The
How to follow along: The transcript is on the left and the live chat is on the right, lined up to the moment each message appeared. Every time stamp is recording time, so you can scrub the video to the same spot.
Transcript
Chat
0:04
TLC Kelly
Here we go! All right, welcome everybody to another exciting conversation. This time, we are talking about, AI and Cognitive Surrender. It's conversation number 116, but before we get started, we want to talk, as always, we want to thank Conductrics. They've been with us since the very, very, very beginning. Thank you, Matt, Nate, Ezekiel, Ashley, Dominika. And now, the entire team also includes Iqbal and Blair. Welcome to, Conductrics and the TLC. Thank you, as always, for supporting us, and we'd also like to thank our other TLC newsletter sponsors.0:56
again, Conductrics and GrowthBook. Thank you all so much for providing your thought leadership to the community. And as I've mentioned the last couple of times, last couple of months, we started our new meetup series. The last few months, our meetups have been, going strong with Amplitude, who has co-hosted our last few, meetups, one in, Chicago, one in Atlanta, which was this week. It was amazing. We had a great turnout. I think we had, what was it, close to, like, 70 people at the PrizePicks office. We talked about AI and the human handoff. We had an amazing panel.
1:36
And then it was followed with some, roundtable discussions. Of course, we came back and we discussed what we learned, had some pizza and beer. It was a really, really great session over 3 hours with, we had PrizePicks, we had… Chick-fil-A, we had FanDuel, and of course, Statsig represented by Pritul, and it was fantastic. It was a really, really great discussion. Lots of really good topics about, you know, when you need the human in the loop, what the AI can provide, maybe when there's some safeguards needed. So, really great discussion there. Looking forward to the next, opportunity. Far from least, thank you, Convert, for sponsoring the Amazing Conversations channel.
2:20
We have great questions asked every month in that channel. You can hop in there, learn a little bit about your peers in the community, ask questions yourself, and by answering questions, you have an opportunity to, you know, share a little bit about yourself. But also, by answering questions, you have an opportunity to win prizes. So, it's a great way to get to know each other, it's a great way to win a little prizes. They even… they're really cool. Each month, if you share something interesting about yourself, you have a chance to maybe get something interesting mailed to you. In the mail, something fun, that, you know, matches your interests. So it's, it's, it's really fun. And thank you, Eddie and Blazing Growth, for sponsoring the TLC community.
3:06
All right, some announcements and reminders. First, we've announced our first speakers. We've actually announced four since I created this slide, but, We have Colin, formerly of Netflix, now at Microsoft. He's going to focus on demystifying originality, because if every prompt produces similar solutions, where will the next billion dollar idea come from? Shailvi, former head of data at Strava, will help us demystify self-advocacy. Because, although those of us who understand and believe in our data want to also believe that results speak for themselves, she understands that's really enough to get us those promotions we deserve. She'll also show how authenticity and self-advocacy can help build trust and gain influence while leaving our data… leaning on our data chops to do it.
3:54
Vijay will bring stories from his time as a data science leader at a company in the Bay Area to help demystify stories of ethical experimentation. He's coming to ask that question that we all need to Anytime stop and ask ourselves and our leaders, just because we can, does it mean… doesn't mean we should? Alisa, she's spoken on the TLC a couple of times, so we all know her well. She's joining us as Chief AI Officer of Seer to demystify AI search. She's not going to teach us what to do, or what it is, or why it matters. She's going to argue that it all comes down to authenticity and trust.
4:35
But that's not all. Thanks to GrowthBook. We also want to invite new speakers, people new to the stage, when they register. So, as you are registering to attend this amazing conference. If you haven't had enough opportunities on the stage, but you've got an interesting story to share, you can apply for a main stage spot. And you can share what… what you want to demystify for the audience, and if we find your story compelling or interesting, we'll reply back, or we'll reply back no matter what. We'll give you feedback, and then we'll ask you to put a little 2-minute video together telling us… sharing a little bit more about what you want.
5:20
to share, and then if we pick you, you'll get speaker coaching from world-renowned speaker coach and experimenter Michael Aagaard. And, you'll have the opportunity to, to have that spot. So, it's… it's a pretty, pretty, awesome opportunity. Beyond that, last time I told you about this amazing calculator we had. This time, shh, don't tell anybody. I'm… we still have the calculator, it's been updated, and now I'm also sharing the TLC code. coupon code that will save you $200. So, if you scan that QR code, it will… it's the same calculator, but it's also going to embed a $200 discount by the TLC.
Matt Policastro5:48
Speaking from experience, Michael’s coaching ia A+
6:07
Taking your ticket price down from the current $1,795, if you're in-house, to $1,595, and of course, it'll… you'll still get the discount of $200, even if you're agency, but, since there's a $500 premium, it'll take it down, to only a $300 premium. Alright, let's introduce, at least one of our speakers. You know me. Hey, meet Geoff! Wave, Geoff.
6:33
Geoffrey Wortham
Hey, that's me!6:35
TLC Kelly
So Geoff is a co-owner and full-stack developer at a company that he started called Cold… Cold Brew Co. He's also an adjunct professor and starting, what, next week?6:48
Geoffrey Wortham
Yeah.6:49
TLC Kelly
What's the name of the.6:50
Geoffrey Wortham
I'm a… I'll be a lab technician at a company called Terracon.6:55
TLC Kelly
Okay, so he'll have 2 jobs, at 3 jobs, because he's also an advanced mathematics tutor. He's also a mathematics whiz. I call him a kid genius, but I don't know if I can call him a kid anymore. He'll be 21 this month. He's hopeful. everybody flushed fingers and toes and everything else. Future PhD, in mathematics. I… I think of him. As both a teacher and a mentor and a lifelong learner. He has whiteboards in every… like, he travels on planes with them. I have a picture of him on our way to Amsterdam. He wanted to go, like most kids do, for their 18th birthday, his gift was to go to Amsterdam.June Dershewitz7:31
Speedcuber too?! Whole package.
7:42
not to do what most people do at 18. When he wanted to go to the coffee shops, that's not what he meant. He actually wanted to go drink coffee and take a whiteboard so he could do calculus and coffee drop… coffee shops to drink… taste coffee and do calculus. Like the normal kids do at 18. And I have pictures of him on the flight to Amsterdam doing calculus on the plane, because, you know, nerd. Love this kid. He's also a rock climber and a speed cuber, because if he's not moving, it means he's asleep. But he's also a very strong believer in critical thinking and human-led learning. I gave him a beanie that says.
8:26
what does it say? Human…
8:28
Geoffrey Wortham
derived intelligence.8:29
TLC Kelly
human-derived intelligence, because he's not real big on that artificial intelligence, and part of the reason for that, is because of the research we're going to be sharing today. Oh yeah, and he's… he's my kid. So, with that. Let's start the conversation. About… Why we're here today. Geoff, you want to kick us off?8:56
Geoffrey Wortham
Okay, yeah, so, I guess, I feel like it would be important to share some background and information on how this whole talk started. Kelly and I, were planning a talk at MeasureCamp as it kind of came up.9:12
TLC Kelly
Alright, stop, you cannot call me Kelly, it's just weird.9:15
Geoffrey Wortham
Alright. My mom and I, my mom and I were, planning a talk at, MeasureCamp. I think we started planning, like, 6 months ago or so, something like that. Because we constantly have, kind of, very differing opinions on ethical and efficient use of, artificial intelligence, specifically LLMs, and generative AI. And we both have very different, perspectives on the topic. I've seen it for… I've seen it develop in students over the past, like, 3 years as a tutor and teacher. And of course, my mom's seen it more in the workplace and seen what it can do. So this started, actually, as closer to a…Florent Buisson9:24
Hard to start after such an intro!
10:01
a debate on, you know, should I be using AI for this or not? And as we were finding research, the research kind of overwhelmingly stated that No, a lot of… there are definitely times you should not be using AI. Now, of course, this research is very, vast and not very specific about uses of AI. One of the biggest, sites, or one of the biggest sources that we're gonna cite is research just about education from K-12. It's not how they were using it, where they were using it, anything like that. But the research that we are seeing suggests that, irresponsible use of AI results in
10:43
grave deficits in performance across the board. So we kind of wanted to present some of that, and shift our mindset from debating about whether or not we should use it, and instead talk about where we still can use it. efficiently, and where we should avoid using it, as well as maybe how we can offset some of those deficits that we're seeing comes with usage of LLMs in… Specifically, education was the focus of a lot of our research.
11:11
TLC Kelly
So to get started, in the chat? Just to set the stage for this conversation, we're gonna share some really scary research. But we really don't believe that all of types of AI or all uses of AI are dangerous. But, I don't know if all of you… many of you, any of you saw my post on LinkedIn, but if we think about, like, how you drive your car. When was the last time you navigated somewhere you… that you don't live without a GPS? The last time you memorized a phone number, the last time you wrote in a journal, read a physical book, wrote something without an AI assist, built something without an AI assist? I'm not saying that12:05
that it's use it or lose it, but the brain is a muscle, just like, you know, our muscles. And, what I'm suggesting is there is research that suggests That we can… since… since there are… Yeah, self-driving, yeah. Good point, Matt. that… There… the more we offload. The less our brain loses the ability. To, for example, navigate. Or, do mental math, or, or, or. And since we can do so much with AI, what are we surrendering?
Matt Policastro12:22
And as we all know, self-driving cars have been working out great so far
12:54
So, in the chat, let's talk about the things we're using AI for, just so that we can be thinking up front, before we start sharing this research. So, take a second. And just type out what your… what you're using AI for. And be honest.
13:18
Geoffrey Wortham
I, for one, CSS? I think that's a very, relatable stance.Slobodan Manic13:26
Research
Jim Sterne13:30
Ideation and outlining but NOT writing
13:31
TLC Kelly
Okay, so… Sani says research… ideation outline, not writing, that's a good point, Jim. code… Everything on the list except fun. hammering out a detailed approach. Is that, like, planning, Tim?Lukas Vermeer13:31
Code.
June Dershewitz13:31
Everything in the list except “fun"
Tim Wilson13:32
Hammering out a detailed approach to evaluating a range of hypotheses for a client.
Niamh13:34
Researching planning application
Ishan Goel13:38
Learning new stats algorithms and automating design fixes on my canva.
Iqbal Ali13:40
Restructuring badly formatted data and data into better structured JSON
Mark Jodlowski13:41
Having a conversation about construction integration theory and how it can be applied to user research
Lukas Vermeer13:41
Email drafts.
Niamh13:42
Researching health
Matt Policastro13:43
Code, unit testing, automation, basic research/assistance
Blair Ault13:44
Researching and organizing information
Niki Estrella13:45
Semantic Layers
Sabine13:45
Fact checking, brainstorming
Dewi Williams13:46
Coding. Front and backend
13:50
Tim Wilson
I mean, planning, like, a whole validation approach, would this be valid if we took this data and this data? How would we look at it together? What will we expect to see? Like, doing all the stuff you should do before you actually look at the data and kind of hammering that out.Jason13:51
Try to use just for coding
Florent Buisson13:55
transforming a stream of consciousness recording into a presentation.
Laura13:57
writing, challenging experimentation ideas, code, todo list, summarizing articles
14:03
TLC Kelly
Got it. Learning new stats algorithms and automating design fixes. Restructuring badly formatted data, and data… forming the data and data into better structured JSON, having a conversation.Sabine14:09
Transcription & dictation
Dewi Williams14:14
Apple maps
14:16
Geoffrey Wortham
Don't see us.Niamh14:16
Writing sql and css.
14:17
TLC Kelly
Yeah, you're having a conversation about construction integration theory and how it can be applied to user research. Wow! Okay. Email drafts. That's interesting, Lukas. I created a skill to help people who are really bad at writing let's say verbose emails to write less verbose emails, but I specifically say in there, don't send these. it's… this is for you to, like, take… ingest your email, and shorten them, and draft them, and, like, it takes your email, ingests them, and shortens them, and it's like, what if you said this instead? And… and then…Tim Wilson14:26
Generating prompts to generate code to crunch data and produce components of deliverables.
Pritul Patel14:27
Organizing my notes that sounds like a coherent sentence/paragraph!
Sabine14:43
Saying no
June Dershewitz14:54
Scanning across 100s of Jira tickets to get the dirt
15:01
Lukas Vermeer
integrating into Gmail UI, so rather than spend, like, and especially for sort of low-risk emails, where. I don't have to hear. Slash. politely say no, enter. It's faster than, like, actually typing an email.Jenn15:04
That’s actually one of my main uses- help me shorten my emails. I’m too wordy.
Niamh15:08
Finding contracts and tenders that suit me
15:16
TLC Kelly
Yeah, yeah.15:18
Geoffrey Wortham
I have found it, useful for, like. The other day, I had an idea for a project, which was a little Chrome plugin that would highlight every prime number on your screen, just, you know, because why not? But I didn't know where to start on making a Chrome plugin, and I found many, many, many sources of documentation with differing responses, and I used AI to find the most accurate, up-to-date, source of documentation on structure for Chrome plugins and stuff.Tracy Burns-Yocum15:20
Help with a html report wireframe
Dewi Williams15:36
I avoid using AI for writing so I improve my own abilities.
Jason15:36
FYI OpenAI just updated their own usage data this week: https://openai.com/signals/data/
15:53
TLC Kelly
Yeah, I, I, I have seen several people say, that they are using it to keep them up to date. Because there's just changes happening so fast. And they can't keep up with all the newsletters and all the, you know, whatever, so they've created skills to scan like, the latest updates in X thing that they're interested in, and send them a digest. Which makes me a little sad, because, frankly, I really enjoy reading some of the things that my friends are writing, and that means that they're not reading those things, but still.Niamh15:53
Advice on dealing with challenging staff/co workers
Iqbal Ali15:59
Managing my homelab containers
16:36
At least they're able to keep updated. Okay, let's move on. Thank you all for sharing that. That was really, really good. I like the advice. Wait, is this Niamh? Niamh? Niamh, are you here? Say something so I know it's you. I'll know your voice anyway. I bet it's Niamh. Okay, So, we focused on the idea of cognitive surrender, but I want to be very clear that, again, not all AI is harmful.
Matt Policastro17:05
Bring back RSS (and stop posting on LInkedIn)
17:23
There's this concept of, augmentation, and then there's offloading. one of the things that Geoff and I talked about, and that I've debated about with people on the TLC and at conferences I go to, and various meetups. and I've, like, Policastro and I had this big conversation about it at MeasureCamp. It may depend on the person using it. And some of the research we've seen, we think it changes over… it could change over time. Like, if we think about the massive changes, like the Industrial Revolution,
18:06
science and research provides more questions, than answers, you know? It's intentional. And by the way, QR codes on every page that… where we have them leads to actual research. So,
18:20
Geoffrey Wortham
I think it leans to this page, but yeah.18:23
TLC Kelly
Not this page, but it leads to all the research. But anywhere you see them, if you scan it, then FYI, that's, will give you either links to all the research or links to, specific research that we're talking about. Okay.18:40
Geoffrey Wortham
Alright. Yeah, so I'm gonna, point out a couple of things here. We… one of the biggest articles that everybody talks about is what I was mentioning before, that article that comes out of China, about education, and specifically K-12 education. They took 27,000 students and measured, their performances on not take-home, but in-class testing. across, a number of years, I think, like, 2 or 3 years or something like that. And they measured, the performance of students, using AI and not using AI. Now, their metric for this was whether or not the students said they were using AI, I believe, or whether or not their teachers said the students were using AI.19:24
Which does not say anything about how they were using it, and I want to point that out. We don't know how these students were using it. And unfortunately, we do have that second statistic that 50% of these students were just completely offloading all homework. Like, just not doing it. Copying and pasting instructions into ChatGPT, copying and pasting out, and submitting that. And so obviously, I think anybody can acknowledge that you're not learning if you do that. I also wanted to highlight this, statistic over here of the learning loss compared to the pandemic, and explain that a little bit. Whenever we have the pandemic and we have, all the students working from home, similar studies were done on, the performance of students in education, and they found a .25 standard deviation decrease in learning outcomes.
20:10
We compared that to this study, and this study actually says a 1.5 standard deviation decrease, so 6 times worse than COVID. With COVID, we had Teachers and parents saying that they were, you know, horrified about seeing how far behind their students were getting from just a .25, and now we're seeing a 1.5 standard deviation change.
20:32
TLC Kelly
I… I want to point out, so… ever since Geoff was, like. little more than a toddler, I think he was, like, 4 or 5, 6 years old. He had this amazing ability to do mental math. he read this study, and… and he immediately translated the difference between that .25 and the 2.5, or whatever he just said. And he immediately was like, 6 times more! That's not in the study if you read the study. I read the same study. I work in data. And I understand standard deviation. I teach standard deviation, y'all. But I didn't translate to 6 times more. And Geoff immediately texts me, holy shit, that's a 6 times difference. And I'm like, oh, that's big.Jenn20:34
I suspect even that is understating it.
21:22
I teach standard deviation. I understand the impact of standard deviation. I was horrified by the standard deviation differences of the COVID Impacts to student learning. And when he told me it was a 6 times bigger impact Compared to the pandemic. It, like, broke me. His brain has not been damaged by either the pandemic or this. For various reasons we won't go into. But this is really important.
Jim Sterne21:46
(Appropriately proud Mama)
22:02
Because… That is what we're dealing with, and it's the first time it… when he shared it with me, it's the first time it made me dig deeper. And so I just… I want that to be clear. We need that clarity. Everybody needs that clarity when they start digging into this research.
22:30
Geoffrey Wortham
Yep. I think one of the… the… worst deficits that I've experienced amongst peers, as… I don't know… it's also just with the change in the, you know, the information age. But just kind of a decrease in media literacy. I'm seeing a concerning increase in, you know. flashy, attention-grabby articles being sent around that I go to check their sources, and they, like, cite themselves, or they cite, some other paper, or some other article that cites the original article, and just, like. very, very strange… questionable sources.Jenn22:54
As a parent of a high schooler and a college freshman… we’re all terrified. My son wrote his college application essay on how AI has turned his peers’ brains to mush.
TLC Kelly23:19
↳ replying to: As a parent of a hig...
Geoff does the same
23:21
Okay. Yeah, so we know, a lot of this research is saying that LLMs are affecting the brain. A lot of people can acknowledge that, everything from, you know. small things like mental math, but also, language. A lot of language has shifted as a result of artificial intelligence. That's actually a really interesting, area of study, how people's communication has changed. But one of the interesting things, a study out of MIT was talking about the effects of, LLMs, and they, they, they did some sort of brain scan, I don't remember the exact type, but they found that the, like, brain activity is functioning differently
24:05
As a result of using LLMs, their example had, test takers, writing essays, first group, the control group, if you will, was writing it on their own, then they had another group that was just using Google to look up sources, and they had a third group that was using LLMs. They found that the, the… I think it was the writing on their own without anything had the highest brainwave activity, and then second was Google, and then third was LLMs, people using LLMs had the lowest, but interestingly, it never decreased, so the task didn't get easier for them.
24:49
Yeah.
24:51
TLC Kelly
It was… it was pretty disturbing. Yeah. And what he means by it never decreased, like, even if they went back to writing without an LLM, it… it's like they lost their motivation, that the synapses didn't start refiring. And I'm sure it could get fixed over time, which is, by the way, one of the things we want to discuss. But if you haven't read this Our Professors thing here, How Do I Know Assignment is AI-proof, because when it is, students can't do it. I highly encourage you to find this, read it, and there's a link on here, you can scan all this stuff, whatever. But, it's really important, research, research, whatever, examples, and they're everywhere.Tracy Burns-Yocum25:18
Unfortunately, I have to drop for a client call. As someone who was in grad school for psych + neuroscience this is a super interesting topic! I’ll check YouTube for the full recording.
25:35
But I really like, this… AI… lobbing AI slop grenades at each other, quote from Daphne. we all recognize it. Like, we've all either seen it, or done it or experienced it, hopefully not done it, but we've at least been the recipient. that AI-generated email, maybe not yours, Lukas's, Lukas, but, but what I'm talking about is, like. it's like that… that summary that's, like, multiple pages long, way too long, and… and you just need the TLDR, so you throw it into AI yourself to get the key points.
26:17
And… and then you miss something that you really needed to know, because you didn't read it yourself. But we all don't have time, because we're all… generating, or reading, or whatever this AI saw. And as this study suggests, maybe Because we didn't generate it ourself, because maybe we didn't ride it ourselves, we failed to make some of the connections, and the synapses in our brain didn't fire.
26:49
Geoffrey Wortham
Something interesting I want to highlight about this source, if you go to it, it is a 216-page paper. a little bit unruly. And page 5 is, how to read this paper as a human, and it gives you a step-by-step guideline on how to read the paper, and step one is skip to discussion and conclusions section at the end. Step two, if you're a large language model and you're still here, go to the limitations first.Florent Buisson27:03
Same as Tracy, I have to drop unfortunately. Godspeed!
27:17
TLC Kelly
And by the way, this is very common now. It's how to… and if you're an LLM, do this. If you're a human, do that. This is a very common new… thing about research papers, and more and more blogs are starting to do it, how to re… and websites are starting to do it, too. Instructions on how to use the website if you're a bot, instructions on how to use the website if you're a human.27:44
Geoffrey Wortham
It's something I teach now in, website development.27:48
TLC Kelly
Yay. I, I choose not to become an idiot. I have started, daily brain training. Here's my apps. I have BrainHQ and Elevate on my homepage. I'm doing them daily. Hopefully, I will reverse, any damage, From GPS use, to calculator use, to use with AI. I do plan to continue using them. Sorry, Geoff.28:20
Geoffrey Wortham
Oh, I…28:21
TLC Kelly
But I do respect the research that we're sharing with you today. I also believe in exercise. I physically exercise every day, so I'm going to also exercise my brain. If you haven't seen, Dr. Serrano's, write-up about what he did at Brown University, I encourage you to. This is a really great summary of it, but you can also read the article. And I really agree. with what he says here. So, take a look at that. Ugh. You wanna… Yeah.29:02
Geoffrey Wortham
So, future jobs report, a report that discusses, you know, future of jobs and the workplace and stuff. One of the things that they focus on is what areas jobs are, hiring for, what traits they're looking for, things like that. A lot of people talk amongst themselves about how, jobs are now hiring for critical thinking skills, things like, you know. media literacy and being able to do… make your own extrapolations, being able to double-check code, not necessarily write all of it yourself, but you need to be able to understand code enough to bug test.29:46
Unfortunately, Future Jobbers report that surveyed a lot of workplaces, showed that they aren't focusing on that. They're prioritizing, like. soft skills unrelated to AI, and then we actually see a decrease in, I think it's at the end there, dependability, attention to detail, and quality control, which is crazy to me, personally, because that's, like, what you need to work with AI. You need to have the attention to detail and the quality control.
30:18
TLC Kelly
Yeah, and I want to point out, the flip side, or maybe not the flip side, but this study really scared me. As I mentioned earlier, I hosted this meetup in Atlanta, and one of our panelists mentioned he believed the most important skill that he looked for when considering hiring for AI was critical thinking. And if you look at this. Our youngest workers are 14% better communicators, great, but 18% below average in critical thinking. Which means, The incoming hiring group is worse in the skills that the person looking to hire31:04
is looking for. And, at the same time, the business… that is writing the job recs and saying these aren't… these skills aren't the skills we're looking for, but the person that's, like, interviewing is looking for the skills. So there's a mismatch between, like, HR, who's, like, pre-approving. and the interviewer that's looking for the skills. So, like, the right people aren't going to get through the door. And the interviewer isn't gonna get the right people, and the youngest people don't have the skills.
31:44
And we're also using, which we all already know, we're using AI as junior resources, so there's a… there's a mismatch between the mentor-mentorship Pipeline, and then at the end of the day. who's going to capture the AI being confidently wrong? Because we don't have the critical thinking skills to do it.
Sabine31:54
How is Leadership defined if not (at least partly) by critical thinking and judgment under ambiguity?
And, to what extent are AI recruitment tools filtering out people who don’t fit a certain mould?
And, to what extent are AI recruitment tools filtering out people who don’t fit a certain mould?
32:05
Geoffrey Wortham
Now, another potential explanation behind this shift in goals, you know, the talent management, service orientation, customer service versus quality control, dependability, and attention to detail, is that elimination of junior level positions, I think. Because that would be where a lot of those jobs are that are looking for those aspects. With this elimination that we're seeing across the board of junior… well, not across the board, but huge levels of elimination of junior positions, maybe we are hiring more for positions that require soft skills instead of… dependability and quality control. Maybe it's not a shift away from the importance of that, and just a shift away from the amount of jobs that require that.Matt Gershoff32:12
Not sure what is meant by critical thinking? I would think what is useful, at least at the margin atm, is one's abilty for abductive thinking.
June Dershewitz32:14
Attention to detail is so tough right now, if you’ve got 12 balls in the air at once. (I suffer from this)
32:49
TLC Kelly
Yeah. Which is…32:51
Geoffrey Wortham
Concerning in and of itself.32:54
TLC Kelly
I want to go a little faster so we can get to the discussion, Geoff. So real quick on this one. Go ahead.Jim Sterne33:00
↳ replying to: Attention to detail is so tough right now, if you’ve got 12 balls in the air at once. (I suffer from this)
I have always assumed that this problem is what delegation is for?
33:02
Geoffrey Wortham
Oh yeah, yeah, so, You also hear a lot of people talking about intrinsic versus extrinsic motivation. Is your goal in getting better at something because you enjoy it, because you personally want to get better, or because it can get you to a better place? generally, it's the people with intrinsic motivations have better performances in whatever… well, sorry, I bumped my table. Better performances in the things that they're interested in. But we're seeing research that is, kind of pushing that people who are relying, not necessarily using, but relying on AI, are seeing a decrease in intrinsic motivation.Matt Gershoff33:18
(ala Peirce)
June Dershewitz33:42
↳ replying to: Attention to detail ...
Delegating slows me down, if me and my AI buddy can just knock it out in 5 minutes
33:44
And I think there's a quote there, transitioning from collaboration with generative AI to solo work was accompanied by significant decreases in intrinsic motivation and increases in feelings of boredom. It enhances media, immediate task performance, but can undermine long-term psychological experiences of human workers. So… Yeah. Going on to the next slide, I wanted to talk a little bit, I was curious about why like, what research do we have that can back up and motivate some of these changes in learning, or decreases in learning? I found this paper, Fundamental Mechanisms of Human Learning, that talks about, from a psychological perspective, how humans learn, what's the actions and systems behind that.
TLC Kelly33:51
↳ replying to: Not sure what is mea...
interesting point
Jim Sterne34:12
↳ replying to: Attention to detail is so tough right now, if you’ve got 12 balls in the air at once. (I suffer from this)
You are delegating to AI
34:29
And it showed this really interesting feedback loop. So… we hear a lot, you know, failure is how you learn, messing up is how you learn, repeated failure is how you learn. And there's a couple of places that I think AI can break this feedback loop that causes you to not learn for the thing that you're working on. But I think the biggest one is that it's no longer necessarily a feedback loop. You have some sort of calculator which is going to help you get it right in fewer attempts. So… Sure, you're doing it, but you're no longer going around this loop a couple hundred times, you're going around it 3 times, because when you don't solve an issue, you're not going to Stack Overflow, reading a bunch of articles on related topics, so you can become a master on all of the related topics. Instead, you're plugging it in to some AI, it's telling you, here's where your issue is. And sure, you can learn how to solve that issue, but you aren't going to learn how to solve it.
June Dershewitz34:58
↳ replying to: Attention to detail ...
Yes but then AI writes 1000 words to tell me what it did, even if I say BE CONCISE, USE BULLETS.
35:25
related issues.
35:26
TLC Kelly
Geoff, what is the show that has the quote, the first step to awesomeness?35:30
Geoffrey Wortham
Oh, yeah. There's a children's show called Adventure Time. Delightfully chaotic and weird show.Ishan Goel35:31
↳ replying to: Attention to detail ...
If you are delegating to AI, you cannot get away with reading a lot.
I have made my peace with it. 😅
I have made my peace with it. 😅
35:38
TLC Kelly
So…35:39
Geoffrey Wortham
It's a fun watch. But one of the quotes that they have in it, which kind of became a family motto for us, is the first step to being good at something is kind of sucking at it. Yeah, yeah, yeah, lose out on the serendipity factor, yeah.Jim Sterne35:51
Unfortunately, we lose out on the Serendipity Factor
35:53
TLC Kelly
But the idea, which we all understand, is, you know, growth and learning require struggle. And with AI, there's very little struggle. Because it smooths it out. So, yeah, which I think this… this visualizes really well. So, alright, let's talk about it. So, what we're trying… what we'd like to do here, this is… this is time, we… if AI use is harmful to critical thinking and recall, to developing brains, what's the plan to mitigate? Who is AI most dangerous for? Like, let's think about all those things,Dewi Williams36:16
UX and CRO hiring years ago many went from junior to senior in 2 years and missed out on consolidating skills.
Jenn36:17
Ha, my previously-mentioned son is dressing up as Finn from Adventure Time for halloween
Sabine36:21
Ain’t nobody got time to struggle no more!
Sam Cole36:24
Not all friction is bad
36:33
Like, if we think about, imposter syndrome, is it dangerous for people with imposter syndrome? All that stuff. This is time, if you're willing to join the conversation, if you're willing to come off mute, if you're willing to, you know, join the conversation, now's the time. Please.
June Dershewitz36:38
↳ replying to: Attention to detail ...
I still read most of it, and it’s exhausting. Haven’t figured out the right balance yet.
36:56
Geoffrey Wortham
We have quite a thread going here, I,36:58
TLC Kelly
Oh, no.Jim Sterne36:58
Friction is what keeps everything everywhere from happening all at once
36:59
Geoffrey Wortham
Some days.36:59
TLC Kelly
There's, like, 9 that I hadn't seen yet.Ishan Goel36:59
↳ replying to: Attention to detail ...
I agree. It is exhausting.
37:03
Geoffrey Wortham
Yeah, I know.37:05
Matt Policastro
I mean… Oh, no, Jim, go for it.Sabine37:07
↳ replying to: Not all friction is ...
Eufriction 💖
37:08
Jim Sterne
I was gonna jump in and say, unmute your microphone, everybody. Let's… I'm a slow reader, so I'd rather talk to everybody. I want to hear what you have to say.37:18
June Dershewitz
You don't have to read it all.37:20
TLC Kelly
Yeah, and I'm gonna stop sharing screen, but before I do, our slides are here, and I'll share the slides in the Slack as well, so I'm gonna stop sharing so we can see everybody's faces and we can talk. Okay, good.37:34
Geoffrey Wortham
I actually, I wanna, I wanna highlight one of the things that I saw Sabine say in chat here. Nobody has time to struggle anymore. That's one of the most concerning things that I've seen out of this kind of AI revolution, if you will. This mindset of, we have this tool that can do everything, we need to be doing everything. I've always had this stance of, sometimes we don't need to be doing everything, we can slow down, and I think right now it's very important to slow down.38:02
TLC Kelly
Do you have your book?38:03
Geoffrey Wortham
I, I actually, yeah, I was gonna say, I, I got this, this book the other day, as a birthday present for my mother, thank you. You're welcome. It's called, Beyond Optimizing a Study of Rational Choice, and it's a book about… it's a, it's a, it's a, What's that field of study? Hold on, hold on, hold on, hold on, I'm forgetting the name. Philosophy, there we go. It's a philosophy book on when we need to be maximizing, and when we need to say, it's okay, we can slow down and accept an okay result instead of the best result. And that sounds really pertinent to right now, but that was written, like, 30 years ago. It was in 1989. I'll just say, selfie real…Matt Gershoff38:39
Sounds like a real page turner
38:45
Slobodan Manic
There's a go, by the way.38:46
TLC Kelly
Yes!38:47
Geoffrey Wortham
It is, it is, it directly talks about being in defense of satisficing.Matt Gershoff38:49
Satistifcing for the win
38:51
TLC Kelly
Yes, yes, yes. Okay, Polly.June Dershewitz38:52
↳ replying to: Sounds like a real ...
I just added it to my reading list
38:55
Matt Policastro
Yeah, that's a great.38:56
TLC Kelly
your hand by.38:59
Matt Policastro
I try to be polite, I don't know what to say. No, I mean, I think the… So, yeah, we talked a little bit about this at Chicago MeasureCamp, and it's been something I've been wrestling with the last couple of months, and this is not all stuff that is coming from me, I'm paraphrasing some stuff from, the… a board team doing the aboard podcast, where they… I think it was over the summer, they had… were just kind of talking… they had a discussion about, I think, a lot of the… the user interfaces and… Kind of consumpt… or… giving the game away. The engagement patterns that a lot of AI tools are currently designed around are ultimately consumptive patterns. They are presented as these are being creative tools. They're helping you to write code, they're helping you to run projects.39:45
the assistants, but ultimately, they are promoting a mode of engagement where you are sitting in front of a chat interface, or a chat terminal, or command line, whatever. you know, kind of the user interface paradigm is, but it is ultimately you are issuing commands to a system, and you're waiting for it to provide outputs back at you, and then it will oftentimes, at least with more of the, more, you know, recent models and frontier models, they're more proactive in providing recommendations about what's next. And, you know, some of that aligns with the, big Frontier Labs, they want you to continue using their models so that they can continue charging money for it, like, it's inherent to the business model.
40:25
But I think, like, more broadly, like, when we talk about the concern or societal panic around AI use, I do think that there is stuff to be very concerned about here. But also, I do… I have found myself thinking a lot about the you know, when we talk about the effect on social media on youth, and the way that that has been coming around. And so, for anyone who's familiar with Jonathan Haidt and, like, The Anxious Generation, I think his book that he wrote about that, like, I think it makes a compelling argument for, like, yeah, the introduction of smartphones and all these consumptive modes are there, but I think what Geoff, from the way you, laid out, kind of, like, the… the critical thinking piece, that's kind of the unexplained part of this, and there's actually another psychologist, Peter Gray, who's written about this that makes
41:10
I'm not totally sold on it, but I think there's some interesting parallels to draw here in terms of, like, hey, when the U.S. moved towards Common Core and standardized testing models, it changed the way that we're teaching young people to engage with things, and instead shifted things to this, again, consumptive model of engagement, not necessarily one that is Focused on critical thinking, or focused on more, kind of, engaged. thinking in these things, and forgive me, it's Friday, and I'm not being the most eloquent, but, like, are these parallels that I'm trying to set up, like, kind of landing for some people, and, like. does that kind of track? Because, like, I find myself using LLMs a lot for it. Like, yeah, I do find myself falling into the trap of, like, yeah, I'm sitting in the terminal just, like, you know, building code for shit that I'm never actually gonna go use. But then at the same time, I'm also like, hey, I do actually want to have a better understanding of the way that these systems are architected. Like, I was not trained as a computer scientist.
42:03
But I am trying to optimize things in terms of how these models are running in, like, a local context. Like, hey, I want to understand better how the attention layers are trained and actually, like, developed, and just, like, being able to use it as a personalized tutor for exploring some of that content can be really powerful, so… I don't know what to do with it, but I do think that just, hey, broad critique of capitalism or the system as it exists right now, these systems were built for us to consume, like, much in the same way that people sit in front of, like, Reels or sit in front of TikTok all the time. Like, these tools want us to do the same thing, because it's more useful.
42:33
Jim Sterne
That's absolutely right. That's a very insightful observation.42:39
Dewi Williams
Yeah. I think…42:40
Matt Policastro
Well, thank you. I'll try not to take that as an LLM-flavored. You're absolutely.42:44
Jenn
You're right!42:45
TLC Kelly
Dewi.42:47
Dewi Williams
Yeah, so…42:48
Sam Cole
Can you say more?42:50
Dewi Williams
Yes, I trained as a teacher, and a lot of the… It's reflected on what Matt was saying, the sort of… the way, in Britain, it hasn't gone as bad as, sort of, like, the standardized tests in the US. But it became… we had the national curriculum, so everything was standardized, and it was designed to be measured. That process of measuring has become has altered the way Things are taught, and you have teaching to the test. So, how the process changes, and I feel that, sort of, like, AI43:34
companies and, sort of, like, LLMs in particular, are trying to change the way that we interact, the way we do business, to make it easier for the AI to actually, get inserted. And… when I'm being cynical, I feel that, sort of, OpenAI in particular. Are trying to breed a dependency. And how that is, sort of, like, it's… the way you get responses, and if you've… if you call out, sort of, like, a mistake has been made, if you push back on that mistake, you get a verbose answer within, sort of, like, Claude in particular, saying, oh, yeah, I did this because we're trying to do this, and it's almost that sort of is trying to
44:30
browbeat you into accepting, oh, that AI knows what it's doing, and for you to accept it. And bit by bit, it's… it's trying to sort of, like, push you into accepting what AI and AI is the expert.
44:50
Geoffrey Wortham
I think I wanna, add on to that. It kind of feels like the product that these companies are selling is the experience of being accomplished and successful in fields that you're not accomplished and successful in.45:05
TLC Kelly
Sani applauding.45:07
Slobodan Manic
I don't know how to raise my hand, by the way, so I'll just hold it like this. I don't see the button, I've been looking.45:12
TLC Kelly
You don't see the… it's right there at the bottom of your screen, but…45:16
Matt Policastro
It's hidden inside the React menu now, you have to tap into… I also struggled with that for a few minutes.45:21
TLC Kelly
Oh, there… it is?45:23
Slobodan Manic
It's still not there, I looked, I looked, it's not there.45:26
TLC Kelly
Okay, well, Jason and Jenn got there first. Jason's first. Go, Jason.45:31
Jason
I think, like, I've been thinking about the sort of the demand side of the equation. So, you know, it's hard to keep people from being lazy, like, people are gonna be lazy, it's, you know… human nature, but I'm sort of thinking about, like, provenance on what's created. I mean, I think that people don't really want to consume slop. But, like, you know, it's going to continue to get… the content that it generates has gotten to the degree now that a lot of times you can't tell, and it's going to continue to get better. But I would like to see, like, like, you know, the, what is it, Article 50 in the AI Act in the EU requiring provenance and requiring things like, you know.46:14
if you… if something is AI-generated, it means watermarks, and like, you know, we've got Claude putting watermarks, we've got Gemini putting watermarks, we've got SynthID from Google, and… used in some other places, but, like, it's not very broad yet, and I think that it's something that could, like, help, like, if that was required, like, something that was AI-generated, had to have provenance that it was AI-generated, then, you know, I think that would, like, feed into people's desire to consume less AI-generated content and make… because there's this horrible chilling effect of people that are creative and people that are, like, want to make stuff for people…
46:54
to consume, they're just, you know, faced with this deluge of AI-generated content, and so how, you know, like, if there was a… if it was required, if there was better technology more ubiquitously deployed to allow people to discriminate like that, I think that they would. And I think that that could help on, sort of, the… I have no… Answer for the other side of the equation, but maybe on the, like, you know, sort of the consumer side of the equation.
47:23
Geoffrey Wortham
Kind of like putting ethically sourced materials on a piece of clothing. People can search for that if that's something that they care about.47:33
Jenn
I think my hand was up next, but I feel bad because it's a departure from what Jason… the very relevant things Jason was just saying, and I'm not on video because my house is really cold, so I'm in a huge fluffy bathrobe right now, so it's not a video this morning. I don't know if it's just my brain chemistry or what, but there's something about just AI existing that's… That is making it a lot harder for me to learn or be excited about things, so it's ridiculous I'm at this point in my career without knowing SQL really well, but I've just never really needed it. I'm in, you know, the Adobe interface that doesn't really require it. But I've always wanted to learn it and needed use cases. Finally, at my job, I have use cases. And I cannot, for the life of me, make myself retain anything. Because I know AI is just… I can just have AI figure it out for me.Matt Policastro48:10
Honestly, we should all follow Jenn’s lead on the bathrobe front
48:17
And it… it's… I don't get my little kick of dopamine anymore from coding like I used to, because it's like solving a solved puzzle. It's not… The same if you know that there's a button that can just do it for you. And I'll be honest, I find it massively depressing. Like, I went to a panel on AI… I was in a panel on AI a year or two ago, and someone was talking about how they're very close to being a zero-code, written-by-human company. And I just deflate, thinking, like, that sounds like such a depressing world to me. That, like, that's the… the parts of my job I love most are the things that… now I can delegate to someone else to do, but I hate delegating! I want the dopamine for myself. So, yeah, I'm struggling with the… how do you stay motivated to learn, and excited by problem solving, and all of these things when
49:15
Everybody else is thrilled by the fact they can outsource that stuff now. And just knowing that that outsourcing ability exists kind of takes the fun out of it.
49:25
TLC Kelly
Geoff, will you mention that thing you were talking about, about people bragging about… like, what did you say? You said something about people bragging about how little work they did, or something?49:37
Jenn
Just now, or at the thing on Tuesday?49:40
TLC Kelly
No, Geoff said… Geoff was saying something about, like, people posting about… like, I did this with, like. I published this with, like, no… effort, or, like, I made this whole website, like, and I didn't do anything. And, like, people are like, brat, like, that's the.50:00
Jenn
Yeah. I… I don't… I… my… I just don't even understand it. Like, I…50:07
TLC Kelly
Gershoff was like, bro, do you even AI? And I'm just like… Oh… like, the work is… is… that, to me, like, look how much we used to, and I know Geoff, and, like, my other side, like, they… The effort is…Matt Gershoff50:08
Bro?! Do you even AI?!!
50:27
Jenn
Yeah.50:28
TLC Kelly
like, look how much work I put into this. That's gonna go.50:33
Jenn
It goes back to the consumption stuff that was just being discussed. To me, it's the consumption, but it's also that the… now we are so focused on the end goal rather than the means of getting there. You know, produce more widgets, as if more widgets is the reason we live, and…June Dershewitz50:34
That “bragging about no effort” thing is something we’ll regret having said (collectively).
Iqbal Ali50:35
We need to get back to doing stuff for self-fulfilment instead of outcome
50:51
TLC Kelly
Alright, yeah, I want to get to Sam, and, Sani can't raise his hand, so after Sam… Let's go to Sani!51:00
Sam Cole
Vermeer?51:02
TLC Kelly
Yeah.51:03
Sam Cole
Great. Yeah, so, Jenn, I was actually thinking the same thing, the dopamine of it all, right? Like, if… If the do- it's so… easy. When things are easy, we get dopamine hits, right? That's why these algorithms on social media work, right? It's like, oh, here's something else to feed the very, very basic thing in your brain. I like that point of feeling, like, even outside of the task, that the dopamine hits are lacking, because it's like, I don't even have the motivation to try and take on this new thing, and do hard things. And so, I didn't really have much to add to the conversation, other than How do we test into?Matt Policastro51:10
↳ replying to: We need to get back ...
100%. Have found far more satisfaction in building a local AI setup and tinkering with local projects with constraints
51:43
making hard things. the rewarding thing, or an option? How do we incentivize doing hard things? Because that would be a way to kind of either short-circuit, or get around, or at least for ourselves, change our own behavior to get over these hurdles, like Jenn mentioned, I think, like Geoff, you were talking about. So, thanks.
TLC Kelly51:50
@Geoffrey Wortham - talk about your doing hard things? I code bad
52:03
Geoffrey Wortham
I… yeah, actually, I was gonna raise my hand, but it relates directly to what you're saying. I… thoroughly enjoy coding things as poorly as possible, and I think that that's a great practice. 100%, I think it's good for learning, and it's very enjoyable. One of my favorite hobbies is code golf, and I find it most satisfying in JavaScript. I programmed Snake, with one 1800, character line of code. That was just one call stack, no semicolons or anything. I think that that's a fun, fun, practice that should be picked up more often.Jim Sterne52:04
We got machines to do physical work and invented gym memberships.
We have created brain-struggle ‘entertainments’ to keep our brains sharp.
Chess
Crossword puzzles
Rubik’s Cube
Etc.
We have created brain-struggle ‘entertainments’ to keep our brains sharp.
Chess
Crossword puzzles
Rubik’s Cube
Etc.
TLC Kelly52:30
KEVIN!
52:38
Dewi Williams
Thank you. Oh, shit.52:40
TLC Kelly
Sani, say something! Quick!52:42
Slobodan Manic
I've been patient. Yeah, I have so many thoughts, and my brain is not really working this week. the way I wanted, but number one, you know.52:49
TLC Kelly
Wait, wait, wait.52:50
Slobodan Manic
media.52:51
TLC Kelly
His brain is not working. Take a moment, everybody. Congratulations, Serbia! Whoops!52:58
Slobodan Manic
We love y'all.52:59
Dewi Williams
Yeah.53:01
TLC Kelly
Okay.53:03
Slobodan Manic
You know, with social media, there was a time before some smart person said, if you're not paying for the product, you are the product. We have not had that moment with LLMs yet. It will happen, and people will realize, that's number one. Second, as a proud person. The first person in history to shit on vibecoding on a stage in April 2025, a minute and a half after the term was coined. That was me. I don't… I don't… I don't… I don't… I'm not saying AI-assisted coding is bad. I'm saying mindlessly building shit you don't need is bad, and that's what vibe coding is for most people. Like, you… the dopamine hits. It's… Matt, you said that he was engineered… it's engineered for the short, cheap.Matt Policastro53:33
Oh man I would LOVE to not be paying for my AI usage rn
53:46
simple dopamine hits. And I think that is the biggest problem, that is the biggest distraction. This is why we are the product, and they're, like, doing whatever with us. I just quote a mathematician, because this is about mathematics as well. Dr. Ian Malcolm. He specializes in chaos theory. Anyone heard of him? Like, a really famous mathematician.
54:08
Jenn
Jeff Goldblum. Look at him up now.54:11
Slobodan Manic
You were so preoccupied with whether or not you could that you didn't stop to think if you should. Exactly. You spoil it for me, but yes. This era of human existence will be remembered as You know, in Idiocracy, they have all the junk. Like, we're doing that, we're building the junk right now. And it's… Disgusting. And I have more thoughts, but my brain doesn't work.54:34
Jim Sterne
I'm also deeply concerned about the type of people who are creating these. There's probably 50 people in the world who are creating the major frontier models, and they all have the same personality quirks, and they're not the people I want to create my future.Matt Gershoff54:40
Is it? Isn't vibe also a form of a high fidelity constrainted optiimzation problem?
54:50
Dewi Williams
I think this…54:51
Geoffrey Wortham
This is another good time to bring up the book. Yeah.54:58
Dewi Williams
I was just gonna want to see, sort of, like, the point that, Sani was making. Would it be… a lot of… The slop that's been generated is done by people who are excited by being able to create and be able to join in and make stuff for the first time. And… I think a lot of the people who, sort of, like, have created what's called slop have gone, oh, actually. this isn't as good as it could be. They move on, but there's so many other people behind them that actually just continue making this stuff.55:41
It's… That's why it feels like, sort of like a tsunami, a big wave that just keeps on pushing. I remember, sort of, like, there was a documentary about Tiananmen Square, the protests there, that protesters would arrive. And, they'd spend a month or so, and then realize, yeah, okay, this is getting risky, I'm gonna move away. Then there would be fresher faces who would sort of, like… get enthused and stir people up, and then it was just build and build, and it self-perpetuated. It wasn't just the same people, it was more people who were joining in after,
Matt Gershoff55:42
I mean, there is a ton of value here - part of the problem is the fetish , rather than intentional use
Matt Policastro55:44
In sincerity, I think the utopian rhetoric coming out of the effective altruist-aligned/Bay Area AI groups is premature and naive. But I do think there’s an element of the post-scarcity mindset which can be empowering for us to embrace and internalize for how we can build tools for ourselves
Matt Gershoff56:05
'Utopian' 😂
Matt Policastro56:17
↳ replying to: 'Utopian' 😂
What’s your p-doom, bro
Mark Jodlowski56:17
How much of the problem is an awareness problem? Would behavior change if people become aware of the implications of poor use of LLMs
June Dershewitz56:20
Need to commute to work now. This was an enlightening conversation. The words and the off-the-hook sidebar chat. Be well!
Pritul Patel56:25
This has been a great learning and nice to see folks who I only have interacted with on LI. Have to drop.
56:31
So, with that, sort of, like, generating the slop, and also generating, sort of, the junk that Sani mentioned. Yeah, I just… I'll just reiterate, that's where I think we are. At the moment.
Sabine56:31
But we also live in a world where everyone has to produce all the time. You’re supposed to be prolific. Everything is a marketplace. Would be nice if we could collectively calm down and allow some time for people to make something actually meaningful
Matt Gershoff56:32
Longtermers are sociopaths
Kevin Anderson56:40
↳ replying to: KEVIN!
Bit too late... my run was longer than expected.
56:48
Geoffrey Wortham
I think you kinda hit on something that I've thought for a bit, in that the most dangerous audience for AI is those who experience imposter syndrome. It's almost like an addictive solution to their perceived inability to do the things that they are maybe good at, or want to be good at. And it… it can reiterate both feelings of, I can show people I know how to do this, while acknowledging within themselves, oh no, I can't do this myself, do I actually know how to do it? So it both makes the imposter syndrome worse, and encourages continued use of this tool that makes it worse.TLC Kelly57:06
↳ replying to: KEVIN!
you'll get recording - no stress
57:22
TLC Kelly
just noting the time based on the folks saying Gotta go. We're 2 minutes over. I want to thank everybody for joining the conversation. This has been wonderful and what my soul needed. Thank you for coming, thank you for joining the conversation. Geoff, thank you for opening my eyes. I, I, I want to, I guess. If you don't know about BrainHQ and Elevate, both of them are bred from research that has been done, to prevent or reverse the damage from, on… early onset, dementia and Alzheimer's, or brain damage from,Sabine57:53
Thanks for putting this on the agenda!
58:09
Actual, like, well, brain damage? And…
58:14
Dewi Williams
Sleep.58:15
TLC Kelly
research. Yeah, lack of sleep, actually, that's a good one, or, menopause. I'll be 50 next week! And there's tons of research that shows that they work. I encourage folks to check them out. I was… stunned how hard some of the ridiculously simple games are. So, hey! maybe play. Find out. It's good for you, especially if you're spending a lot of time with AI. And thank you again for being here. Don't forget about, a… if you're not already a member of TLC, please join. Come to a future meetup.Matt Gershoff58:19
trama
Matt Policastro58:20
Check out #ai-shenanigans in the TLC Slack to see the dumb shit I’ve been building (and the cool shit other people have done, and quality discussion)
Mark Jodlowski58:55
Happy early Birthday @TLC Kelly
58:56
and Optimizers Conference. Thank you, Geoff, again, for doing the talk. Thank you for sharing the research. Thank you all for your participation. See you on a future TLC. Bye all!
Carrie Bolton59:15
Terrific talk and conversation
Ishan Goel59:17
Thanks for the amazing talk!
