WEBVTT 00:00:00.000 --> 00:00:02.319 Share with me your first, last name, name of 00:00:02.319 --> 00:00:05.000 your business, and where you're located at. Javonca, 00:00:05.099 --> 00:00:08.140 so great to be here, and thanks for having me 00:00:08.140 --> 00:00:10.839 on your show, and awesome to meet you. So my 00:00:10.839 --> 00:00:14.019 name's Alec Crawford. I'm founder and CEO of 00:00:14.019 --> 00:00:17.019 Artificial Intelligence Risk, Inc., and I'm calling 00:00:17.019 --> 00:00:19.679 in today from Greenwich, Connecticut, which is 00:00:19.679 --> 00:00:22.980 right outside New York City, where I was born 00:00:22.980 --> 00:00:27.179 and grew up. I love that. Is that where home 00:00:27.179 --> 00:00:30.019 is? Yeah, so Greenwich is home. I'm in New York 00:00:30.019 --> 00:00:32.840 probably, you know, once a week. My sister still 00:00:32.840 --> 00:00:36.259 lives in New York and I'm in there obviously 00:00:36.259 --> 00:00:40.159 for business. We focus primarily on integrating 00:00:40.159 --> 00:00:44.060 high risk AI safely. So that's banks, insurance 00:00:44.060 --> 00:00:46.840 companies, but also health care companies or 00:00:46.840 --> 00:00:49.299 anyone anywhere that has kind of really sensitive 00:00:49.299 --> 00:00:52.079 data they need to protect. Right. Which is important 00:00:52.079 --> 00:00:55.079 with AI. And by the way, never type your social 00:00:55.079 --> 00:00:59.420 security number into AI. You know, and before, 00:00:59.799 --> 00:01:02.600 as I'll share with everyone, is the fact that 00:01:02.600 --> 00:01:05.480 I had questions because you said you were here 00:01:05.480 --> 00:01:07.480 in San Antonio and you were doing some things. 00:01:07.540 --> 00:01:11.700 Do me a favor and just explain AI risk and the 00:01:11.700 --> 00:01:14.099 details in which you explain to me of what you 00:01:14.099 --> 00:01:16.969 do. yeah absolutely so what we do is integrate 00:01:16.969 --> 00:01:19.769 high -risk ai safely so it starts with the world's 00:01:19.769 --> 00:01:23.950 best models like openai and and gemini but private 00:01:23.950 --> 00:01:27.069 versions so the data doesn't leak out they can't 00:01:27.069 --> 00:01:29.810 train on your data super super important then 00:01:29.810 --> 00:01:33.310 we had put a user interface in uh so everybody 00:01:33.310 --> 00:01:36.030 can use all those different models safely from 00:01:36.030 --> 00:01:38.989 one place right which is cool and then we add 00:01:38.989 --> 00:01:41.290 ai agents in there we're like which are like 00:01:41.290 --> 00:01:44.379 digital employees that do Different jobs, whether 00:01:44.379 --> 00:01:46.379 you're in the call center or the chief financial 00:01:46.379 --> 00:01:49.819 officer's office or check your email or whatever, 00:01:50.019 --> 00:01:52.439 write emails for you. And then most important, 00:01:52.560 --> 00:01:57.680 we wrap it all in what I call AIGRCC or governance 00:01:57.680 --> 00:02:00.840 risk, regulatory compliance and cybersecurity 00:02:00.840 --> 00:02:04.260 layer to protect all of that and make sure you're 00:02:04.260 --> 00:02:06.420 doing everything legally. Because obviously, 00:02:06.500 --> 00:02:08.120 if you're a health care company or a bank, you 00:02:08.120 --> 00:02:10.400 got lots of rules you got to follow. And then 00:02:10.400 --> 00:02:13.580 we help our customers kind of. customize it and 00:02:13.580 --> 00:02:16.139 kind of learn to fish. And it's just been such 00:02:16.139 --> 00:02:20.400 an amazing journey. How can you do this without 00:02:20.400 --> 00:02:24.020 the risk? Everything is a risk, right? But there 00:02:24.020 --> 00:02:27.780 has to be a clearly defined level of risk for 00:02:27.780 --> 00:02:32.020 something like that. I have to know, where did 00:02:32.020 --> 00:02:37.360 the idea come from? Oh, yeah. Such a great question. 00:02:38.520 --> 00:02:43.139 So I retired. I was a chief risk officer at Lord 00:02:43.139 --> 00:02:45.599 Abbott, mid -side asset manager, retired from 00:02:45.599 --> 00:02:49.580 there, ended 22. And of course, my great love 00:02:49.580 --> 00:02:52.219 is computer science. I was building neural networks 00:02:52.219 --> 00:02:56.139 from scratch at Harvard in 1987 and always loved 00:02:56.139 --> 00:02:59.800 computers, but then spent years in finance and 00:02:59.800 --> 00:03:02.539 investing and risk management. And it was kind 00:03:02.539 --> 00:03:05.449 of like... But like the Reese's Peanut Butter 00:03:05.449 --> 00:03:07.490 Cup commercial, like, hey, you got your chocolate 00:03:07.490 --> 00:03:09.270 and my peanut butter. No, you got your peanut 00:03:09.270 --> 00:03:11.909 butter and my chocolate. I was like, oh, risk 00:03:11.909 --> 00:03:15.930 and AI, right? Like manage the risk of AI. So 00:03:15.930 --> 00:03:18.930 it just kind of hit me on the head and also seemed 00:03:18.930 --> 00:03:22.770 like such a really, really something that the 00:03:22.770 --> 00:03:25.189 world really needed, right? Like if we can't 00:03:25.189 --> 00:03:29.539 control the risks of AI over the next. 10 years, 00:03:29.680 --> 00:03:31.599 I mean, you've seen all these things about, you 00:03:31.599 --> 00:03:34.219 know, existential risk to humanity of AI, right? 00:03:34.300 --> 00:03:37.360 Like, if we need anything, we need this, you 00:03:37.360 --> 00:03:41.620 know, controlling the risks of AI. And with that 00:03:41.620 --> 00:03:45.560 said, I love the Reese's commercial you just 00:03:45.560 --> 00:03:49.699 did. That was a whole commercial. And I love 00:03:49.699 --> 00:03:53.039 it. With that, I mean, it's crazy, because when 00:03:53.039 --> 00:03:55.919 AI came for us, it was like maybe about five 00:03:55.919 --> 00:03:59.240 years ago, but for you. How long have you been 00:03:59.240 --> 00:04:03.259 working with AI? Oh, yeah, since 1987. And, you 00:04:03.259 --> 00:04:04.639 know, obviously people are playing around with 00:04:04.639 --> 00:04:06.860 ChatGPT and things like that. It's become very 00:04:06.860 --> 00:04:09.560 accessible. You know, back then or really before 00:04:09.560 --> 00:04:12.240 ChatGPT came out, it was really kind of computer 00:04:12.240 --> 00:04:16.019 science, like experts in this, right? And so 00:04:16.019 --> 00:04:18.720 a neural network is basically what people call 00:04:18.720 --> 00:04:22.660 deep learning today. And it's what allows ChatGPT 00:04:22.660 --> 00:04:25.360 to work, right? It's a large language model. 00:04:25.519 --> 00:04:28.639 So that's really the precursor. to that. It's 00:04:28.639 --> 00:04:31.139 become much more technical and complicated since 00:04:31.139 --> 00:04:33.519 I did it with these things called transformers. 00:04:33.519 --> 00:04:36.339 And basically the tech has changed for sure, 00:04:36.399 --> 00:04:38.899 but the basic concept is the same. Let's figure 00:04:38.899 --> 00:04:41.480 out effectively how to mimic the human brain 00:04:41.480 --> 00:04:46.199 statistically and have an AI be able to talk 00:04:46.199 --> 00:04:49.439 to you. And what I did realize back in the 80s 00:04:49.439 --> 00:04:51.439 was, look, we just don't have enough computing 00:04:51.439 --> 00:04:54.819 power or memory to make this work. It did okay. 00:04:54.939 --> 00:04:56.879 I was basically teaching computers to play poker. 00:04:57.339 --> 00:05:00.540 Yes. Yes. Right. And bet and bluff. And they 00:05:00.540 --> 00:05:02.579 did a good job. They basically they could almost 00:05:02.579 --> 00:05:07.079 beat me. Right. But now poker bots are the best 00:05:07.079 --> 00:05:08.980 poker players in the world. They can beat any 00:05:08.980 --> 00:05:13.839 human. Right. So so it's it's AI has been a complete 00:05:13.839 --> 00:05:16.660 game changer. And we've seen it over time when. 00:05:17.360 --> 00:05:20.399 Deep Blue beat the world champion in chess and 00:05:20.399 --> 00:05:24.220 then AlphaGo beat the world champion in Go. And 00:05:24.220 --> 00:05:27.240 then they started beating poker players and everyone 00:05:27.240 --> 00:05:31.480 else. And look, it's the quote is basically software 00:05:31.480 --> 00:05:34.660 is eating the world, right? Like that's a pretty 00:05:34.660 --> 00:05:37.800 famous quote about software. And now it's AI 00:05:37.800 --> 00:05:40.519 is eating the world. And I think that there's 00:05:40.519 --> 00:05:42.480 a couple of really important things there. One 00:05:42.480 --> 00:05:45.839 is like. you got to get on board with AI. If 00:05:45.839 --> 00:05:48.000 you're not using AI today, like you may not have 00:05:48.000 --> 00:05:49.899 a job two years from now, unless you're a real 00:05:49.899 --> 00:05:53.139 estate agent, which I think you're good. Because 00:05:53.139 --> 00:05:55.079 look, there are just some roles which people 00:05:55.079 --> 00:05:56.680 are always going to want to talk to a human, 00:05:56.779 --> 00:06:03.220 right? Yes. But look, if you're, you know, if 00:06:03.220 --> 00:06:06.300 you're processing, you know, forms in a bank 00:06:06.300 --> 00:06:09.199 somewhere, like find a new job fast, right? Because 00:06:09.199 --> 00:06:12.899 AI can do that. The second part is we've got 00:06:12.899 --> 00:06:15.740 to be able to do this safely and securely. And 00:06:15.740 --> 00:06:17.540 what people need to realize, whether you're an 00:06:17.540 --> 00:06:20.720 individual or work at a bank, is that if you 00:06:20.720 --> 00:06:22.399 just go to the web and start using Gemini or 00:06:22.399 --> 00:06:24.740 ChatGPT, like all that information is public. 00:06:25.259 --> 00:06:27.660 That is not safe. You have to be super careful. 00:06:27.720 --> 00:06:30.079 Do not put personal information on there. Do 00:06:30.079 --> 00:06:32.379 not put your health conditions on there. Like 00:06:32.379 --> 00:06:34.879 that's just not a great idea, right? You've got 00:06:34.879 --> 00:06:38.339 to be super, super careful about that. So those 00:06:38.339 --> 00:06:40.250 are really the two things to think about. And 00:06:40.250 --> 00:06:42.509 that was just that was very interesting. You 00:06:42.509 --> 00:06:46.470 said that because a lot of people, most people 00:06:46.470 --> 00:06:50.569 are utilizing chat GPT to sort of I used to call 00:06:50.569 --> 00:06:54.389 it diagnosing themselves. So to chat knows themselves 00:06:54.389 --> 00:06:58.209 like when it comes to medical. But my my aha 00:06:58.209 --> 00:07:01.110 moment again, which was you said is that when 00:07:01.110 --> 00:07:06.199 we use. chat GPT or Gemini, and someone puts 00:07:06.199 --> 00:07:08.879 our name in there, it can literally pull from 00:07:08.879 --> 00:07:12.319 the information, vice versa. Am I correct? Like 00:07:12.319 --> 00:07:14.939 taking whatever you put in your chat and using 00:07:14.939 --> 00:07:18.060 it when somebody searches your name? Yeah, absolutely. 00:07:18.240 --> 00:07:21.279 So one of the things that, I'll give an example. 00:07:21.540 --> 00:07:23.939 So one of the things that OpenAI does is they 00:07:23.939 --> 00:07:27.699 make entire chat conversations available for 00:07:27.699 --> 00:07:30.360 research purposes. Now, this is not... If you 00:07:30.360 --> 00:07:32.680 have a corporate version or you're using our 00:07:32.680 --> 00:07:34.459 version of software where everything's inside 00:07:34.459 --> 00:07:36.439 a firewall, of course, they can't do that. Right. 00:07:36.699 --> 00:07:38.519 But if you're just on the Internet saying, hi, 00:07:38.639 --> 00:07:41.220 my name's Javanka McGee and I have this medical 00:07:41.220 --> 00:07:42.860 condition, I have questions about it or I have 00:07:42.860 --> 00:07:45.600 symptoms, I have questions about this. That may 00:07:45.600 --> 00:07:50.600 be may be released six months later for research 00:07:50.600 --> 00:07:53.639 purposes, for people to do research on AI. And 00:07:53.639 --> 00:07:55.480 one of the things that the Atlantic magazine 00:07:55.480 --> 00:07:59.430 did a few months ago. is they published excerpts 00:07:59.430 --> 00:08:03.009 of those because it's publicly available. And 00:08:03.009 --> 00:08:05.569 there were people in there. There was there's 00:08:05.569 --> 00:08:09.230 a young woman there who was worried about, you 00:08:09.230 --> 00:08:11.870 know, her weight, having a conversation with 00:08:11.870 --> 00:08:13.370 Chachi Petit. There was a guy thinking about 00:08:13.370 --> 00:08:14.970 getting divorced from his wife. These are all 00:08:14.970 --> 00:08:17.689 super, super private conversations that were 00:08:17.689 --> 00:08:21.709 literally published in The Atlantic magazine. 00:08:21.750 --> 00:08:25.569 So it is not safe. You know, to go out there 00:08:25.569 --> 00:08:28.569 and do that. You have to be very, very careful 00:08:28.569 --> 00:08:29.829 about what you do. And people just start paying 00:08:29.829 --> 00:08:32.909 attention to that. Right. And you need to pay 00:08:32.909 --> 00:08:34.669 attention to that. I never even thought about 00:08:34.669 --> 00:08:36.169 that, Alec. I'm sitting here like, wouldn't the 00:08:36.169 --> 00:08:38.490 heck have I asked Chad GPT in the last year? 00:08:40.149 --> 00:08:43.870 With all of this said, you're so passionate about 00:08:43.870 --> 00:08:46.049 it. And there's one word that I love and I use 00:08:46.049 --> 00:08:49.809 it respectfully. But I'm like, oh, he's a computer 00:08:49.809 --> 00:08:54.480 nerd. I love it. Remember when the word nerd 00:08:54.480 --> 00:08:58.460 used to be such a bad thing? Yeah. Now as a form 00:08:58.460 --> 00:09:01.279 of like intelligence, like at a level that's 00:09:01.279 --> 00:09:04.379 like beyond I could think of. So Alec Crawford, 00:09:04.440 --> 00:09:08.080 you're like this super nerd that I love. Yeah. 00:09:08.080 --> 00:09:11.919 Yeah. I've moved away from I was very introverted 00:09:11.919 --> 00:09:14.720 as a kid. Like I taught myself the program at 00:09:14.720 --> 00:09:19.879 age 13. I got my first computer in 1979. Right. 00:09:19.940 --> 00:09:23.409 Like. I was definitely that generation. But look, 00:09:23.490 --> 00:09:27.230 as you work in larger institutions, as you know, 00:09:27.389 --> 00:09:29.950 because you were at USAA, like you have to become 00:09:29.950 --> 00:09:33.350 more outgoing. Yes. Like if you want to if you 00:09:33.350 --> 00:09:35.669 want to manage people and get promoted, like 00:09:35.669 --> 00:09:37.870 you can't just sit in your cubicle all day and 00:09:37.870 --> 00:09:39.970 take numbers to a computer. That ain't going 00:09:39.970 --> 00:09:42.250 to work. Right. So I effectively had to transition 00:09:42.250 --> 00:09:45.309 from being an introvert to more of an extrovert. 00:09:45.429 --> 00:09:47.590 Right. Although certainly at times I could be 00:09:47.590 --> 00:09:50.159 at some. Big event where I know no one and go, 00:09:50.220 --> 00:09:52.419 ooh, this is a little scary, but you kind of 00:09:52.419 --> 00:09:54.799 have to force yourself to interact with people. 00:09:54.960 --> 00:09:57.860 So for better or worse, I've become a little 00:09:57.860 --> 00:10:00.059 more nerdy in that I'm running an AI company, 00:10:00.159 --> 00:10:02.000 but a little less nerdy in that I'm going to 00:10:02.000 --> 00:10:05.159 lots of conferences and now have, I was joking 00:10:05.159 --> 00:10:08.500 with my co -founder, Joe, who you met, because 00:10:08.500 --> 00:10:10.539 when I started the company, I was like, I've 00:10:10.539 --> 00:10:13.440 got 10 ,000 public speaking hours as a Wall Street 00:10:13.440 --> 00:10:15.980 analyst. Like I've done this like a million times. 00:10:16.080 --> 00:10:19.379 I've spoken to. groups of 10 ,000 people and 00:10:19.379 --> 00:10:21.620 rooms with three people in them. Right. But tons 00:10:21.620 --> 00:10:23.539 of public speaking. And then I looked at him 00:10:23.539 --> 00:10:29.820 and go, Oh, it's up to 15 ,000 now. Wow. Thousand 00:10:29.820 --> 00:10:32.279 hours of public speaking. Right. Like, like you 00:10:32.279 --> 00:10:34.259 cannot be an introvert to do that. Tell me about 00:10:34.259 --> 00:10:37.419 your vision in all of this. Yeah. Thank you so 00:10:37.419 --> 00:10:41.399 much. Yeah. Look, my vision is for companies 00:10:41.399 --> 00:10:45.820 and people to be able to use AI safely and securely, 00:10:46.080 --> 00:10:47.879 right? So we talked a lot about that earlier, 00:10:48.019 --> 00:10:50.279 where if you do things wrong, you could just 00:10:50.279 --> 00:10:52.500 lose your data, right? The other thing is, look, 00:10:52.559 --> 00:10:55.000 AI is not that trustworthy right now, right? 00:10:55.120 --> 00:10:58.080 You can ask it questions, get the wrong answer, 00:10:58.279 --> 00:11:00.860 but it's going to sound really authoritative, 00:11:00.919 --> 00:11:06.149 right? And it's also easy to fool AI into things. 00:11:06.230 --> 00:11:08.429 Like there was an incident over the last couple 00:11:08.429 --> 00:11:13.370 of weeks where people had fooled Grok, one of 00:11:13.370 --> 00:11:16.029 the AI systems out there, and kind of that it 00:11:16.029 --> 00:11:21.169 was misrepresenting facts, right? So it's super 00:11:21.169 --> 00:11:24.190 important to just remember right now with AI, 00:11:24.289 --> 00:11:27.509 trust, but verify, right? If you can answer from 00:11:27.509 --> 00:11:30.269 there, especially if it's important. Check a 00:11:30.269 --> 00:11:32.769 source, ask your friends. I can't tell you how 00:11:32.769 --> 00:11:36.090 much stuff someone I know will send me a text 00:11:36.090 --> 00:11:37.669 or whatever, like, oh, yeah, I told me this. 00:11:37.730 --> 00:11:39.309 I'm like, that's not right. You know, like you 00:11:39.309 --> 00:11:42.070 got to go check that. Right. So trust but verify. 00:11:42.190 --> 00:11:49.610 So our vision is for the future of AI to be to 00:11:49.610 --> 00:11:53.929 be trustworthy and safe and good for everybody. 00:11:54.649 --> 00:11:57.649 The other part of my vision. is that aligned 00:11:57.649 --> 00:12:00.610 with that is 10 years from now i want everybody 00:12:00.610 --> 00:12:04.429 to have an ai co -pilot uh that will help them 00:12:04.429 --> 00:12:06.509 do their job better that will take the stuff 00:12:06.509 --> 00:12:08.370 out your job you hate like filling out paperwork 00:12:08.370 --> 00:12:11.429 whatever a the ai is going to do that right um 00:12:11.429 --> 00:12:15.850 what i don't want to see is 15 unemployment because 00:12:15.850 --> 00:12:18.230 people's jobs are taken by ai and that's something 00:12:18.230 --> 00:12:22.470 i think all of us as a community of citizens 00:12:22.470 --> 00:12:27.200 and business leaders And government leaders really 00:12:27.200 --> 00:12:29.960 need to think very hard about how do we achieve 00:12:29.960 --> 00:12:34.120 my vision of everybody as a copilot instead of 00:12:34.120 --> 00:12:39.080 accidentally ending up with 15 % unemployment. 00:12:39.179 --> 00:12:41.399 We have to think about that now or it will happen 00:12:41.399 --> 00:12:47.740 later. And I'm not by far an expert nor experienced 00:12:47.740 --> 00:12:51.059 in all of this, but you just said copilot and 00:12:51.059 --> 00:12:56.860 I'm thinking. we have to verify. And if we continue 00:12:56.860 --> 00:12:59.940 to keep using it without verification and putting 00:12:59.940 --> 00:13:02.500 it out there, I think the verifier would be a 00:13:02.500 --> 00:13:06.000 human. So we should not get to the 15 % of unemployment 00:13:06.000 --> 00:13:09.919 because there's a thought, and I don't like using 00:13:09.919 --> 00:13:12.120 the word should, there's a thought to make sure 00:13:12.120 --> 00:13:17.220 that you have somebody clearly define and go 00:13:17.220 --> 00:13:19.259 through that. That you're putting out something 00:13:19.259 --> 00:13:21.919 that's trustworthy, not even just because of 00:13:21.919 --> 00:13:23.779 that, but as a company. That's our goal, right? 00:13:23.940 --> 00:13:27.100 As a consumer, I would want to know that what 00:13:27.100 --> 00:13:29.700 you're telling me, because that's your expertise, 00:13:29.820 --> 00:13:32.700 is right rather than you just going on AI and 00:13:32.700 --> 00:13:34.379 sending me the email. You know what I mean? So 00:13:34.379 --> 00:13:36.659 I never even thought about that until you said 00:13:36.659 --> 00:13:39.500 that. Yeah, totally agree. Well, look, even in 00:13:39.500 --> 00:13:41.179 parts of the government, they're trying to have 00:13:41.179 --> 00:13:44.470 AI be a judge and check the other AI. That does 00:13:44.470 --> 00:13:47.429 not work. Do not do that. Right. You need a human 00:13:47.429 --> 00:13:50.389 in the loop on a lot of this stuff. And you also 00:13:50.389 --> 00:13:53.850 need a safety valve. Right. So Colorado passed 00:13:53.850 --> 00:13:57.549 an AI act. It goes to effect in February. And 00:13:57.549 --> 00:14:01.029 one of the things they say is. If you're a bank, 00:14:01.110 --> 00:14:04.190 you can have AI grant a loan or deny a loan. 00:14:04.529 --> 00:14:06.669 But the consumer then has the ability to go back 00:14:06.669 --> 00:14:08.549 to the bank and say, I want a human to check 00:14:08.549 --> 00:14:12.850 it. Right. So that's very important. Right. Like 00:14:12.850 --> 00:14:16.769 and that should be part of your standard operating 00:14:16.769 --> 00:14:20.710 procedure as a financial institution is you've 00:14:20.710 --> 00:14:22.230 got to have that safety valve because, you know, 00:14:22.230 --> 00:14:25.049 AI is not always going to be right. The other 00:14:25.049 --> 00:14:28.230 issue, of course, is you're training AI on data. 00:14:28.649 --> 00:14:33.070 And sometimes you have, you know, limited use 00:14:33.070 --> 00:14:37.960 cases. Right. So, for example. If I'm a bank 00:14:37.960 --> 00:14:39.620 and I've got data and everybody I've granted 00:14:39.620 --> 00:14:43.379 loans to and an 87 year old comes and asks for 00:14:43.379 --> 00:14:45.360 a loan, like I might not have any data there. 00:14:45.759 --> 00:14:48.519 It might simply turn that person down because 00:14:48.519 --> 00:14:51.679 it's never seen an 87 year old applicant. Right. 00:14:52.080 --> 00:14:55.679 Ever. Is that fair? Absolutely not. That's ageism. 00:14:55.740 --> 00:14:59.100 Right. So you've got to be super careful about 00:14:59.100 --> 00:15:05.570 these things. And I think that. That's going 00:15:05.570 --> 00:15:08.169 to be critical over the next few years is just 00:15:08.169 --> 00:15:14.049 making sure that we focus on eliminating bias 00:15:14.049 --> 00:15:17.549 and discrimination when AI is being used. Super 00:15:17.549 --> 00:15:22.370 important. What's your why? Yeah, I'm really 00:15:22.370 --> 00:15:26.149 doing it for everyone else, not for me, right? 00:15:26.350 --> 00:15:31.690 Like, so think about all the people out there 00:15:31.690 --> 00:15:34.730 that don't understand. When they type in their 00:15:34.730 --> 00:15:36.929 medical condition to chat to people on the Internet, 00:15:37.049 --> 00:15:38.570 that that's something that can be discovered. 00:15:38.649 --> 00:15:41.370 Right. Think about all the companies out there 00:15:41.370 --> 00:15:44.990 that ran out there and implemented AI first. 00:15:45.210 --> 00:15:46.990 And now they're sitting there going like, oh, 00:15:47.169 --> 00:15:52.110 is this safe? Yeah. Like super, super have to 00:15:52.110 --> 00:15:54.629 make sure that number one, people are educated 00:15:54.629 --> 00:15:58.129 about this. Yes. And number two, we have companies 00:15:58.129 --> 00:16:01.870 that are not. just focused on AI use cases, but 00:16:01.870 --> 00:16:05.490 focused on AI trust, safety, compliance. So what 00:16:05.490 --> 00:16:07.870 I call AI, GRCC, or governance, risk, compliance, 00:16:07.970 --> 00:16:09.649 and cybersecurity. I'm going to give you a crazy 00:16:09.649 --> 00:16:16.110 statistic. So in the US, about 25, 30 % of all 00:16:16.110 --> 00:16:21.029 software spending is on cybersecurity, as well 00:16:21.029 --> 00:16:23.190 as governance, risk, and compliance. The rest 00:16:23.190 --> 00:16:27.049 is like CRMs and other kind of stuff, right? 00:16:27.769 --> 00:16:33.490 Word, all that kind of stuff. In AI, less than 00:16:33.490 --> 00:16:37.950 1 % is being spent on trust and safety. Oh, wow. 00:16:38.169 --> 00:16:43.710 That is crazy, right? That should be 30%. Think 00:16:43.710 --> 00:16:46.889 about how powerful AI is and people are spending 00:16:46.889 --> 00:16:50.470 basically no money right now on trust and safety. 00:16:50.549 --> 00:16:54.389 That has to change. Yeah. And I don't want to 00:16:54.389 --> 00:16:57.070 take up too much of your time, but do you believe 00:16:57.070 --> 00:17:02.029 that this has grown so rapidly that no one really 00:17:02.029 --> 00:17:05.609 thought about what it was going to be used for 00:17:05.609 --> 00:17:07.829 and really just thought maybe it was just for, 00:17:07.890 --> 00:17:10.349 you know, just to actually talk to and get some 00:17:10.349 --> 00:17:13.390 quick answers? When you started this out in the 00:17:13.390 --> 00:17:16.029 80s, did you know that this is where we were 00:17:16.029 --> 00:17:18.970 going to go with this government? I mean, banks. 00:17:19.329 --> 00:17:22.230 I didn't think so. Yeah, no, we've been thinking 00:17:22.230 --> 00:17:24.009 about this for a long time. This guy named Ray 00:17:24.009 --> 00:17:28.849 Kurzweil, who has been predicting this for decades, 00:17:29.170 --> 00:17:36.869 literally since the 1970s. And so, yeah, I knew 00:17:36.869 --> 00:17:38.049 it was going to happen. I didn't know when it 00:17:38.049 --> 00:17:39.490 was going to happen, but I knew it was going 00:17:39.490 --> 00:17:43.150 to happen. And now I think the big change is 00:17:43.150 --> 00:17:47.730 it's progressing so rapidly that what I thought 00:17:47.730 --> 00:17:50.750 would take two years has taken six months. Yes. 00:17:50.960 --> 00:17:54.920 Right. And so maybe what I thought would take 00:17:54.920 --> 00:17:58.880 30 years will take five. And that is what people 00:17:58.880 --> 00:18:03.759 call artificial general intelligence. And I think 00:18:03.759 --> 00:18:07.319 we're going to have something that looks like 00:18:07.319 --> 00:18:12.500 human intelligence within years. And I don't 00:18:12.500 --> 00:18:14.200 think it really will be human intelligence. I 00:18:14.200 --> 00:18:16.359 think it will kind of fool a lot of people. I 00:18:16.359 --> 00:18:20.089 think to truly get. artificial general intelligence, 00:18:20.490 --> 00:18:22.269 which, by the way, is different from consciousness, 00:18:22.710 --> 00:18:25.589 I think it is going to take longer. It's just 00:18:25.589 --> 00:18:28.569 very, very complex. And we will probably have 00:18:28.569 --> 00:18:31.269 something, maybe it's within my lifetime, maybe 00:18:31.269 --> 00:18:34.630 not. And maybe that involves a combination of 00:18:34.630 --> 00:18:37.809 AI and quantum computing that will effectively 00:18:37.809 --> 00:18:41.549 be conscious at some point. Because look, humans 00:18:41.549 --> 00:18:47.150 are smart. Humans are creative. Once we build... 00:18:47.559 --> 00:18:50.019 an ai as smart as a human what's the first thing 00:18:50.019 --> 00:18:53.099 we do we say go build something smarter yes yes 00:18:53.099 --> 00:18:56.460 right and eventually we get super intelligence 00:18:56.460 --> 00:19:00.839 within a relatively brief period of time and 00:19:00.839 --> 00:19:03.819 it's going to be a matter of what are the parameters 00:19:03.819 --> 00:19:06.779 around that how do we How do we stay safe? Thank 00:19:06.779 --> 00:19:08.779 you for being here today. I'm really happy that 00:19:08.779 --> 00:19:10.819 you tuned in to Vision Pros Live. I'm looking 00:19:10.819 --> 00:19:14.059 forward to seeing your reactions as these episodes 00:19:14.059 --> 00:19:15.859 continue to move forward. This is going to get 00:19:15.859 --> 00:19:18.500 more and more fun. We'll have more and more engagement 00:19:18.500 --> 00:19:20.559 as well. We'll invite people to participate in 00:19:20.559 --> 00:19:22.740 the show. And thank you for giving us your time 00:19:22.740 --> 00:19:25.019 and attention. Have an excellent time building 00:19:25.019 --> 00:19:27.359 out your vision and becoming a vision pro yourself.