WEBVTT 00:00:00.000 --> 00:00:01.860 When we first started hearing about generative 00:00:01.860 --> 00:00:04.679 AI everywhere, the story was pretty clear, wasn't 00:00:04.679 --> 00:00:08.699 it? This amazing new tool for, well, for corporations. 00:00:08.880 --> 00:00:11.380 Yeah, absolutely. We all pictured, you know, 00:00:11.560 --> 00:00:13.699 businesses getting huge productivity boosts, 00:00:13.800 --> 00:00:16.660 coders, marketers, analysts, making everything 00:00:16.660 --> 00:00:19.140 faster. Using it as a purely professional thing. 00:00:19.709 --> 00:00:22.370 But that picture, it's almost completely flipped 00:00:22.370 --> 00:00:24.469 now. We're diving into this really fascinating 00:00:24.469 --> 00:00:27.789 analysis today. They looked at how 700 million 00:00:27.789 --> 00:00:30.429 people are actually using ChatGBT in the real 00:00:30.429 --> 00:00:33.109 world. And the results are? Well, they're pretty 00:00:33.109 --> 00:00:36.170 surprising. Yeah, profound is a good word. It 00:00:36.170 --> 00:00:38.429 turns out, for almost three -quarters of users, 00:00:38.770 --> 00:00:40.990 AI isn't really that corporate tool. It's more 00:00:40.990 --> 00:00:43.810 like a personal friend or a tutor, maybe even 00:00:43.810 --> 00:00:46.009 a travel agent sometimes. Welcome to the deep 00:00:46.009 --> 00:00:48.820 dive. This study gives us this incredible window 00:00:48.820 --> 00:00:51.600 into what 700 million people are doing with AI. 00:00:52.340 --> 00:00:55.280 And it shows it's changing not just what we get 00:00:55.280 --> 00:00:57.759 done, but really how we do things, how we learn, 00:00:58.060 --> 00:01:00.359 how we edit our own writing. Even how we ask 00:01:00.359 --> 00:01:02.600 for advice on really personal stuff, like relationships. 00:01:03.079 --> 00:01:05.780 So our mission today is to unpack these real 00:01:05.780 --> 00:01:08.099 habits. We want to look at the shifts, especially 00:01:08.099 --> 00:01:12.120 in professional skills, and... Maybe touch on 00:01:12.120 --> 00:01:14.680 some of the risks. There's this one big ethical 00:01:14.680 --> 00:01:17.260 flag they raised called decision laundering that 00:01:17.260 --> 00:01:19.379 we definitely need to talk about. Okay, let's 00:01:19.379 --> 00:01:21.040 get into it. Where do we start? That personal 00:01:21.040 --> 00:01:23.840 takeover seems huge. It really is. So finding 00:01:23.840 --> 00:01:28.099 number one, the split. Only about 27 % of all 00:01:28.099 --> 00:01:30.280 the chat interactions they analyzed were actually 00:01:30.280 --> 00:01:33.640 for work. Just over a quarter, wow. Which means... 00:01:33.310 --> 00:01:36.530 The vast majority, 73%, is all personal stuff. 00:01:36.569 --> 00:01:38.689 People exploring hobbies, learning new things, 00:01:38.829 --> 00:01:41.129 managing their daily lives. It's become this 00:01:41.129 --> 00:01:44.349 kind of ultimate personal assistant. And it happens 00:01:44.349 --> 00:01:46.549 so fast. As Sadi mentioned, just a year ago, 00:01:46.810 --> 00:01:49.629 personal use was around 53%. Right. From barely 00:01:49.629 --> 00:01:52.409 half to almost three quarters in just 12 months. 00:01:52.609 --> 00:01:55.390 That's a really rapid, fundamental shift in how 00:01:55.390 --> 00:01:57.510 we're integrating this tech into our lives. So 00:01:57.510 --> 00:02:00.310 why? Why are we suddenly outsourcing so much 00:02:00.310 --> 00:02:04.629 of our curiosity and daily admin? to an AI? Well, 00:02:04.790 --> 00:02:07.450 a few reasons probably. The AI feels like this 00:02:07.450 --> 00:02:09.330 incredibly knowledgeable friend who's always 00:02:09.330 --> 00:02:14.090 there, 3 a .m., no problem. And crucially, there's 00:02:14.090 --> 00:02:17.050 no judgment. Ah, that's a big one. Yeah. If you 00:02:17.050 --> 00:02:20.509 feel kind of silly asking a coworker, what exactly 00:02:20.509 --> 00:02:23.110 is inflation again? Or, you know, how do I fix 00:02:23.110 --> 00:02:26.930 this leaky faucet? Uh, the AI just answers. No 00:02:26.930 --> 00:02:29.030 eye rolling. It takes away that social friction, 00:02:29.250 --> 00:02:31.370 that little bit of anxiety about asking something 00:02:31.370 --> 00:02:33.409 basic, and I guess the corporate side plays into 00:02:33.409 --> 00:02:35.900 it too. companies being nervous about data leaks. 00:02:36.240 --> 00:02:38.020 Definitely. There's still a lot of hesitation 00:02:38.020 --> 00:02:40.680 about employees pasting sensitive company info 00:02:40.680 --> 00:02:43.860 into these public tools. So that naturally pushes 00:02:43.860 --> 00:02:46.080 some usage towards personal accounts, personal 00:02:46.080 --> 00:02:49.099 questions at home. OK, so what are people actually 00:02:49.099 --> 00:02:52.020 doing in that big 73 % slice? What are the main 00:02:52.020 --> 00:02:54.599 personal uses? The top three are pretty foundational. 00:02:54.780 --> 00:02:57.460 First is just seeking information. But people 00:02:57.460 --> 00:02:59.520 want direct answers, not just a page of links 00:02:59.520 --> 00:03:01.539 like Google gives you. Right, like, give me a 00:03:01.539 --> 00:03:03.360 quick healthy recipe that doesn't use chicken. 00:03:03.340 --> 00:03:06.819 not here are 10 recipe blogs. Exactly. Second 00:03:06.819 --> 00:03:10.659 is writing, helping with emails to friends, social 00:03:10.659 --> 00:03:12.860 media captions, maybe even writing a little poem 00:03:12.860 --> 00:03:15.439 or something, polishing personal communication. 00:03:16.180 --> 00:03:18.900 And third is practical guidance, just straightforward 00:03:18.900 --> 00:03:21.919 how -to questions. How do I plan a three -day 00:03:21.919 --> 00:03:24.719 trip to Paris? Or how do I change a bike tire? 00:03:24.840 --> 00:03:26.719 That kind of thing. It really feels like we're 00:03:26.719 --> 00:03:30.300 offloading that initial. Drudgery the basic research 00:03:30.300 --> 00:03:32.740 phase of figuring things out. Yeah, which brings 00:03:32.740 --> 00:03:35.520 up a really interesting point How does this massive 00:03:35.520 --> 00:03:38.639 shift towards personal AI use? Change how we 00:03:38.639 --> 00:03:41.520 even think about productivity day -to -day. That's 00:03:41.520 --> 00:03:44.120 a good question I guess I mean these productivity 00:03:44.120 --> 00:03:46.360 isn't just about work output anymore. It's also 00:03:46.360 --> 00:03:48.460 about learning things faster in your personal 00:03:48.460 --> 00:03:51.300 life We're just handling those mundane life tasks 00:03:51.300 --> 00:03:53.080 more efficiently. So you have more brain space 00:03:53.080 --> 00:03:55.340 for other things Yeah, basically freeing up time 00:03:55.340 --> 00:03:57.719 by handling the small stuff faster. Okay, so 00:03:57.719 --> 00:04:00.949 let's pivot back then. What about the 27 % that 00:04:00.949 --> 00:04:03.530 is for work? What's happening inside that professional 00:04:03.530 --> 00:04:06.449 bubble? Right. So within that 27%, the biggest 00:04:06.449 --> 00:04:09.770 chunk, maybe unsurprisingly, about 40 % of all 00:04:09.770 --> 00:04:11.770 work -related tasks have something to do with 00:04:11.770 --> 00:04:14.650 writing. Makes sense. Lots of emails, reports, 00:04:14.870 --> 00:04:17.009 memos. But here's the really interesting twist. 00:04:17.410 --> 00:04:21.120 When people use AI for writing at work, Two -thirds 00:04:21.120 --> 00:04:24.860 of the time, about 66%, they're asking it to 00:04:24.860 --> 00:04:27.279 edit or change text they've already written themselves. 00:04:27.920 --> 00:04:30.060 So they're not just saying, write me a report 00:04:30.060 --> 00:04:33.060 on Q3 sales. Exactly. They're not asking for 00:04:33.060 --> 00:04:35.759 new stuff from a blank slate nearly as often. 00:04:36.079 --> 00:04:38.180 It seems people have caught on pretty quickly 00:04:38.180 --> 00:04:41.959 to what some call AI slop. AI slop? OK, define 00:04:41.959 --> 00:04:44.879 that. It's that, you know, generic, robotic -sounding 00:04:44.879 --> 00:04:47.360 text the AI spits out if you ask it to write 00:04:47.360 --> 00:04:49.639 something complex from scratch. It often likes 00:04:49.639 --> 00:04:52.459 real context, sounds bland, it usually needs 00:04:52.459 --> 00:04:54.399 a ton of editing anyway. Right. It doesn't sound 00:04:54.399 --> 00:04:56.860 like you or understand the nuances of your specific 00:04:56.860 --> 00:04:59.120 situation. Precisely. So the smarter approach, 00:04:59.199 --> 00:05:01.680 the one people seem to be adopting, is you write 00:05:01.680 --> 00:05:03.839 the first draft, you put your own knowledge, 00:05:03.959 --> 00:05:06.060 your tone, your context in there. You anchor 00:05:06.060 --> 00:05:08.259 it. Yeah, exactly. And then you ask the AI to 00:05:08.259 --> 00:05:10.240 act like an editor, clean it up, make it more 00:05:10.240 --> 00:05:11.759 concise, check the grammar, maybe make it sound 00:05:11.759 --> 00:05:14.079 more professional. You provide the core, the 00:05:14.079 --> 00:05:16.319 AI provides the polish. They give some examples 00:05:16.319 --> 00:05:18.589 in the study, right? Like, for a sick email? 00:05:18.670 --> 00:05:20.529 Yeah, a good one. Instead of just write a sick 00:05:20.529 --> 00:05:22.949 email, the user drafts something simple like, 00:05:23.370 --> 00:05:25.810 hey boss, woke up feeling rough, can't come in, 00:05:25.829 --> 00:05:28.189 we'll check email later. Then they feed that 00:05:28.189 --> 00:05:30.810 to the AI and say, make this sound more professional 00:05:30.810 --> 00:05:32.870 and formal. Okay, that makes a lot of sense. 00:05:33.490 --> 00:05:36.589 It's using the AI's strength refining language 00:05:36.589 --> 00:05:38.889 without relying on it for the core message or 00:05:38.889 --> 00:05:43.029 context. I have to say, I still rattle with prompt 00:05:43.029 --> 00:05:46.360 drift myself sometimes, you know. Trying to get 00:05:46.360 --> 00:05:48.980 the AI to generate exactly what I want from scratch. 00:05:49.399 --> 00:05:51.720 It can be tough to keep it on track. Using it 00:05:51.720 --> 00:05:53.439 as an editor after I've written the main points 00:05:53.439 --> 00:05:56.540 feels, well, way more effective usually. Yeah, 00:05:56.540 --> 00:05:58.920 it avoids that whole battle of trying to steer 00:05:58.920 --> 00:06:01.579 its massive generalization engine. What else 00:06:01.579 --> 00:06:03.519 are people doing for work? Well, summarizing 00:06:03.519 --> 00:06:07.300 is big. Taking a long report, say 20 pages, and 00:06:07.300 --> 00:06:09.439 asking for the key bullet points, that's super 00:06:09.439 --> 00:06:11.560 useful. Oh yeah, definitely. Brainstorming, too. 00:06:11.920 --> 00:06:14.339 Like, give me 10 creative ideas for a new coffee 00:06:14.339 --> 00:06:16.649 shop. getting that initial list to react to. 00:06:16.930 --> 00:06:18.350 What about programming? I thought that would 00:06:18.350 --> 00:06:20.649 be higher. Surprisingly low, actually. Only about 00:06:20.649 --> 00:06:23.829 4 .2 % in these general chats. The thinking is 00:06:23.829 --> 00:06:25.990 that coders are probably using more specialized 00:06:25.990 --> 00:06:28.569 tools, like GitHub Cotilot, that are built right 00:06:28.569 --> 00:06:31.389 into their workflow. Ah, OK. That makes sense. 00:06:31.509 --> 00:06:33.910 Different tools for different jobs. So if the 00:06:33.910 --> 00:06:36.629 key professional skill is becoming editing and 00:06:36.629 --> 00:06:40.209 refining AI output rather than just pure creation, 00:06:41.129 --> 00:06:43.829 what does that mean for human creativity? especially 00:06:43.829 --> 00:06:46.629 in that first crucial drafting stage. It suggests 00:06:46.629 --> 00:06:49.290 the human role is really shifting. Maybe less 00:06:49.290 --> 00:06:52.110 about being the initial author from zero and 00:06:52.110 --> 00:06:54.569 more about being the final judge, the curator, 00:06:54.970 --> 00:06:56.889 the one who adds the essential context and makes 00:06:56.889 --> 00:06:58.750 the final call. Okay, let's talk about where 00:06:58.750 --> 00:07:01.810 the lines get blurry between tool and something 00:07:01.810 --> 00:07:04.029 else. The study mentioned teaching and learning. 00:07:04.230 --> 00:07:06.810 Yeah, about 10 % of usage falls into that category. 00:07:07.189 --> 00:07:09.089 And there's a good side and a worrying side here. 00:07:09.209 --> 00:07:11.649 Okay, the good. The good part is the AI can be 00:07:11.649 --> 00:07:14.110 an incredibly patient tutor. You can ask it to 00:07:14.110 --> 00:07:16.230 explain something complex like photosynthesis 00:07:16.230 --> 00:07:18.730 over and over in simpler terms, like you're 10 00:07:18.730 --> 00:07:21.589 years old, and it won't get annoyed. That's pretty 00:07:21.589 --> 00:07:24.290 powerful for self -learning. Definitely. But 00:07:24.290 --> 00:07:26.589 the worrying part, and this is especially true 00:07:26.589 --> 00:07:29.550 for students, is the hallucination problem. Right. 00:07:29.769 --> 00:07:31.829 We should quickly define that. AI hallucination 00:07:31.829 --> 00:07:34.209 is basically AI makes up confidence sounding 00:07:34.209 --> 00:07:36.790 stuff that is factually wrong. Exactly. Like 00:07:36.790 --> 00:07:39.310 that funny example they use the AI inventing 00:07:39.310 --> 00:07:42.009 a fictional John Avocado as the inventor of avocado 00:07:42.009 --> 00:07:44.870 toast back in the 50s. It sounds plausible, but 00:07:44.870 --> 00:07:47.009 it's completely made up. And if students just 00:07:47.009 --> 00:07:49.439 accept that without checking. They risk learning 00:07:49.439 --> 00:07:52.100 things that just aren't true. It trains a habit 00:07:52.100 --> 00:07:54.100 of acceptance rather than critical thinking, 00:07:54.319 --> 00:07:57.360 which is... not great. Yeah, definitely not. 00:07:57.720 --> 00:07:59.660 What about the more social uses? This is where 00:07:59.660 --> 00:08:02.500 it gets even blurrier. About 2 % of messages 00:08:02.500 --> 00:08:05.180 were people asking for relationship advice. Wow. 00:08:05.519 --> 00:08:09.449 Asking AI about... Boyfriend problems, family 00:08:09.449 --> 00:08:11.990 issues. Yeah, things like that. And the why is 00:08:11.990 --> 00:08:14.089 probably similar to asking basic questions. It's 00:08:14.089 --> 00:08:15.850 always available and there's zero judgment. You 00:08:15.850 --> 00:08:17.410 don't have to worry about burdening a friend 00:08:17.410 --> 00:08:20.230 or feeling embarrassed. But the risk there seems 00:08:20.230 --> 00:08:23.850 really high. Enormous. AI doesn't have feelings. 00:08:24.009 --> 00:08:26.610 It doesn't understand love or grief or jealousy. 00:08:26.970 --> 00:08:29.990 It just recognizes patterns in data about how 00:08:29.990 --> 00:08:31.990 humans talk about those things. So the advice 00:08:31.990 --> 00:08:35.500 is going to be generic. based on averages, not 00:08:35.500 --> 00:08:38.500 on your specific nuanced human situation. Exactly. 00:08:38.799 --> 00:08:41.379 It lacks that essential human context and, frankly, 00:08:41.659 --> 00:08:44.340 any real moral compass. Relying on it for deep 00:08:44.340 --> 00:08:47.120 emotional guidance feels, well, pretty concerning. 00:08:47.419 --> 00:08:49.679 And there was even small talk. Another 2 % was 00:08:49.679 --> 00:08:52.259 just basic small talk. Hi, how are you? Tell 00:08:52.259 --> 00:08:55.110 me a joke. Things like that. Which on one level 00:08:55.110 --> 00:08:57.889 is harmless, but it does suggest that for some 00:08:57.889 --> 00:09:00.090 people the line between tool and companion is 00:09:00.090 --> 00:09:02.389 basically gone It could even point to you know 00:09:02.389 --> 00:09:05.509 loneliness replacing potentially tricky human 00:09:05.509 --> 00:09:08.669 chats with easy, predictable AI ones. And this 00:09:08.669 --> 00:09:10.490 all ties into a bigger trend they saw, right? 00:09:10.610 --> 00:09:12.870 Moving from doing to asking. Yeah, that was a 00:09:12.870 --> 00:09:15.690 key shift. Asking the AI to actively do a task 00:09:15.690 --> 00:09:18.850 made up about 35 % of usage, but asking it for 00:09:18.850 --> 00:09:21.809 information, advice, or opinions, asking or consulting 00:09:21.809 --> 00:09:24.289 that was over half, almost 52%. So we want the 00:09:24.289 --> 00:09:26.690 AI to help us think, help us decide, more than 00:09:26.690 --> 00:09:29.110 we want it to just perform tasks for us. Seems 00:09:29.110 --> 00:09:31.779 that way. Which leads to another question. If 00:09:31.779 --> 00:09:34.820 AI is always there, this patient, non -judgmental 00:09:34.820 --> 00:09:38.639 listener, does that constant availability maybe 00:09:38.639 --> 00:09:42.159 reduce our own ability or willingness to engage 00:09:42.159 --> 00:09:45.279 in those messy, sometimes difficult but ultimately 00:09:45.279 --> 00:09:47.620 deeper human connections? It feels like it could, 00:09:47.820 --> 00:09:51.299 yeah. It risks making us less reliant on the 00:09:51.299 --> 00:09:54.179 hard work of real, nuanced human interaction. 00:09:54.440 --> 00:09:56.500 which is where real growth happens. Okay, let's 00:09:56.500 --> 00:09:58.860 talk about skill levels. The study had some really 00:09:58.860 --> 00:10:01.899 interesting insights into how AI affects beginners 00:10:01.899 --> 00:10:05.490 versus experts. It's almost like A tool of inversion. 00:10:05.690 --> 00:10:07.950 How so? For beginners, people who are novices 00:10:07.950 --> 00:10:10.289 at something, AI is amazing. It can help you 00:10:10.289 --> 00:10:13.350 leap from knowing basically nothing, 0%, up to 00:10:13.350 --> 00:10:15.970 maybe 80 % proficiency really fast. Like the 00:10:15.970 --> 00:10:17.889 coding example. If you've never written HTML, 00:10:18.049 --> 00:10:20.169 you can ask for a simple website for your, I 00:10:20.169 --> 00:10:22.190 don't know, local bakery. And boom, you get a 00:10:22.190 --> 00:10:24.870 decent functional starting point. It dramatically 00:10:24.870 --> 00:10:27.309 lowers the barrier to entry for learning practical 00:10:27.309 --> 00:10:30.110 new skills. It makes starting less intimidating. 00:10:30.330 --> 00:10:33.279 But for experts, it's different. Completely different 00:10:33.279 --> 00:10:35.960 story. Experts usually already know that basic 00:10:35.960 --> 00:10:39.899 80%. They operate in that final, tricky 20 % 00:10:39.899 --> 00:10:43.080 where nuance, deep experience, and subtle judgment 00:10:43.080 --> 00:10:46.059 matter most. And that's where AI struggles. Often, 00:10:46.179 --> 00:10:48.960 yeah. This leads to what the study called knowledge 00:10:48.960 --> 00:10:51.960 dilution. The AI might get the basics right, 00:10:52.039 --> 00:10:54.360 but it messes up the critical details that define 00:10:54.360 --> 00:10:56.960 true expertise. Like the Hemingway example they 00:10:56.960 --> 00:10:59.559 mentioned, an AI might mimic the short sentences. 00:10:59.759 --> 00:11:01.559 Right, it gets the superficial style points, 00:11:01.879 --> 00:11:04.360 but it misses the underlying tone, the subtext, 00:11:04.480 --> 00:11:07.179 the why behind Hemingway's choices. So the expert 00:11:07.179 --> 00:11:09.440 ends up spending their valuable time just correcting 00:11:09.440 --> 00:11:13.039 the AI's generic, slightly off output. So the 00:11:13.039 --> 00:11:15.379 AI helps the novice more than the expert in a 00:11:15.379 --> 00:11:17.899 way. The expert still has to provide that crucial 00:11:17.899 --> 00:11:20.840 last 20 % of human insight. Exactly. The real 00:11:20.840 --> 00:11:23.200 value add is still human at the highest levels. 00:11:23.580 --> 00:11:25.899 And thinking about who's using this, the study 00:11:25.899 --> 00:11:27.799 found that half of all messages are coming from 00:11:27.799 --> 00:11:31.120 Gen Z. Half? Wow. I mean, imagine scaling that 00:11:31.120 --> 00:11:32.940 learning curve, or maybe that dependency curve, 00:11:33.399 --> 00:11:36.080 to potentially billions of queries globally from 00:11:36.080 --> 00:11:38.620 just that generation. That's huge. It is. And 00:11:38.620 --> 00:11:41.139 it drives that worry we talked about. Are young 00:11:41.139 --> 00:11:43.220 people becoming too dependent? Are they using 00:11:43.220 --> 00:11:45.899 it to bypass the struggle of learning, asking 00:11:45.899 --> 00:11:48.919 for the whole essay? Instead of using it smartly 00:11:48.919 --> 00:11:51.600 for an outline or to check their arguments, it's 00:11:51.600 --> 00:11:53.440 the difference between cheating yourself and 00:11:53.440 --> 00:11:55.720 augmenting yourself. Right. It always comes down 00:11:55.720 --> 00:11:58.200 to how you use the tool. But the temptation to 00:11:58.200 --> 00:12:00.899 take the easy route is definitely there. On a 00:12:00.899 --> 00:12:02.600 positive note, though, they mentioned the gender 00:12:02.600 --> 00:12:06.340 gap is closing. Yeah, that's great news for accessibility. 00:12:06.799 --> 00:12:09.879 It's almost even now, about 48 % male users. 00:12:10.240 --> 00:12:12.679 And they did see some slight tendency differences 00:12:12.679 --> 00:12:15.559 in how people use it. Like what? Female users 00:12:15.559 --> 00:12:17.860 tended to lean a bit more towards help with writing 00:12:17.860 --> 00:12:20.419 and practical guidance, like planning or organizing 00:12:20.419 --> 00:12:23.000 things. Male users leaned a bit more towards 00:12:23.000 --> 00:12:25.559 technical help, coding, and straightforward information 00:12:25.559 --> 00:12:28.419 seeking. Just tendencies, though, not hard rules, 00:12:28.480 --> 00:12:30.679 obviously. Interesting. So this brings up a big 00:12:30.679 --> 00:12:34.059 question for the future, then. If AI gets really 00:12:34.059 --> 00:12:37.120 good at handling that first 80 % of almost any 00:12:37.120 --> 00:12:39.799 task, giving everyone a solid starting point, 00:12:41.159 --> 00:12:43.159 how do future experts actually develop those 00:12:43.159 --> 00:12:45.059 foundational skills, the stuff you learned in 00:12:45.059 --> 00:12:47.139 that initial struggle? That's the million dollar 00:12:47.139 --> 00:12:49.590 question, isn't it? It suggests the future of 00:12:49.590 --> 00:12:52.190 developing expertise might be less about mastering 00:12:52.190 --> 00:12:55.149 the basics from scratch and more about learning 00:12:55.149 --> 00:12:58.370 how to rigorously question, validate, and ultimately 00:12:58.370 --> 00:13:01.470 perfect the AI's initial 80 % output. OK, let's 00:13:01.470 --> 00:13:03.850 try to bring this all together. We've seen AI 00:13:03.850 --> 00:13:06.789 shift dramatically towards personal use, redefining 00:13:06.789 --> 00:13:09.250 productivity around learning and life admin. 00:13:10.440 --> 00:13:12.740 Professionally, the key skill seems to be evolving 00:13:12.740 --> 00:13:16.100 towards editing, refining, judging AI output, 00:13:16.519 --> 00:13:18.799 rather than just raw creation from zero. Right. 00:13:19.120 --> 00:13:21.840 And AI literacy, knowing how to prompt, well, 00:13:22.159 --> 00:13:24.340 critically evaluate the answers, spot the hallucinations, 00:13:24.379 --> 00:13:26.480 that's becoming absolutely essential for everyone. 00:13:26.620 --> 00:13:29.399 But you mentioned a final major risk they highlighted. 00:13:29.639 --> 00:13:31.500 Decision laundering. Yeah, and this one feels 00:13:31.500 --> 00:13:34.100 like the most significant ethical red flag decision 00:13:34.100 --> 00:13:36.179 laundering. Okay. What exactly is that? It's 00:13:36.179 --> 00:13:39.019 when someone usually in a position of power uses 00:13:39.019 --> 00:13:42.139 an AI to help make a really difficult decision 00:13:42.139 --> 00:13:45.860 maybe one with Serious moral weight right or 00:13:45.860 --> 00:13:47.539 impacting people's lives like who to lay off 00:13:47.539 --> 00:13:50.000 based on some performance data Okay, and then 00:13:50.000 --> 00:13:52.460 they essentially blame the AI for the outcome 00:13:52.460 --> 00:13:54.620 to avoid taking personal responsibility They 00:13:54.620 --> 00:13:56.940 say well the algorithm crunched the numbers and 00:13:56.940 --> 00:13:59.799 this is what it decided They launder their difficult 00:13:59.799 --> 00:14:02.240 choice through the perceived objectivity of the 00:14:02.240 --> 00:14:06.000 machine. Wow, that's Yeah, that's really problematic. 00:14:06.139 --> 00:14:09.299 It's deeply concerning because AI no matter how 00:14:09.299 --> 00:14:12.740 smart it seems has no actual morals It has no 00:14:12.740 --> 00:14:15.620 empathy no understanding of human context or 00:14:15.620 --> 00:14:18.720 consequences Beyond the data it was trained on 00:14:18.720 --> 00:14:22.059 so offloading those fundamentally human judgments 00:14:22.059 --> 00:14:24.460 the ones that require courage and ethical consideration 00:14:24.960 --> 00:14:27.220 onto an algorithm. It's basically abdicating 00:14:27.220 --> 00:14:29.740 our responsibility. Humans have to own the final 00:14:29.740 --> 00:14:32.240 decision, especially the hard ones. We can use 00:14:32.240 --> 00:14:34.840 AI as a tool to inform us, but the judgment call, 00:14:34.860 --> 00:14:37.580 that has to remain human. This whole study really 00:14:37.580 --> 00:14:40.600 paints a picture of AI becoming less of a simple 00:14:40.600 --> 00:14:44.179 tool and more of a constant advisor, right? Deeply 00:14:44.179 --> 00:14:46.159 embedded in our personal lives, helping us think, 00:14:46.279 --> 00:14:48.580 helping us decide. We're becoming editors of 00:14:48.580 --> 00:14:50.779 choices, not just text. Yeah, we're outsourcing 00:14:50.779 --> 00:14:53.629 the cognitive friction, maybe. So maybe the final 00:14:53.629 --> 00:14:56.389 thought to leave people with is this. If we rely 00:14:56.389 --> 00:14:59.610 on AI so much as an advisor, if our main role 00:14:59.610 --> 00:15:02.789 becomes editing the choices it presents, are 00:15:02.789 --> 00:15:05.929 we, over time, subtly losing our own ability 00:15:05.929 --> 00:15:09.029 to generate truly original ideas, or maybe more 00:15:09.029 --> 00:15:11.169 importantly, to develop and trust our own moral 00:15:11.169 --> 00:15:13.330 conviction when things get tough? That's definitely 00:15:13.330 --> 00:15:15.090 something worth thinking about as we all keep 00:15:15.090 --> 00:15:17.629 using these incredibly powerful tools every day. 00:15:17.730 --> 00:15:19.629 Thank you for diving deep into this fascinating 00:15:19.629 --> 00:15:21.570 study with us today. Pre -music.