660 --> 00:07:06.060
And this obviously, this technology doesn't do that, but I have this prior technology mental block that I think is causing some of the issues with me doing the interaction with AI because that previous learning is telling me like, no, you have to be precise, otherwise, it's gonna mess it up.
00:07:06.540 --> 00:07:08.780
Don't worry about being precise.
00:07:09.180 --> 00:07:09.820
Gotcha.
00:07:09.900 --> 00:07:11.100
I'm checking that off.
00:07:11.340 --> 00:07:18.700
What I'll say about agents, how I use them, is they're essentially chatbots that have arms and legs.
00:07:18.860 --> 00:07:36.860
And so, as normally, if you're using ChatGPT, for example, you might be like, Well, I want to create a blog and it will give you information, and you go back and forth, and then you take that information and you might paste it into your blog.
00:07:37.100 --> 00:07:39.500
In this case, it works autonomously.
00:07:39.580 --> 00:07:45.420
So I might have that back and forth and say, okay, now great, go to my website, update my language.
00:07:45.740 --> 00:07:49.420
It just takes the human out of the loop so much.
00:07:49.580 --> 00:07:52.379
They they they refer to it as the sandwich or the burger.
00:07:52.540 --> 00:07:57.980
You have human on the front end, so we'll have a back and forth with AI, figuring out what exactly we want to do.
00:07:58.140 --> 00:08:04.860
Once we have a plan, you send it out to go do it, and then at the end you check and you you act as a human in the loop.
00:08:04.939 --> 00:08:10.460
But all that cutting and pacing, going back and forth, is eliminated.
00:08:10.700 --> 00:08:19.340
All of the busy work, all of the kind of boring stuff that you don't want to do can be outsourced at this point.
00:08:20.300 --> 00:08:37.179
So when you say that, what comes to mind with me is that you are giving access to this AI, to things that require your passwords and you know, your different basically personal information that I think would give some people pause.
00:08:37.340 --> 00:08:44.460
I mean, is this all retained that you I mean, have you looked up and see like how is this retained within Chat GPT?
00:08:44.700 --> 00:08:58.059
Um, you know, like what safeguards are there that it doesn't just go gonzo one day and all of a sudden you have 50 appointments with a massage therapist, you know, in California or something.
00:08:58.299 --> 00:09:04.299
Well, you got to be careful because the truth is agents have gone and deleted files that were very important.
00:09:04.539 --> 00:09:07.100
So you need to make sure that you sandbox.
00:09:07.340 --> 00:09:13.340
So the first thing you do is you start out and give it access to an email account that is um burning.
00:09:15.019 --> 00:09:15.659
Yeah.
00:09:15.899 --> 00:09:16.220
Yeah.
00:09:16.460 --> 00:09:20.220
So I'll give you a use of one of my, I'll give you an example of a use case.
00:09:20.460 --> 00:09:25.340
I get a lot of Google Scholar alerts that are just AI articles.
00:09:25.500 --> 00:09:27.580
I get six emails a day.
00:09:27.899 --> 00:09:33.740
And within those, about every day it works out to about 48 different articles.
00:09:34.060 --> 00:09:37.180
I have an email account that's just a burner account.
00:09:37.259 --> 00:09:43.340
I use it for nothing but that, and I let AI have complete read access to it.
00:09:43.500 --> 00:09:47.259
It can't send anything out from there, but it can read those.
00:09:47.420 --> 00:09:49.899
I definitely wouldn't want my bank statements going there.
00:09:50.060 --> 00:09:54.379
I certainly wouldn't want patient or student information to go in there.
00:09:54.779 --> 00:09:57.580
But that's that's one way.
00:09:57.740 --> 00:10:03.740
And then I give it access to folders, but those folders are sandboxed.
00:10:03.820 --> 00:10:13.820
So it has its own folder within documents that it can go, write files, edit files, delete files, but it cannot leave that folder.
00:10:14.220 --> 00:10:21.420
So when you talk about that folder, is that like for me, I have Dropbox, and let's say I wanted to give you access to a folder.
00:10:21.659 --> 00:10:29.820
I could request and say, hey, give this access, but that doesn't mean you have carte blanche in my whole Dropbox, you know, account.
00:10:29.980 --> 00:10:32.060
You just have the access to one folder.
00:10:32.220 --> 00:10:41.899
And I guess, you know, this is where I need to do a little more research, but like to get the agents going, is there a particular agent button for our common like Google Gemini?
00:10:42.379 --> 00:10:44.620
I know you've been using Chet or Claude.
00:10:44.779 --> 00:10:49.740
I mean, is there something that's labeled agent on the side and you click on that and start setting things up?
00:10:49.899 --> 00:10:54.620
I mean, where would people get a better idea of how to even try this out?
00:10:54.860 --> 00:10:56.620
Because I I wouldn't even know.
00:10:56.860 --> 00:11:00.139
You know, I guess I could ask Chet, like, how do I use your agent?
00:11:00.220 --> 00:11:05.500
But yeah, I guess I should say that I'm using agent as a fairly loose term.
00:11:05.659 --> 00:11:08.220
I think this is more like agentic AI.
00:11:08.300 --> 00:11:15.500
That means it has some autonomy and can work on its own, can go to achieve goals with feedback from me.
00:11:16.060 --> 00:11:19.899
The two there, and there's a wide spectrum.
00:11:20.060 --> 00:11:55.550
You have like open cloud where people are running their own open weight or open source.
00:11:55.629 --> 00:12:04.750
It's just models that they're running on their own machine, or they might be using indirect access to other models.
00:12:05.389 --> 00:12:15.550
But the easiest entry point would be either codecs, which is open AI, so it's essentially Chat GPT, or Claude Anthropic Cowork.
00:12:16.110 --> 00:12:16.190
Okay.
00:12:16.750 --> 00:12:35.389
But I know mostly Codex because if you've ever used CLOD, you know that the the use limits are pretty stringent with tokens, just meaning they give you a set amount of use of the product, and when you use too much, it will say, hey, you're done for three hours.
00:12:35.550 --> 00:12:39.389
Actually, Codex will do that too, but it's much more generous.
00:12:40.030 --> 00:12:52.670
And you say this, is this where people would need paid accounts regardless, and they still have the token limits, or is this the free accounts that you uh run against the token limits more typically?
00:12:53.389 --> 00:12:56.110
You should have a paid account, just full stop.
00:12:56.509 --> 00:13:04.190
You get access to better models, it's$20 a month, which is an extremely good deal.
00:13:04.430 --> 00:13:09.870
Right now, this whole industry is being subsidized by the companies.
00:13:10.030 --> 00:13:10.269
Right.
00:13:10.509 --> 00:13:15.790
So with a$20 account, you can get hundreds of dollars worth of compute use.
00:13:16.030 --> 00:13:20.670
With a$200 account, I've heard estimates as high as like$8,000.
00:13:21.150 --> 00:13:27.070
So they're losing a lot of money right now because it partially because they want to get your use.
00:13:27.230 --> 00:13:28.990
They want to get you as a customer.
00:13:29.310 --> 00:13:31.629
But maybe that$20, it's worth it.
00:13:31.790 --> 00:13:34.670
I don't know if the free accounts of these have access.
00:13:34.830 --> 00:13:35.550
I doubt it.
00:13:35.710 --> 00:13:37.790
I don't know because I have paid accounts.
00:13:37.950 --> 00:13:42.830
Um, I could Google it or I could ask AI right now, but pay for the$20 for one of them.
00:13:42.910 --> 00:13:45.710
And and my guess would be ChatGPT.
00:13:46.030 --> 00:13:53.790
You have to, of course, remember that it's not HIPAA or FERPA compliant unless you have some sort of enterprise account.
00:13:54.030 --> 00:14:04.190
And you sure of that through someone at your company or organization or district, but or ChatGPT for teachers, you can have a FERPA.
00:14:04.990 --> 00:14:05.230
Right.
00:14:05.389 --> 00:14:16.750
For the like Google, Google Gemini, uh, if you have an enterprise through that, you should be having some FERPA compliance in regards to any of the things you use within the school's account.
00:14:16.990 --> 00:14:22.190
Just because you work for a school and you set up a free account or something, that doesn't apply, right?
00:14:22.269 --> 00:14:24.830
It has to be within that email account.
00:14:25.150 --> 00:14:44.750
So just pivoting a little bit, how would someone use an agent or agentic AI within kind of mental health or within the school kind of setting, knowing that institutions may be a little reluctant to have these agentics running around doing autonomous things?
00:14:45.470 --> 00:14:49.870
I wouldn't start by using it for healthcare or for school services.
00:14:49.950 --> 00:14:56.990
I would start by finding other repetitive time sucks in your life that you you want to automate.
00:14:57.310 --> 00:15:04.350
And so in your case, I know it was uh really looking up research, which resonates with me because I am a research hound.
00:15:04.430 --> 00:15:15.389
I mean, I'm always downloading articles and putting them away and you know, throwing them through Notebook LM, uh, which I think you know could also be used for mental health, obviously.
00:15:15.550 --> 00:15:22.750
So think about all of the areas, even though Adam and I are very techie oriented and use it a lot for AI.
00:15:22.990 --> 00:15:34.910
You can imagine uh doing it for, say, reading comprehension interventions and coming up with the latest ways of tracking that or the novel ways to use curriculum-based measurement in a particular age group.
00:15:35.070 --> 00:15:59.070
You know, whatever setting that you're working in, one of the things that I think we have challenges with as practitioners is that continuing ed aspect and where we're advancing our knowledge to be the most cutting edge, hopefully, in whatever area that we are providing services in, because that's going to help our students or our clients.
00:15:59.389 --> 00:16:02.190
And it doesn't need to be where you're spending a lot of money.
00:16:02.269 --> 00:16:18.350
For that 20 bucks uh for one of these, you know, big clawed Chat GPT, you can get and be more well informed from a scientific journal kind of standpoint than I think you could ever when being in practice before.
00:16:18.509 --> 00:16:27.950
Like you don't have to be affiliated with a university anymore because Google Scholar and Google Scholar Labs has access to a tremendous range of journals.
00:16:28.110 --> 00:16:34.509
It's not comprehensive, but it's a heck of a lot more than, say, 10 years ago, most practitioners have.
00:16:34.590 --> 00:16:50.590
And I think if you use agentic AI to help boil it down, you give it some structure, set up a dummy email like Adam was talking about, I think it really can get you up to speed on stuff that will have a real impact on the way that you practice.
00:16:51.070 --> 00:16:57.550
When you say, I mean, that seems like a perfect use case because nobody's really worried about a journal article somehow getting loose.
00:16:57.629 --> 00:16:59.629
Um, you know, it's open source anyway.
00:16:59.710 --> 00:17:00.750
It already has it.
00:17:00.830 --> 00:17:03.150
The AI already has that information.
00:17:03.389 --> 00:17:06.670
And yet we can easily apply it to our clientele.
00:17:07.470 --> 00:17:07.869
Yeah.
00:17:08.029 --> 00:17:13.549
And I think it's not going to be too long before we're going to have HIPAA compliant agents.
00:17:13.710 --> 00:17:19.389
So getting an understanding of how this all works will just help you to be ahead of the curve.
00:17:24.349 --> 00:17:32.589
Right, as opposed to trying to go from crawl to run right away, which can be very messy if you've ever seen a child that tries to go from crawl to run.
00:17:32.910 --> 00:17:35.469
It's just a disaster, face plants all around.
00:17:35.629 --> 00:17:39.149
We don't want that, the equivalent when we're using AI.
00:17:39.389 --> 00:17:43.869
And I think already, you know, we've got a sense there's a wide range out there.
00:17:44.029 --> 00:17:52.349
And one of the things Adam and I are going to run into is we're going to get a little too tech wonky sometimes because we've been living in this space for so long.
00:17:52.509 --> 00:17:57.789
And some of the shortcuts and things that he tells me and I tell him makes perfect sense to us.
00:17:58.029 --> 00:18:04.989
But as listeners, you know, it may, you may have a little bit more time to get up to speed.
00:18:05.069 --> 00:18:12.909
So we'll try to, I'll try to put the brakes on every now and then so we can explain a little further detail about what exactly we mean with some of this stuff.
00:18:13.069 --> 00:18:17.869
Because we didn't really pick the easiest topic to introduce people to our podcast.
00:18:18.029 --> 00:18:28.349
I think this is probably one of the most cutting-edge aspects from a general public, anyway, that most general public could utilize are these agentic AIs.
00:18:28.509 --> 00:18:32.589
I'm not talking about the programmers and all that, you know, that's really high in the weeds.
00:18:32.829 --> 00:18:40.909
But uh, would you agree, or would you say there's some other stuff out there that people need to be kind of aware of, even if they haven't quite used them yet?
00:18:41.149 --> 00:18:45.069
Well, I was thinking when you said that, yeah, maybe this isn't the best topic for the first one.
00:18:45.469 --> 00:18:47.309
But so yeah, I don't know.
00:18:47.469 --> 00:18:56.669
Uh I can think of a couple of use cases for one for school psychologists and one for uh for people in private practice right off the bat.
00:18:56.909 --> 00:18:57.469
Yeah.
00:18:57.709 --> 00:19:03.389
So one thing that school psychologists should be doing, and they don't do as much because they don't have time, is program evaluation.
00:19:04.909 --> 00:19:12.349
AI, create a survey with me, upload the survey to Google Forms, whatever.
00:19:12.749 --> 00:19:17.309
All essentially with no work on my side.
00:19:17.869 --> 00:19:18.589
Very little.
00:19:18.749 --> 00:19:23.389
I mean, let's say you want to be careful, you don't want to give access to your Google Forms.
00:19:23.709 --> 00:19:28.109
I can understand that, but it could still generate the document that you could upload.
00:19:28.349 --> 00:19:30.429
But at this point, go ahead.
00:19:30.909 --> 00:19:37.949
I was gonna say with the Google forms, would that be something that a say school psychologist could do within the enterprise suite?
00:19:38.109 --> 00:19:54.429
Like, does Google, because that's not one you've mentioned, does Google have some agent equivalency that we talked about with both Claude and ChatGPT that people in the school might be able to utilize um for doing something like that uh in a survey?
00:19:54.989 --> 00:19:59.789
It's funny, Google has um kind of fallen behind in 2026.
00:20:00.669 --> 00:20:05.789
I thought ChatGPT won the race for 23, 24.
00:20:05.949 --> 00:20:08.669
I thought 2025 was Google's year.
00:20:08.909 --> 00:20:17.149
I think that the first part of 2026 was Cloud, and now I think ChatGPT if you're looking at just the big models.
00:20:17.869 --> 00:20:26.509
But yeah, there are agent platforms for um Gemini, but I'm not seeing their use very often.
00:20:26.669 --> 00:20:32.909
I think it's I think it's probably more within enterprises where they're being built.
00:20:40.749 --> 00:21:01.949
Well, I was gonna say what I've been hearing lately is, and these are broad stereotypes, but the most consumer-focused AI, and by that I mean like general public, is more Gemini, like a lot of tools are coming out for it, that they're being very like that's the primary, what they're chasing.
00:21:02.189 --> 00:21:16.269
Whereas Claude is chasing more a business environment, and chat is starting to move a little bit more towards that business versus consumers, you know, general consumers or the public.
00:21:16.509 --> 00:21:24.189
And you can see that, like, you know, they had their little video creation snippet thing on ChatGPT that they decided to jettison.
00:21:24.829 --> 00:21:33.229
You know, I also have seen it in ChatGPT decided to go ahead and lessen where you could find learn mode.
00:21:33.389 --> 00:21:37.229
It made a big deal about how you could learn tutoring, you know, through it.
00:21:37.309 --> 00:21:40.669
And now it's still there, but I noticed they've buried it.
00:21:40.749 --> 00:21:46.589
It's not even on like the more button and then the more button, and you have to like actually ask, like, what happened to it?
00:21:46.669 --> 00:21:49.149
And they're like, well, we didn't want to feature it anymore.
00:21:49.229 --> 00:21:50.589
It still has that capability.
00:21:50.749 --> 00:22:10.589
Well, what it tells me is that they're de-emphasizing the usage for, say, a student population or a more public population, they're shifting to that more programming and like this maybe agentic view compared to say Gemini, where you can still feel see like learn mode is pretty prominently displayed.
00:22:10.829 --> 00:22:13.389
Would you agree with that kind of characterization?
00:22:13.549 --> 00:22:15.789
Or is it I'm sure there's some subtleties there?
00:22:16.029 --> 00:22:41.549
Would you say it's a decent characterization that Gemini has continued a more consumer slash student focus with things like learn mode, where chat is now pivoting more to business, and Claude has always been fairly more business oriented and has not really made as much effort to be more public consumer kind of something you would use in schools, unless you get the higher use.
00:22:42.509 --> 00:22:43.069
That's fair.
00:22:43.149 --> 00:22:43.389
Yeah.
00:22:43.469 --> 00:22:48.669
So Jim and I is very much more personal use, and the other two are more enterprise.
00:22:48.829 --> 00:23:03.149
And with that, you know, I did remember that about a month ago, Jim and I came out with Spark, which is their agent, but I don't think of it as much of an agent because it's more like help me do shopping, help me, you know, give me my daily brief.
00:23:03.309 --> 00:23:21.469
Like once again, kind of that consumer, yeah, you know, individual wants to gab something to help their life out a little bit, but this is not like a business oriented, give me a spreadsheet that's going to plan out, you know, our sales growth for the next quarter kind of thing.
00:23:21.629 --> 00:23:26.109
And I mean, that makes sense because we know that Gemini is embedded in the schools.
00:23:26.189 --> 00:23:31.549
They've Google Tools, the Google ecosystem has been offered in the public schools for free.
00:23:32.189 --> 00:23:34.109
For well over a decade now.
00:23:34.269 --> 00:23:48.509
And as they roll out their Gemini, you know, the idea is, and it's a good business idea, is that we want to go ahead and have all of our people learning and getting used to our AI so that they just continue using it.
00:23:48.669 --> 00:23:56.669
I mean, to me, that's the only explanation that anybody uses Facebook anymore, is because they grew up with it and that's what they're used to.
00:23:56.829 --> 00:24:00.269
Is it the superior social media network that's the best?
00:24:00.989 --> 00:24:02.029
Yeah, exactly.
00:24:02.269 --> 00:24:05.229
Laughable, but yet, you know, millions of people still use it.
00:24:05.389 --> 00:24:09.149
A lot of people in school psychology, you know, still use it.
00:24:09.229 --> 00:24:14.349
Um, and so I guess that's their approach, but is it the best product for what you're trying to get at?
00:24:14.509 --> 00:24:25.629
And I think with the agentic AIs, that it's kind of the theme for our episode today, uh you the answer would be yes, uh, you know, agents for Claude and Chat GPT are going to be better.
00:24:25.869 --> 00:24:29.709
But, you know, you have to, and I I always go back to this, Adam.
00:24:29.789 --> 00:24:31.709
You're like, why are you using this?
00:24:31.949 --> 00:24:36.509
It's a cool tool, but is there utility to learning this at this point?
00:24:36.669 --> 00:24:42.029
And if not, maybe you just wait a while until a use becomes apparent.
00:24:42.109 --> 00:24:50.989
Uh, hopefully, though, our little chat, we've given you a couple uses that you can try out things that might be quite helpful, especially if you're listening to this.
00:24:51.149 --> 00:24:52.989
We're recording this in the summer months.
00:24:53.069 --> 00:25:07.149
And the summer for me is always a time to have a little bit of fun and trying out new stuff when I'm not so pressed with all of the daily requirements of being in academia or being in the schools or you know, otherwise.
00:25:07.309 --> 00:25:09.709
So this is a good time to take advantage of that.
00:25:09.949 --> 00:25:11.549
Adam, you have any other suggestions?
00:25:11.709 --> 00:25:14.189
You said you had two agentic agent ideas.
00:25:14.349 --> 00:25:23.389
One was using the program evaluation and Google surveys or something like that, and kind of going that route, but you also had one for maybe people in private practice.
00:25:23.629 --> 00:25:33.229
If I had a private practice and I was interested in driving business my way, I would have, if you have a website, you get all sorts of analytics.
00:25:33.549 --> 00:25:38.189
And those analytics never get looked at because it's just who has the time for that?
00:25:38.429 --> 00:25:41.709
I can point AI at my website.
00:25:41.789 --> 00:25:50.189
It can look at the clicks, which which pages get the most interaction, what I can do to improve the language.
00:25:50.589 --> 00:25:53.949
And this is these are things I used to do, but I would cut and paste and cut and paste.
00:25:54.109 --> 00:25:56.509
Now I can just say, go read my website.
00:25:56.749 --> 00:26:02.589
Go find me five other private practices in my area that have similar websites.
00:26:03.549 --> 00:26:05.309
Doing a market analysis.
00:26:05.469 --> 00:26:09.549
What can I do to put myself out ahead?
00:26:09.629 --> 00:26:25.389
I mean, if you're at all interested in business in that sort of way, it can be a way to increase visibility, which would be really good for someone who's either got inside private private practice or new to practice, or just wants to increase their visibility.
00:26:25.869 --> 00:26:28.189
So I do it in two ways with my website.
00:26:28.349 --> 00:26:30.429
I give it access to all my analytics.
00:26:30.669 --> 00:26:36.669
Google also gives me Google analytics that tell me uh where traffic's coming from.
00:26:36.909 --> 00:26:48.589
And then I also say, okay, now work as a marketing agent, go look at the other other companies that might be competitors.
00:26:48.749 --> 00:26:51.309
There's not a whole lot, luckily, right now.
00:26:52.669 --> 00:26:55.309
And just give me ideas for how to improve.
00:26:55.629 --> 00:27:01.229
And you know, a half an hour or an hour later, I'll get some feedback and I'll be like, oh, that's a good idea.
00:27:01.389 --> 00:27:09.789
No, that's not, or like dig in more here, and it's just back and forth where it's essentially like having an intern that I set loose to do research on the internet.
00:27:10.029 --> 00:27:15.869
So anything you want to find on the internet, you can have it go and do that.
00:27:16.429 --> 00:27:48.189
Well, it sounds like even if you were going to say go into a salary negotiation with an agency or even your school district, though I know school districts aren't necessarily always into bargaining, but you could certainly go, you know, look up the last six months of help wanted in the area, you know, see if you can find any published salaries regarding these types of positions, what is the average, you know, and that would be a really good way of using it because I can only imagine the amount of hours that you might spend otherwise trying to ascertain.
00:27:48.269 --> 00:27:58.509
And I think that's one of the things because in you know, our fields, whether it's school psychology or even the broader kind of clinical, there are not enough of us out there.
00:27:58.669 --> 00:28:11.869
And so we are kind of in the bargaining, you know, we we have a little bit of a leg up because we can provide a skill that is not in high, uh well, is in high demand, but not a lot of very low supply.
00:28:12.029 --> 00:28:21.149
And so how can you bargain effectively where you're not just, you know, I want a million dollars a year, you know, that's obviously laughable, but you don't want to downplay either.
00:28:21.309 --> 00:28:30.669
And I think in fields uh with psychology, traditionally we don't really do a good job advocating for our own salaries, perks, and things like that.
00:28:30.829 --> 00:28:37.149
I find that in my private practice I have to keep encouraging my employees and things of like, hey, no, you're worth it.
00:28:37.309 --> 00:28:40.669
Like, you know, this is this is what you need to be arguing for.
00:28:40.829 --> 00:28:45.709
Like if you're working for a district, yes, they have a salary scale, but guess what?
00:28:45.869 --> 00:28:49.629
They make exceptions when they can't find anybody and have to have them.
00:28:49.789 --> 00:28:57.469
You know, so don't continually kind of shoot yourself in the foot because you need to look out for yourself in addition to the students.
00:28:57.549 --> 00:29:05.789
And one of the ways is that you get properly compensated for you know what is a hard job, especially when you're in season uh in school.
00:29:05.949 --> 00:29:26.749
So I keep thinking, as we chat longer, Adam, I keep thinking of other uses that aren't necessarily technically using personal information, which I think from what you've been telling me um in our conversation here, maybe that is not a starting point or even anything you should do at this point with a gentic kind of AI processes.
00:29:27.949 --> 00:29:33.869
I would stay away from giving it access to anything that you um want to make it out of that.
00:29:34.029 --> 00:29:36.589
You don't want fed into AI.
00:29:36.669 --> 00:29:38.989
So no PII, no PHI.
00:29:39.229 --> 00:29:39.469
Yeah.
00:29:39.789 --> 00:29:40.109
Yeah.
00:29:40.349 --> 00:29:41.469
I I would agree.
00:29:41.549 --> 00:29:47.949
Um, and just a quick PII, PHI, it's either personal identifiable information or personal health information.
00:29:48.269 --> 00:29:49.469
Protected health information.
00:29:49.789 --> 00:29:51.309
Protected, you know, I don't know why.
00:29:51.389 --> 00:29:55.469
It's because PII and PHI, they're basically very similar.
00:29:55.629 --> 00:29:58.429
And you're right, it's protected health information.
00:29:58.749 --> 00:30:08.429
But, you know, we just need to be very aware, it and that's any identifiable characteristics of, of course, any clients, students that you might be working with.
00:30:08.749 --> 00:30:13.149
So all of these what even if it's redacted.
00:30:13.709 --> 00:30:14.189
Right.
00:30:15.469 --> 00:30:21.469
And remember, AI, of all things, is really good at summarizing and finding things, right?
00:30:21.549 --> 00:30:26.189
And boiling down lots of information, condensing it to a high probability.
00:30:26.429 --> 00:30:28.909
By definition, like that's how the whole system works.
00:30:29.069 --> 00:30:36.509
So can you imagine any better kind of detective in trying to figure out who you've tried to blind some sort of information from?
00:30:36.829 --> 00:30:40.989
AI is going to be able to do that, probably most effectively of all its functions.
00:30:41.149 --> 00:30:43.789
And so just don't use any of it.
00:30:43.869 --> 00:30:46.829
And I would just steer clear if students keep it broad.
00:30:47.069 --> 00:31:01.149
If you're going to use these agentic AIs, I think as we go on with our future podcasts, uh, we will get into more use cases of AI where you can use PII, and that's within particular systems that have been set up for that purpose.
00:31:01.229 --> 00:31:14.669
But these general purpose usage, including especially agentics, I would say, you know, keep it broad, uh, can still be very, very helpful without, you know, even skirting the issues of legality and ethics.
00:31:14.989 --> 00:31:15.709
Agreed.
00:31:16.029 --> 00:31:16.749
All right, Adam.
00:31:16.829 --> 00:31:24.989
So believe it or not, we've survived our first half hour, and hopefully so have our listeners, as we kind of get into this whole uh back and forth.
00:31:25.149 --> 00:31:36.349
But uh, you want to leave our listeners with anything uh at the end of kind of a takeaway for them to maybe try in the next couple weeks before they come back and hear our next discussion?
00:31:36.669 --> 00:31:38.109
I think just have fun.
00:31:38.509 --> 00:31:39.469
Be curious.
00:31:40.029 --> 00:31:41.709
That's in in general in life.
00:31:41.869 --> 00:31:44.189
Having fun being curious is a good thing.
00:31:44.269 --> 00:32:02.269
But for AI, I mean, if you can if you can embrace the idea that this is all new and this is all going to change, and this is something that can be fun, then it's gonna be a lot better than feeling like I have to stay on top of this to stay relevant.
00:32:02.429 --> 00:32:06.589
Um that's that's just not a good, that's not a good way to to feel.
00:32:06.749 --> 00:32:08.109
It's not a good approach.
00:32:08.349 --> 00:32:26.909
So if you just think of this as like, wow, this is a really cool, interesting thing that I can learn about that will increase efficiency and can help improve the services I provide, you're gonna, you're gonna have a lot more, you're just gonna be in a better headspace and you're gonna learn more and save time.
00:32:27.789 --> 00:32:30.749
Yeah, I totally agree about the curiosity aspect.
00:32:30.909 --> 00:32:35.549
I mean, that's why I've liked technology over all these years and reviewed so much.
00:32:35.789 --> 00:32:41.629
It's not just AI, you know, it's how could this be utilized, how could this enhance what I can do?
00:32:41.789 --> 00:32:45.149
How can this enhance what other practitioners can do?
00:32:45.309 --> 00:32:54.509
You know, I think most of us got into it, right, to do well by our clients and students, and we want to be able to spend most of our time doing that.
00:32:54.669 --> 00:33:10.509
We don't want to spend it filling out forms, we don't want to spend it doing a lot of these kind of background things that these agencies now have a promise to maybe alleviate so that we can really focus on the core of why majority of us got into the field.
00:33:10.749 --> 00:33:17.389
So uh, but once again, not because you have to, because that's just where you see the promise.
00:33:17.469 --> 00:33:22.109
And some of these are gonna fall flat on their face and just be utter failures.
00:33:22.269 --> 00:33:27.709
I'm still struggling with this comic strip idea that I have uh that just hasn't quite come to fruition.
00:33:28.029 --> 00:33:30.669
But I was curious and I'll get it there.
00:33:30.829 --> 00:33:34.989
Maybe I'm just waiting for the technology to make it easier on me with these AIs.
00:33:35.069 --> 00:33:38.189
Like if something doesn't work, give it two months, return to it.
00:33:38.269 --> 00:33:44.589
You might find all of a sudden there's a button that'll do it, you know, in a tenth of the speed it would have taken two months ago.
00:33:44.669 --> 00:33:47.469
Um, you know, the field is just changing so quickly.
00:33:47.549 --> 00:33:51.869
And so, you know, being curious like that, I think is just a requirement.
00:33:52.029 --> 00:33:53.629
Otherwise, you're gonna get buried.
00:33:53.789 --> 00:33:59.149
This is not behavior analysis where you learn it once and it won't change in 50 years, kind of thing.
00:33:59.309 --> 00:34:01.709
You know, this is something that's changing every week.
00:34:01.949 --> 00:34:05.949
I know when I do my presentation, I looked back at what I was doing a year ago.
00:34:06.109 --> 00:34:11.630
I have like five slides left out of a hundred that I haven't had to change or update.
00:34:11.710 --> 00:34:21.389
And Adam, you probably have gone through three cycles of you know, full cycles of slide presentations that you just don't even use because they're no longer relevant.
00:34:21.710 --> 00:34:23.630
Yeah, it's it's changing pretty rapidly.
00:34:23.789 --> 00:34:33.469
And I will say I met with someone in OpenAI last week who does the chet for educators, and that was my biggest ask because he was like, What would you like to see?
00:34:33.710 --> 00:34:40.110
I said, I'd like to see a FERPA compliant a way to use codecs in a FERPA compliant environment.
00:34:40.430 --> 00:34:42.670
And he took notes.
00:34:42.910 --> 00:34:44.110
I don't know what that means.
00:34:44.990 --> 00:34:48.829
Notes are good, notes are not dependent on what's in here anymore, you know.
00:34:49.150 --> 00:34:59.390
But I'm hoping, you know, that that one of these frontier AI companies will make this happen sooner than later because it's gotta happen.
00:35:00.350 --> 00:35:01.550
Yeah, I agree.
00:35:01.950 --> 00:35:02.829
Imagine it not.
00:35:03.630 --> 00:35:06.670
All right, well, I think we're gonna go ahead and wrap up then.
00:35:06.829 --> 00:35:11.630
Uh, thanks for listening to our initial Beyond the Box AI and mental health.
00:35:11.789 --> 00:35:19.309
I'm Dan Florell with Adam Lockwood, and we hope that you enjoyed it enough to come back in a couple of weeks.
00:35:19.390 --> 00:35:25.789
Uh, this is part of the Cite the Practice Podcast Network, and see you next time.
