WEBVTT 00:00:00.000 --> 00:00:02.859 Welcome to the Innovation Conversation, a podcast 00:00:02.859 --> 00:00:05.419 about innovators, both in business and real life. 00:00:05.900 --> 00:00:08.439 Hosted by myself, Ricardo Rescual and Harry McCona. 00:00:08.800 --> 00:00:11.619 Today's episode is sponsored by Hyperskill. Hyperskill 00:00:11.619 --> 00:00:14.119 is a learning and training platform that enables 00:00:14.119 --> 00:00:16.339 people from all over the world to learn new tech 00:00:16.339 --> 00:00:18.739 skills. If you're looking to learn new tech skills, 00:00:19.000 --> 00:00:20.940 this is the platform to choose. You can find 00:00:20.940 --> 00:00:23.739 out more about them on hyperskill .org. Today 00:00:23.739 --> 00:00:26.390 we are joined by Elon Buckner. one of the godfathers 00:00:26.390 --> 00:00:30.429 of AI. Ellen, thank you so much for being here 00:00:30.429 --> 00:00:33.649 and being one of our guests in the podcast. If 00:00:33.649 --> 00:00:35.210 you'd like to introduce yourself to the audience, 00:00:35.789 --> 00:00:41.369 please. Sure, let's see. I've had about a 30 00:00:41.369 --> 00:00:44.829 -year career. I co -founded and sold a software 00:00:44.829 --> 00:00:50.030 company to Thomson Reuters. I was a hedge fund 00:00:50.030 --> 00:00:53.859 portfolio manager. I built a practice selling 00:00:53.859 --> 00:00:58.200 AI to asset managers, consulting practice. I 00:00:58.200 --> 00:01:02.579 spent three, four years in big tech at Microsoft 00:01:02.579 --> 00:01:05.859 and Google. I was most recently head of AI at 00:01:05.859 --> 00:01:11.659 a company called Factset. And just recently left 00:01:11.659 --> 00:01:16.540 about a month, month and a half ago to explore 00:01:16.540 --> 00:01:21.319 a startup project. I have so many questions lined 00:01:21.319 --> 00:01:24.659 up, it's absolutely crazy. So one of the things 00:01:24.659 --> 00:01:26.959 we normally like to ask in the podcast, it's 00:01:26.959 --> 00:01:30.400 about your own journey as an entrepreneur. So 00:01:30.400 --> 00:01:32.280 how hard was it, because you sold the company 00:01:32.280 --> 00:01:34.659 a couple of years ago, so how hard was it to 00:01:34.659 --> 00:01:37.359 get started, get funding, get any funding? So 00:01:37.359 --> 00:01:38.760 do you want to tell us a bit more about that 00:01:38.760 --> 00:01:41.560 journey? Because you are a very successful entrepreneur, 00:01:41.900 --> 00:01:44.760 right? Not everyone can actually go through an 00:01:44.760 --> 00:01:47.640 exit and an exit to Thompson. Thompson writer, 00:01:47.799 --> 00:01:49.859 right? So it's it's not an easy thing. So you 00:01:49.859 --> 00:01:51.739 want to tell us a bit more about that? Thank 00:01:51.739 --> 00:01:54.659 you, Ricardo. You know, every every journey, 00:01:55.019 --> 00:01:57.400 entrepreneurial journey that I have is completely 00:01:57.400 --> 00:02:00.260 different. I haven't cracked the code yet. But 00:02:00.260 --> 00:02:02.640 I can tell you about my first one. And it was 00:02:02.640 --> 00:02:05.799 it was pretty amazing. I was I was a couple of 00:02:05.799 --> 00:02:09.819 years out of college. And I got together with 00:02:09.819 --> 00:02:15.180 a few a few contacts through Family Connection. 00:02:15.389 --> 00:02:19.050 and they were just getting started. It was basically 00:02:19.050 --> 00:02:26.389 me and maybe one or two other co -founders. We 00:02:26.389 --> 00:02:28.229 were just in an empty office and we were trying 00:02:28.229 --> 00:02:34.370 to figure out what to do. We started out bootstrapping. 00:02:35.349 --> 00:02:37.909 The bootstrapping recipe that I'm familiar with 00:02:37.909 --> 00:02:41.930 is to do consulting. Back then, my background 00:02:41.930 --> 00:02:48.699 was in web design. I had worked for this early 00:02:48.699 --> 00:02:51.699 web design leader called agency .com. They did 00:02:51.699 --> 00:02:54.080 the first website for American Express and for 00:02:54.080 --> 00:02:58.120 Itachi and for a few other major brands. And 00:02:58.120 --> 00:03:01.479 my partner came from Microsoft. He was pretty 00:03:01.479 --> 00:03:03.919 senior in the consulting organization there. 00:03:04.319 --> 00:03:07.879 And our idea was that we would make something 00:03:07.879 --> 00:03:12.080 that merges the beauty of web design and the 00:03:12.080 --> 00:03:14.759 software depth of you know, the Microsoft way 00:03:14.759 --> 00:03:17.360 of building software. This was back in like 1994, 00:03:18.039 --> 00:03:23.120 95. And we would try to kind of build websites 00:03:23.120 --> 00:03:26.060 that are not just skin deep, but actually functional. 00:03:26.979 --> 00:03:32.340 And that worked out pretty well. You know, we 00:03:32.340 --> 00:03:34.879 got the first few clients through just outreach 00:03:34.879 --> 00:03:37.919 to our networks. And then we actually connected, 00:03:37.979 --> 00:03:41.780 we partnered with the Microsoft consulting organization, 00:03:41.879 --> 00:03:46.449 which Just by way of background, back then and 00:03:46.449 --> 00:03:49.030 still now, consulting is a very small part of 00:03:49.030 --> 00:03:51.990 Microsoft's business. They don't really especially 00:03:51.990 --> 00:03:54.909 like to do consulting. It's less profitable than 00:03:54.909 --> 00:03:59.509 building software. Back then, demand for Microsoft 00:03:59.509 --> 00:04:06.289 was booming just after Windows 95 was released 00:04:06.289 --> 00:04:09.629 and all this other stuff. The consulting organization 00:04:09.629 --> 00:04:12.460 in New York, was just getting hammered. There 00:04:12.460 --> 00:04:16.459 was just way, way too many requests. And so we 00:04:16.459 --> 00:04:19.120 were like the Statue of Liberty. There's a plaque 00:04:19.120 --> 00:04:21.660 on the Statue of Liberty that says, give me your 00:04:21.660 --> 00:04:24.860 poor and wretched masses. So we said, give me 00:04:24.860 --> 00:04:27.199 all your clients that don't pay the bills. They 00:04:27.199 --> 00:04:30.459 have unreasonable expectations. They are paying 00:04:30.459 --> 00:04:33.079 to work with, there's scope creep, all the ones 00:04:33.079 --> 00:04:35.579 that you don't like working with, and you will 00:04:35.579 --> 00:04:37.959 never hear from them again. That was our pitch. 00:04:40.540 --> 00:04:44.819 And it worked. We got a couple of referrals from 00:04:44.819 --> 00:04:47.620 Microsoft Consulting, and we took care of them. 00:04:48.579 --> 00:04:52.500 And that became kind of the early start to our 00:04:52.500 --> 00:04:57.879 business. And then over time, it turned out that 00:04:57.879 --> 00:05:01.199 a couple of our clients requested substantially 00:05:01.199 --> 00:05:04.939 the same thing. And that thing was... Anyway, 00:05:05.079 --> 00:05:09.259 so it was my role to... I basically, I noticed 00:05:09.259 --> 00:05:11.040 it, I wrote a memo about it and I said, you know 00:05:11.040 --> 00:05:13.079 what, we can keep doing consulting but it would 00:05:13.079 --> 00:05:16.360 be even better if we had a product because that 00:05:16.360 --> 00:05:18.660 would be a nice compliment to our business and 00:05:18.660 --> 00:05:20.939 it would be really cool to build it and it would 00:05:20.939 --> 00:05:27.379 be profitable. And so I guess became the product 00:05:27.379 --> 00:05:30.160 person and I took that product from the memo 00:05:30.160 --> 00:05:33.060 to where it became something like 70 % of the 00:05:33.060 --> 00:05:35.860 revenue of the company and the reason why we 00:05:35.860 --> 00:05:39.339 got acquired. that product would you know it 00:05:39.339 --> 00:05:43.079 would now be called social networking but back 00:05:43.079 --> 00:05:45.579 then there's no such thing that label didn't 00:05:45.579 --> 00:05:48.000 exist so we called it Bolton boards or community 00:05:48.000 --> 00:05:51.560 and there's also no software as a service we 00:05:51.560 --> 00:05:56.600 shipped it as a CD. I remember those, I remember 00:05:56.600 --> 00:06:00.540 CD -ROMs. That's right so you pop the CD on your 00:06:00.540 --> 00:06:04.060 server and you get all these features and it 00:06:04.060 --> 00:06:06.600 was a lot cheaper than having to sort of build 00:06:06.600 --> 00:06:09.839 those things yourself. Like the moment that we 00:06:09.839 --> 00:06:13.180 hit real sort of product market fit, it was kind 00:06:13.180 --> 00:06:18.399 of interesting because, you know, we were, it's 00:06:18.399 --> 00:06:21.220 not that we were limping along, but we were chugging 00:06:21.220 --> 00:06:24.259 along, it wasn't anything special. And then, 00:06:24.259 --> 00:06:27.259 you know, the dot com wave hit, all these companies 00:06:27.259 --> 00:06:31.399 raised money. And you raise money, the VC tells 00:06:31.399 --> 00:06:33.399 you there's a checklist of things you have to 00:06:33.399 --> 00:06:36.699 do. And usually, Number two or three item on 00:06:36.699 --> 00:06:38.459 that checklist is build a community around your 00:06:38.459 --> 00:06:41.680 website or around whatever you do. You need a 00:06:41.680 --> 00:06:45.040 community around. And so back then, there was 00:06:45.040 --> 00:06:49.040 just no software to do that. And our pitch to 00:06:49.040 --> 00:06:52.019 all of these entrepreneurs was we'll help you 00:06:52.019 --> 00:06:54.860 take number three off your checklist and you'll 00:06:54.860 --> 00:06:57.120 spend, say, a quarter of a million dollars instead 00:06:57.120 --> 00:06:59.459 of one million dollars. And you'll spend like 00:06:59.459 --> 00:07:02.600 30 days instead of six months to build it. It's 00:07:02.600 --> 00:07:03.959 going to work better than whatever you're going 00:07:03.959 --> 00:07:07.240 to build in six months. It was super expensive, 00:07:07.680 --> 00:07:13.660 right? Yeah, it was enterprise -grade software, 00:07:14.899 --> 00:07:18.060 and it was built to scale. So some of our clients, 00:07:18.980 --> 00:07:23.939 our software powered... When CNBC launched their 00:07:23.939 --> 00:07:26.439 website, their first website, and they had discussions 00:07:26.439 --> 00:07:30.930 about stocks, that was using our software. So 00:07:30.930 --> 00:07:33.470 it was maybe 20 ,000 discussion boards about 00:07:33.470 --> 00:07:36.529 each security. When Healthy on WebMD launched 00:07:36.529 --> 00:07:39.730 and they had discussions with doctors online 00:07:39.730 --> 00:07:43.069 that was using our software. So we tried to build 00:07:43.069 --> 00:07:44.829 software that was really high quality, that would 00:07:44.829 --> 00:07:49.069 scale quite a bit, would be fast. So that's an 00:07:49.069 --> 00:07:52.930 incredible story because I think people can only 00:07:52.930 --> 00:07:56.649 aspire to have been there when it was kind of 00:07:56.649 --> 00:07:58.370 exploding, right? So just before the dot -com 00:07:58.370 --> 00:08:00.379 and then catching all the dot -com, and then 00:08:00.379 --> 00:08:02.220 the massive explosion that happened and then 00:08:02.220 --> 00:08:05.259 everyone would kind of I wouldn't say started 00:08:05.259 --> 00:08:07.019 having a personal computer but they did start 00:08:07.019 --> 00:08:08.660 having access to the internet and started using 00:08:08.660 --> 00:08:10.459 the internet quite a lot right in the beginning 00:08:10.459 --> 00:08:13.620 of 2000 so it's really the time to be alive I 00:08:13.620 --> 00:08:17.060 guess and how right now I know you're quite active 00:08:17.060 --> 00:08:20.279 in the AI how did you transition from you know 00:08:20.279 --> 00:08:22.220 having your own company then selling the company 00:08:22.220 --> 00:08:24.620 and then how do you progress towards AI because 00:08:24.759 --> 00:08:26.660 It's not new for you. You haven't been in the 00:08:26.660 --> 00:08:28.959 space for much, much longer than anyone else 00:08:28.959 --> 00:08:34.460 talking about it. So how did it happen? I got 00:08:34.460 --> 00:08:38.559 my start working on AI about 20 years ago or 00:08:38.559 --> 00:08:43.100 so. And I was working for this consulting company 00:08:43.100 --> 00:08:46.399 that was in two lines of business. One of them 00:08:46.399 --> 00:08:48.679 was management consulting, strategy consulting. 00:08:49.220 --> 00:08:50.980 They recruited me right out of business school 00:08:50.980 --> 00:08:53.610 for that kind of business. And the other line 00:08:53.610 --> 00:08:55.350 of business was what they called data mining. 00:08:57.210 --> 00:09:01.350 And it turned out that they were, you know, we 00:09:01.350 --> 00:09:04.409 were okay at strategy consulting. We would compete 00:09:04.409 --> 00:09:06.990 with, I don't know, McKinsey, Bain, et cetera. 00:09:07.230 --> 00:09:11.769 We were a little bit cheaper. But the work that 00:09:11.769 --> 00:09:14.169 this company was doing in data mining was five, 00:09:14.190 --> 00:09:17.149 10 years ahead of its time. And so I was hired 00:09:17.149 --> 00:09:20.330 for business number one, and I switched to business 00:09:20.330 --> 00:09:23.220 number two as soon as I could because I was just 00:09:23.220 --> 00:09:26.240 shocked and amazed at what they were doing. And 00:09:26.240 --> 00:09:32.360 back then, there were no language models. The 00:09:32.360 --> 00:09:37.419 work was done in SAS. And everything was really 00:09:37.419 --> 00:09:41.840 slow. I started building marketing and risk models 00:09:41.840 --> 00:09:44.960 for one of our biggest clients was American Express. 00:09:45.600 --> 00:09:49.720 So American Express, it's not that they didn't 00:09:49.720 --> 00:09:53.980 use any math. It's that before we came on the 00:09:53.980 --> 00:09:57.139 scene, they would use, let's say, to make a decision 00:09:57.139 --> 00:09:59.500 about whether to give you a line of credit or 00:09:59.500 --> 00:10:01.720 not, they would use a really simple model, like 00:10:01.720 --> 00:10:03.919 linear regression or logistic regression or something 00:10:03.919 --> 00:10:10.940 like that. And we came in in the early aughts, 00:10:11.000 --> 00:10:15.039 early 2000s, and we basically brought forward, 00:10:15.600 --> 00:10:18.259 so linear regression was invented, I don't know, 00:10:18.379 --> 00:10:24.500 in 18 whatever 1870s 1860s was used by some statisticians 00:10:24.500 --> 00:10:27.039 maybe even older than that logistic regressions 00:10:27.039 --> 00:10:30.659 from the 1960s and American Express in like in 00:10:30.659 --> 00:10:33.799 the year 2000 they were using like 40 to 140 00:10:33.799 --> 00:10:37.860 year old technology and we came in and we proposed 00:10:37.860 --> 00:10:40.200 new modeling techniques that had been invented 00:10:40.200 --> 00:10:43.940 in like 1998 and so it was it was like shooting 00:10:43.940 --> 00:10:46.679 fish in a barrel it was just It was bringing 00:10:46.679 --> 00:10:49.659 a tank to a knife fight. It was completely unfair. 00:10:50.320 --> 00:10:52.919 The new techniques worked way better than the 00:10:52.919 --> 00:10:56.840 new regression. So they were able to get much 00:10:56.840 --> 00:10:59.980 better statistical results for all sorts of decisions 00:10:59.980 --> 00:11:02.559 like who should get credit, how much credit should 00:11:02.559 --> 00:11:06.379 they get, who do you target the marketing proposal 00:11:06.379 --> 00:11:09.379 to, who should get to the front of the queue 00:11:09.379 --> 00:11:12.299 when they call for help, all sorts of resource 00:11:12.299 --> 00:11:17.049 allocation decisions. It was an eye -opener for 00:11:17.049 --> 00:11:21.610 me. I was just, I mean, I was talking to customers. 00:11:21.610 --> 00:11:24.950 I was during the day and I was learning the math 00:11:24.950 --> 00:11:29.610 at night and I was having a lot of fun. And then 00:11:29.610 --> 00:11:34.490 it turned out that the business of strategy consulting, 00:11:34.929 --> 00:11:38.490 at least my perspective of it, shifted, and the 00:11:38.490 --> 00:11:42.929 perspective of my clients, shifted from the traditional 00:11:42.929 --> 00:11:47.429 management consulting is top -down. Your client 00:11:47.429 --> 00:11:50.409 asks you the big question, what should our strategy 00:11:50.409 --> 00:11:53.389 be over the next five years? And you start by 00:11:53.389 --> 00:11:55.309 researching the world, researching the market, 00:11:55.629 --> 00:11:57.889 researching this client. And you identify a couple 00:11:57.889 --> 00:12:02.190 of themes or a couple of really big ideas. And 00:12:02.190 --> 00:12:04.629 then you drill down into how to do them. And 00:12:04.629 --> 00:12:07.090 you drill down into what do they mean for this 00:12:07.090 --> 00:12:09.309 business and for that business and for the third 00:12:09.309 --> 00:12:12.529 business. And then you translate that into kind 00:12:12.529 --> 00:12:14.409 of a marching orders, like how do you execute 00:12:14.409 --> 00:12:17.029 it month by month, et cetera. So that is, to 00:12:17.029 --> 00:12:21.049 me, that is like classic top down pyramid principle 00:12:21.049 --> 00:12:26.250 management consulting. And I found that the paradigm 00:12:26.250 --> 00:12:30.710 had shifted from that to sort of bottom up strategy 00:12:30.710 --> 00:12:32.970 consulting. What is bottom up strategy consulting? 00:12:33.509 --> 00:12:35.850 You don't start by asking like big questions 00:12:35.850 --> 00:12:39.529 out of Wikipedia or out of like secondary research. 00:12:39.970 --> 00:12:42.360 You start by interrogating the data. What does 00:12:42.360 --> 00:12:46.080 the data say? And the data will, so first you 00:12:46.080 --> 00:12:47.820 figure out, you collect the data if you don't 00:12:47.820 --> 00:12:50.399 have it, or you buy it. And you interrogate it. 00:12:50.740 --> 00:12:52.799 And then you, you know, the data gives you some 00:12:52.799 --> 00:12:55.179 hypotheses, you validate them. Once you have 00:12:55.179 --> 00:12:57.399 some hypotheses, then you do the rest of it. 00:12:57.399 --> 00:13:00.360 Figure out, you know, how to use them. And I 00:13:00.360 --> 00:13:03.840 found that model of consulting to be so much 00:13:03.840 --> 00:13:06.440 more satisfying than doing sort of top -down 00:13:06.440 --> 00:13:09.830 research, because when I would get questioned 00:13:09.830 --> 00:13:12.190 on something, I'd have backup, there'd be some, 00:13:13.049 --> 00:13:14.789 you know, I'm a numbers guy, there'd be some 00:13:14.789 --> 00:13:17.289 scientific basis for what I'm saying. Now it's 00:13:17.289 --> 00:13:19.409 not always the right, it's not the only right 00:13:19.409 --> 00:13:21.370 answer, it's not the only way to do it, but it's 00:13:21.370 --> 00:13:23.809 a way that really worked for me and it worked 00:13:23.809 --> 00:13:28.629 for more and more of my clients and so I learned 00:13:28.629 --> 00:13:32.509 that way of thinking at this company and then, 00:13:32.509 --> 00:13:34.809 you know, after the company, that the company 00:13:34.809 --> 00:13:38.379 grew, got acquired, I left and I started doing 00:13:38.379 --> 00:13:40.200 some of my own consulting and this was the way 00:13:40.200 --> 00:13:45.039 that served me. But you know, Ricardo, I've been 00:13:45.039 --> 00:13:48.700 doing this for 20 years and I have never seen, 00:13:48.779 --> 00:13:51.360 like what's been happening in AI over the last 00:13:51.360 --> 00:13:53.980 year has just been completely different from 00:13:53.980 --> 00:13:56.720 the last, the 19 years preceded that. I was just 00:13:56.720 --> 00:13:58.879 about to ask you about that because it feels 00:13:58.879 --> 00:14:01.370 that people... You know everyone's talking about 00:14:01.370 --> 00:14:03.730 AI now and rightfully so but they don't actually 00:14:03.730 --> 00:14:06.309 realize there has been a lot of growth up until 00:14:06.309 --> 00:14:08.830 this stage right now is just the point is just 00:14:08.830 --> 00:14:10.669 the tip of the iceberg there's actually quite 00:14:10.669 --> 00:14:13.889 a big chunk of ice underneath you know the waves 00:14:13.889 --> 00:14:16.029 so I was just about to ask you what do you think 00:14:16.029 --> 00:14:19.350 is happening now how do you see this how impactful 00:14:19.350 --> 00:14:23.690 do you think the current evolution of AI is for 00:14:23.690 --> 00:14:26.669 for everything that we do? You know I can tell 00:14:26.669 --> 00:14:30.159 you that the feeling that I get from being in 00:14:30.159 --> 00:14:33.480 this moment, I feel really fortunate. And it's 00:14:33.480 --> 00:14:36.379 very similar to the feeling that I remember having 00:14:36.379 --> 00:14:40.179 in like 1995 working in an internet company. 00:14:40.659 --> 00:14:44.879 I mean, it's really exciting. And at the same 00:14:44.879 --> 00:14:50.159 time, it feels so, so early. It feels like we're 00:14:50.159 --> 00:14:52.879 trying to build these things. We're excited about 00:14:52.879 --> 00:14:55.899 these new capabilities. But there's so much left 00:14:55.899 --> 00:14:59.169 to do. There's just this wide, wide open space, 00:14:59.350 --> 00:15:03.610 and we're just at the tip of the iceberg. It's 00:15:03.610 --> 00:15:05.929 not even the first inning. It's like the first 00:15:05.929 --> 00:15:08.110 five minutes of the game. That's what it feels 00:15:08.110 --> 00:15:13.049 like to me. But in terms of what actually happened, 00:15:13.190 --> 00:15:16.230 I don't have an especially contrarian view. I 00:15:16.230 --> 00:15:20.309 think the fact that ChatGPT is the fastest growing 00:15:20.309 --> 00:15:23.909 product in history and suddenly unlocked all 00:15:23.909 --> 00:15:26.389 of these capabilities and showed that you could 00:15:26.389 --> 00:15:29.309 just You could just throw some more data and 00:15:29.309 --> 00:15:32.210 compute at this problem and get a magic moment. 00:15:32.470 --> 00:15:36.110 And now that toothpaste is out of the tube and 00:15:36.110 --> 00:15:39.610 lots of other people are able to replicate parts 00:15:39.610 --> 00:15:42.730 of it. It's very, very exciting. It feels like 00:15:42.730 --> 00:15:48.029 to me, it feels at least as impactful as the 00:15:48.029 --> 00:15:51.629 creation of the internet. And maybe at least 00:15:51.629 --> 00:15:54.210 as impactful as maybe even bigger than that. 00:15:54.210 --> 00:15:56.980 Maybe it's electricity, maybe it's fire. but 00:15:56.980 --> 00:15:59.419 I'm quite quite optimistic about it. I think 00:15:59.419 --> 00:16:02.080 it's going to be a new revolution, the same way 00:16:02.080 --> 00:16:04.059 we had all these revolutions throughout history. 00:16:04.799 --> 00:16:07.519 I think we're at that stage where everything's 00:16:07.519 --> 00:16:09.980 going to change and rightfully so and for the 00:16:09.980 --> 00:16:17.000 best. I was listening to a famous podcast the 00:16:17.000 --> 00:16:19.019 other day and he was saying it's the same change 00:16:19.019 --> 00:16:23.059 we watch when people transition from having horses 00:16:23.059 --> 00:16:25.860 everywhere to having automobiles. And that happened 00:16:25.860 --> 00:16:28.639 overnight, right? It was night and day. And I 00:16:28.639 --> 00:16:29.960 think that's probably what's going to happen 00:16:29.960 --> 00:16:33.059 with this type of AI. Shouldn't we be concerned? 00:16:33.820 --> 00:16:35.399 Because there's a lot, you know, if you read, 00:16:35.399 --> 00:16:37.600 I don't know, those type of weird newspapers 00:16:37.600 --> 00:16:39.440 and say, Oh my God, AI is coming to take your 00:16:39.440 --> 00:16:41.960 job in this. But to be honest, I think we're 00:16:41.960 --> 00:16:43.279 at that stage where it's actually quite good. 00:16:43.399 --> 00:16:44.700 But should we be concerned? What do you think? 00:16:45.820 --> 00:16:47.759 Yeah, it's a really good question. I actually 00:16:47.759 --> 00:16:50.820 guess that a lot, Ricardo. So I think that there 00:16:50.820 --> 00:16:54.399 are there are two future scenarios that could 00:16:54.399 --> 00:16:57.379 cause concern. And I think that they're actually 00:16:57.379 --> 00:16:59.399 very, very different from each other. And one 00:16:59.399 --> 00:17:04.500 really is or assuming one isn't. So the first 00:17:04.500 --> 00:17:08.460 one is AI is gonna take my job. That is a very 00:17:08.460 --> 00:17:12.359 concerning future scenario. The second one is 00:17:12.359 --> 00:17:17.720 AI is gonna take my freedom. So I actually think 00:17:17.720 --> 00:17:20.460 that the first scenario where AI is gonna take 00:17:20.460 --> 00:17:24.450 my job That is, it looks really concerning, it 00:17:24.450 --> 00:17:26.829 sounds really scary, but I actually think that 00:17:26.829 --> 00:17:30.150 that is the utopian scenario. There are a lot 00:17:30.150 --> 00:17:35.109 of jobs that I wish AI would take. There are 00:17:35.109 --> 00:17:38.269 a lot of people that are doing jobs that are 00:17:38.269 --> 00:17:41.109 just gonna be better done by a computer, and 00:17:41.109 --> 00:17:44.009 they're not gonna be jobs for humans anymore. 00:17:45.690 --> 00:17:47.890 Those humans are gonna be much happier not doing 00:17:47.890 --> 00:17:52.000 those jobs. I think also that when people are 00:17:52.000 --> 00:17:54.500 looking at that first scenario of AI is going 00:17:54.500 --> 00:17:59.619 to take my job, they stop visualizing it at the 00:17:59.619 --> 00:18:03.000 moment that the job is lost. But I don't think 00:18:03.000 --> 00:18:05.220 they fully appreciate all of the knock -on effects. 00:18:06.000 --> 00:18:09.359 So if AI takes all of these jobs, that means 00:18:09.359 --> 00:18:12.140 that it's going to be able to do these jobs better 00:18:12.140 --> 00:18:15.420 and cheaper. That means that the things that 00:18:15.420 --> 00:18:17.859 these jobs produce are going to become so much 00:18:17.859 --> 00:18:21.309 better and cheaper. So imagine that you need 00:18:21.309 --> 00:18:23.130 a proofreader, or you need a lawyer, you need 00:18:23.130 --> 00:18:27.569 an editor for this podcast, or a booker to help 00:18:27.569 --> 00:18:30.670 you find guests so you can do a little bit better 00:18:30.670 --> 00:18:34.289 than me next time. I thought we were going to 00:18:34.289 --> 00:18:36.990 do any better than you, and I think we've reached 00:18:36.990 --> 00:18:40.089 the pinnacle of the highest quality guests ever. 00:18:40.549 --> 00:18:43.569 Thank you, Ricardo. But you can imagine how, 00:18:43.569 --> 00:18:46.690 as an entrepreneur, you consume all of these 00:18:46.690 --> 00:18:49.630 services. Maybe you do them yourself, maybe as 00:18:49.630 --> 00:18:52.569 you grow, you'll decide to outsource them. So 00:18:52.569 --> 00:18:55.069 today, your option is basically outsource it 00:18:55.069 --> 00:18:57.329 to a human and hopefully they're good at what 00:18:57.329 --> 00:19:00.369 they do. And imagine that in that future, and 00:19:00.369 --> 00:19:02.650 maybe this future is not so far away, you can 00:19:02.650 --> 00:19:04.990 actually outsource it to a machine. That machine 00:19:04.990 --> 00:19:06.869 does a really good job of booking guests for 00:19:06.869 --> 00:19:11.470 you. That means that there's going to be the 00:19:11.470 --> 00:19:14.349 quality of your podcast goes up, the cost of 00:19:14.349 --> 00:19:16.789 producing it goes down. You get to keep more 00:19:16.789 --> 00:19:20.009 money from All the other sources of revenue that 00:19:20.009 --> 00:19:22.309 you have the reason why you do this podcast whether 00:19:22.309 --> 00:19:25.509 it's advertising or you know your other businesses 00:19:25.509 --> 00:19:30.150 and so All the services that you consume get 00:19:30.150 --> 00:19:33.630 cheaper. You will feel richer. You will be richer 00:19:33.630 --> 00:19:36.970 And it's not just you it's like everybody that's 00:19:36.970 --> 00:19:39.410 consuming all of these white -collar services, 00:19:39.410 --> 00:19:42.349 you know professional services architects and 00:19:42.349 --> 00:19:45.869 lawyers and designers and programmers and all 00:19:45.869 --> 00:19:49.960 these things become cheap so To me, if you zoom 00:19:49.960 --> 00:19:53.220 out on that society, it means that all the things 00:19:53.220 --> 00:19:56.460 that we need become really cheap. So we will 00:19:56.460 --> 00:20:00.160 feel very rich. Just imagine you're now a billionaire 00:20:00.160 --> 00:20:03.880 and all these things that you want to buy, they 00:20:03.880 --> 00:20:06.160 don't mean anything. The prices don't mean anything 00:20:06.160 --> 00:20:08.519 to you. You can afford as much as you want. And 00:20:08.519 --> 00:20:11.319 so I think that that's a world where there's 00:20:11.319 --> 00:20:14.779 huge deflation. The prices of lots of goods and 00:20:14.779 --> 00:20:17.000 services drop because the AI is doing them more 00:20:17.000 --> 00:20:19.640 efficiently. If there's so much deflation, that 00:20:19.640 --> 00:20:22.920 means basically we all feel richer, even if we 00:20:22.920 --> 00:20:26.079 don't make the numbers might not be higher. It's 00:20:26.079 --> 00:20:28.480 not like my income went up, but my expenses went 00:20:28.480 --> 00:20:31.980 down. And so I feel richer and I am richer. And 00:20:31.980 --> 00:20:35.039 I think in that world, just fewer people need 00:20:35.039 --> 00:20:41.319 to work. So in that world, I think that those 00:20:41.319 --> 00:20:46.819 people who lost their jobs, maybe they'll find 00:20:46.819 --> 00:20:50.160 other jobs, you know, developing, you know, something 00:20:50.160 --> 00:20:53.720 that's higher value. Some of them will become, 00:20:53.720 --> 00:20:56.819 you know, maybe engineering. Instead of being 00:20:56.819 --> 00:20:59.519 a lawyer, you might be an engineer training law 00:20:59.519 --> 00:21:02.319 chatbots or something like that. But it's not 00:21:02.319 --> 00:21:05.180 necessarily like not everybody's going to switch. 00:21:05.359 --> 00:21:08.339 Some people will become artists because whatever, 00:21:08.339 --> 00:21:10.539 you know, basically whatever change they have 00:21:10.539 --> 00:21:13.359 left in their pocket is enough to live on. And 00:21:13.359 --> 00:21:16.150 as a society, I also think that we'll be able 00:21:16.150 --> 00:21:17.970 to afford to support these people because we 00:21:17.970 --> 00:21:20.549 will have so much because all these things become 00:21:20.549 --> 00:21:24.109 cheaper. So I think that world is a world where 00:21:24.109 --> 00:21:26.789 there's going to be something like universal 00:21:26.789 --> 00:21:29.069 basic income and we'll just be able to take care 00:21:29.069 --> 00:21:31.430 of all these people. It's not going to be a problem. 00:21:31.970 --> 00:21:34.430 That is one of those that you never know how 00:21:34.430 --> 00:21:37.710 it's going to go, right? But I was just thinking 00:21:37.710 --> 00:21:40.690 about something that people being scared about 00:21:40.690 --> 00:21:42.990 all this transformation and how it's going to 00:21:42.990 --> 00:21:45.440 impact. I think people fail to realize that they 00:21:45.440 --> 00:21:48.660 can actually upskill themselves by using AI. 00:21:49.599 --> 00:21:51.960 For example, think about tutoring and you're 00:21:51.960 --> 00:21:53.400 in school, you're trying to learn something, 00:21:53.519 --> 00:21:56.380 the teachers are not that good. Nowadays, you 00:21:56.380 --> 00:21:57.839 can probably go on YouTube, you can probably 00:21:57.839 --> 00:21:59.640 go on Google and find the right information, 00:21:59.819 --> 00:22:01.960 but it never really does explain it to the level 00:22:01.960 --> 00:22:04.720 detail you need to. However, on ChatGPT, it will 00:22:04.720 --> 00:22:06.640 tell you exactly what you need to do, how you 00:22:06.640 --> 00:22:10.299 need to do it. It just opens up possibilities, 00:22:10.339 --> 00:22:13.000 right? It's a bit like having access to books 00:22:13.000 --> 00:22:15.839 in the 15th century for the printing press You'll 00:22:15.839 --> 00:22:18.319 be the same thing the same change, right? So 00:22:18.319 --> 00:22:20.759 it's really like that regard I would say it's 00:22:20.759 --> 00:22:23.920 like going from a world where you have access 00:22:23.920 --> 00:22:26.599 to books To a world where you have access to 00:22:26.599 --> 00:22:29.779 a tutor like a really good tutor that knows you 00:22:29.779 --> 00:22:31.980 The tutor knows what you like and what you don't 00:22:31.980 --> 00:22:34.619 like, you know, whether you like maybe you like 00:22:34.619 --> 00:22:37.970 learning by reading or by listening or by watching 00:22:37.970 --> 00:22:40.309 things. And it knows what you're strong at and 00:22:40.309 --> 00:22:42.690 what you're weak at. And it basically helps you. 00:22:42.910 --> 00:22:46.130 It's a personalized tutor. So that personalized 00:22:46.130 --> 00:22:48.589 tutor is, you know, it's very expensive in the 00:22:48.589 --> 00:22:51.470 pre -AI world. Only a few people can afford one. 00:22:51.849 --> 00:22:53.549 But in the post -AI world, you know, everybody 00:22:53.549 --> 00:22:56.390 could, you know, learn that way and yeah, can 00:22:56.390 --> 00:22:58.990 be very effective. Now we were talking about 00:22:58.990 --> 00:23:01.650 some of the dangers of AI which is also a very 00:23:01.650 --> 00:23:03.890 hot topic So what do you think will be like in 00:23:03.890 --> 00:23:06.289 terms of the government's using it and every 00:23:06.289 --> 00:23:08.789 being? Everyone being repressed and we have killer 00:23:08.789 --> 00:23:10.390 robots walking down the street, which by the 00:23:10.390 --> 00:23:15.309 way, it's completely People are absolutely crazy 00:23:15.309 --> 00:23:17.089 when I say that but what do you think? Those 00:23:17.089 --> 00:23:18.509 are some of the things we should be concerned 00:23:18.509 --> 00:23:20.849 about because there is quite a lot of capabilities 00:23:20.849 --> 00:23:23.049 and I know some of it But I'm pretty sure you 00:23:23.049 --> 00:23:25.410 know a lot more about it than I do. Should we 00:23:25.410 --> 00:23:28.059 be concerned about it or not at all? You know, 00:23:28.059 --> 00:23:30.079 Ricardo, I may be one of those crazy people. 00:23:30.240 --> 00:23:33.059 I am pretty concerned about it, actually. So 00:23:33.059 --> 00:23:35.220 when we talked about the two scenarios before, 00:23:35.720 --> 00:23:38.200 where the first scenario is AI takes my job, 00:23:38.779 --> 00:23:41.339 and the other scenario is AI takes my freedom, 00:23:41.859 --> 00:23:44.480 I'm really, really worried about that scenario. 00:23:44.619 --> 00:23:47.420 There's a couple of different ways that AI can 00:23:47.420 --> 00:23:53.400 be used to take our freedom. No, the simplest 00:23:53.400 --> 00:23:55.660 way to think about it is the like the terminator 00:23:55.660 --> 00:23:58.420 scenario. So we all we've all seen the movies, 00:23:58.420 --> 00:24:01.299 you know, approximately, you know, approximately 00:24:01.299 --> 00:24:04.980 every future science fiction story about AI is 00:24:04.980 --> 00:24:07.819 a disaster story. And it's it's really scary. 00:24:08.440 --> 00:24:12.480 And working with this technology, I cannot tell 00:24:12.480 --> 00:24:15.339 you that it's like there's no way for it to happen. 00:24:15.460 --> 00:24:18.200 Like I cannot I can't push that probability down 00:24:18.200 --> 00:24:22.470 to zero. And it's it's very, very scary. Now, 00:24:22.789 --> 00:24:24.549 it's not like, you know, I'm not losing sleep 00:24:24.549 --> 00:24:27.829 over it, but I think that there's, you know, 00:24:27.930 --> 00:24:32.650 there's a real chance that it's called takeoff, 00:24:32.890 --> 00:24:35.630 where the capabilities of AI, you know, they 00:24:35.630 --> 00:24:37.630 kind of grow and grow and grow, and then they 00:24:37.630 --> 00:24:41.809 suddenly take off at a rate that is just too 00:24:41.809 --> 00:24:44.609 fast for the scientists to contain. I think that 00:24:44.609 --> 00:24:48.559 is, it's a real risk. I'm worried about it. And 00:24:48.559 --> 00:24:51.180 also, you know, even if you ignore takeoff, even 00:24:51.180 --> 00:24:54.279 if takeoff is too science fiction to think about, 00:24:54.740 --> 00:24:59.720 AI is the kind of technology where it's really 00:24:59.720 --> 00:25:05.079 asymmetrical. It's not like nuclear power, for 00:25:05.079 --> 00:25:07.859 example. So, you know, if you're worried about 00:25:07.859 --> 00:25:13.519 nuclear proliferation, to create a nuclear bomb 00:25:13.519 --> 00:25:16.539 requires the resources of a nation state. It's 00:25:16.539 --> 00:25:19.000 incredibly difficult. You need to set up this 00:25:19.000 --> 00:25:21.640 enrichment program. You need really high -grade 00:25:21.640 --> 00:25:23.900 scientists and equipment and all these. You have 00:25:23.900 --> 00:25:25.980 to buy these things that are not available for 00:25:25.980 --> 00:25:29.400 sale. And so it takes so much money and so much 00:25:29.400 --> 00:25:33.079 time that only a few countries in the world have 00:25:33.079 --> 00:25:34.779 been able to do it. And those countries have 00:25:34.779 --> 00:25:37.700 been able to contain this technology to some 00:25:37.700 --> 00:25:41.240 extent. It's not perfect either. AI is not like 00:25:41.240 --> 00:25:46.980 that. AI, as an individual, you can have a breakthrough. 00:25:47.900 --> 00:25:50.319 Like today, we're not too late. We're early enough 00:25:50.319 --> 00:25:52.940 in the process where an individual can publish 00:25:52.940 --> 00:25:55.559 a paper and invent something that nobody's ever 00:25:55.559 --> 00:25:59.940 thought of. And that thing significantly increases 00:25:59.940 --> 00:26:02.059 the capabilities of the state of the art of what 00:26:02.059 --> 00:26:04.299 you can do with AI. I can give you lots of examples 00:26:04.299 --> 00:26:07.460 of papers like that. But from a risk perspective, 00:26:08.660 --> 00:26:12.660 just imagine that a single scientist could invent 00:26:12.660 --> 00:26:15.309 something that makes I don't know, that makes 00:26:15.309 --> 00:26:18.329 nuclear bombs more destructive or more attainable. 00:26:19.250 --> 00:26:23.289 That would be scary, right? So with AI, it's 00:26:23.289 --> 00:26:25.769 not just that an individual can create a breakthrough 00:26:25.769 --> 00:26:28.930 to improve, to increase the power. It's also 00:26:28.930 --> 00:26:35.349 that training these large models and all of the 00:26:35.349 --> 00:26:38.029 infrastructure around AI is getting cheaper at 00:26:38.029 --> 00:26:43.890 a very, very fast rate. Yeah, so it's really 00:26:43.890 --> 00:26:46.150 asymmetrical. By the way, I should mention, these 00:26:46.150 --> 00:26:50.769 ideas are not mine. I read a book that I'd like 00:26:50.769 --> 00:26:54.210 to recommend to you and your viewers. The book 00:26:54.210 --> 00:26:57.869 is called The Coming Wave. It's written by Mustafa 00:26:57.869 --> 00:27:04.029 Suleiman, who is one of the co -founders of DeepMind, 00:27:04.930 --> 00:27:08.190 Google DeepMind. And it's mostly a book about 00:27:08.190 --> 00:27:12.500 what can go wrong. 10 % of the book are about 00:27:12.500 --> 00:27:20.759 what we can do about it Yeah, yeah, so so it's 00:27:20.759 --> 00:27:23.299 a it's a really it's a really interesting book 00:27:23.299 --> 00:27:26.700 in the last chapter of that book he talks about 00:27:26.700 --> 00:27:31.400 a framework for For defending against these these 00:27:31.400 --> 00:27:34.680 problems and it's You know, I'm not going to 00:27:34.680 --> 00:27:36.759 do it justice, but it's it's these concentric 00:27:36.759 --> 00:27:38.960 circles It's the way the way that you design 00:27:38.960 --> 00:27:42.819 any kind of security system is not a single barrier 00:27:42.819 --> 00:27:45.500 or single line of defense, but many. So that 00:27:45.500 --> 00:27:48.119 if you breach the external line of defense, you 00:27:48.119 --> 00:27:49.480 have the internal one, you have the internal 00:27:49.480 --> 00:27:54.319 one. It's like a bullseye. And so, again, I don't 00:27:54.319 --> 00:27:57.220 remember all of them, but one of them starts 00:27:57.220 --> 00:27:59.759 very, very close to the researchers. And it basically 00:27:59.759 --> 00:28:04.900 says that every model should have a built -in 00:28:04.900 --> 00:28:08.039 auditing system that describes what the model 00:28:08.039 --> 00:28:12.259 does and is available to not just the team researching 00:28:12.259 --> 00:28:14.259 it, but to other people that want to inspect 00:28:14.259 --> 00:28:17.720 that model. So it's kind of like, you know, you 00:28:17.720 --> 00:28:19.500 should be able to have like a little diagnostic, 00:28:19.559 --> 00:28:21.799 like you have a diagnostic port on your car. 00:28:21.980 --> 00:28:24.420 You know, you can plug in a USB or something. 00:28:24.940 --> 00:28:27.460 The car will tell you. All cars have it since 00:28:27.460 --> 00:28:29.440 then. All cars. Right. That's right. So the car 00:28:29.440 --> 00:28:32.440 will tell you kind of how it's feeling, you know, 00:28:32.440 --> 00:28:35.119 if some things are rusty, some things are according 00:28:35.119 --> 00:28:38.059 to spec or not. So what he suggests is there 00:28:38.059 --> 00:28:39.920 should be, and it's a standard port, like the 00:28:39.920 --> 00:28:43.519 same USB thing will work on any car and can read 00:28:43.519 --> 00:28:45.740 data and say cars have standardized on this. 00:28:46.039 --> 00:28:48.799 So he's suggesting that should be a diagnostic 00:28:48.799 --> 00:28:52.119 port on AI models. Whether that model is built 00:28:52.119 --> 00:28:57.440 by OpenAI or Anthropic or by WeChat, there should 00:28:57.440 --> 00:29:03.720 be a standard way to inspect them. Another wider 00:29:03.720 --> 00:29:08.380 sort of circle that he talks about is that all 00:29:08.380 --> 00:29:11.700 AI labs should run and publish their test results 00:29:11.700 --> 00:29:15.839 in a way that in a public place so that individual 00:29:15.839 --> 00:29:18.920 researchers can inspect those results and spot 00:29:18.920 --> 00:29:22.079 any kind of problem. So there would be like some 00:29:22.079 --> 00:29:24.400 kind of protocol, they publish it, and you and 00:29:24.400 --> 00:29:27.339 I can go check that removed data and see if we, 00:29:27.539 --> 00:29:30.900 you know, and that's transparency. So he basically 00:29:30.900 --> 00:29:33.400 has like, you know, six or seven or eight of 00:29:33.400 --> 00:29:35.279 these concentric circles and they go all the 00:29:35.279 --> 00:29:38.849 way from the individual team up to something 00:29:38.849 --> 00:29:41.849 that requires the society to make an investment. 00:29:42.890 --> 00:29:46.329 But I hope some of these things come true. I 00:29:46.329 --> 00:29:49.470 hope that these mechanisms get adopted. How do 00:29:49.470 --> 00:29:56.730 you prevent certain states from developing capabilities 00:29:56.730 --> 00:29:59.109 like this? Is there a way of controlling it or 00:29:59.109 --> 00:30:01.269 not at all? Because right now, as you're talking 00:30:01.269 --> 00:30:07.400 about nuclear nuclear technology, and we know 00:30:07.400 --> 00:30:10.539 there are certain constraints about getting uranium, 00:30:10.920 --> 00:30:13.299 getting plutonium, then reaching them, and we 00:30:13.299 --> 00:30:15.700 can track that by satellite, saying, okay, you 00:30:15.700 --> 00:30:17.759 have this, you came from that source, so forth, 00:30:17.779 --> 00:30:20.660 so on. How do we stop countries like, I'm going 00:30:20.660 --> 00:30:22.480 to give North Korea, I hope I never go to North 00:30:22.480 --> 00:30:24.319 Korea because then I won't be allowed in, I'll 00:30:24.319 --> 00:30:26.500 probably get stuck there, but how do we prevent 00:30:26.500 --> 00:30:30.240 countries like North Korea from creating capabilities 00:30:30.240 --> 00:30:35.900 that would So how do we solve that? Yeah, it's 00:30:35.900 --> 00:30:38.599 a serious question, Ricardo. I don't have an 00:30:38.599 --> 00:30:42.180 answer. I think that the only way to compel nation 00:30:42.180 --> 00:30:44.599 states to do something is through some kind of 00:30:44.599 --> 00:30:47.480 an international treaty and political pressure. 00:30:49.359 --> 00:30:52.940 There's no magic bullet here. Some of these things 00:30:52.940 --> 00:30:56.480 require very, very broad agreement and broad 00:30:56.480 --> 00:31:00.319 coalitions to make something happen, but not 00:31:00.319 --> 00:31:05.019 all of it. So there are some things that we can 00:31:05.019 --> 00:31:08.440 just start doing the right thing locally. And 00:31:08.440 --> 00:31:10.619 there are some things that just will require 00:31:10.619 --> 00:31:13.339 massive agreement and will probably take a while. 00:31:13.619 --> 00:31:17.720 In that sense, I think that this problem is not 00:31:17.720 --> 00:31:23.599 unlike the environmentalist movement. We have 00:31:23.599 --> 00:31:27.680 to take care of our planet. And how do we force 00:31:27.680 --> 00:31:30.299 the worst polluter to stop polluting? it's going 00:31:30.299 --> 00:31:32.880 to be really hard to force the worst polluter. 00:31:33.059 --> 00:31:35.400 But it doesn't mean that we should just like 00:31:35.400 --> 00:31:37.740 throw up our hands and say, well, if China is 00:31:37.740 --> 00:31:40.279 going to keep polluting in, you know, such and 00:31:40.279 --> 00:31:42.619 such province, it means that I don't have to 00:31:42.619 --> 00:31:45.500 worry about the environment at all. So you kind 00:31:45.500 --> 00:31:47.819 of you take whatever steps you can. You start 00:31:47.819 --> 00:31:49.960 with your own bed, right? You make your bed, 00:31:50.220 --> 00:31:52.660 keep your, you know, you lower your own carbon 00:31:52.660 --> 00:31:55.440 footprint, think personally, and then you think 00:31:55.440 --> 00:31:58.720 about your circle of friends. your local community 00:31:58.720 --> 00:32:02.000 and kind of grow from there but then the question 00:32:02.000 --> 00:32:05.420 would be because we can always the way we present 00:32:05.420 --> 00:32:07.859 ourselves and the way we organize ourselves it 00:32:07.859 --> 00:32:11.180 is done in respect with our values right and 00:32:11.180 --> 00:32:13.960 much more so however when we are living in a 00:32:13.960 --> 00:32:16.059 global society the other fellow might not have 00:32:16.059 --> 00:32:17.880 the same values and they might respect the same 00:32:17.880 --> 00:32:20.950 thing so it's a bit like going to war When you 00:32:20.950 --> 00:32:23.170 do go to war, you need to be aware that your 00:32:23.170 --> 00:32:25.650 enemy might not have the same rules of engagement. 00:32:25.910 --> 00:32:28.789 They will not comply with the Geneva Convention, 00:32:28.789 --> 00:32:30.910 for example. So they'll fight in a very dirty 00:32:30.910 --> 00:32:34.269 way. How do we stop this from happening? It creates 00:32:34.269 --> 00:32:37.710 this dilemma. It's going to be tricky, no? Yeah, 00:32:37.849 --> 00:32:39.109 it's definitely going to be tricky, Ricardo. 00:32:39.269 --> 00:32:43.690 I wish I had a quick answer for you. But I think 00:32:43.690 --> 00:32:47.890 that the stakes of solving this with AI are very, 00:32:47.890 --> 00:32:52.890 very high. Because if we don't get it right in 00:32:52.890 --> 00:32:56.549 warfare, then the risk is that there's going 00:32:56.549 --> 00:32:58.609 to be a war that doesn't follow the rules and 00:32:58.609 --> 00:33:00.910 it's really going to hurt the combatants in that 00:33:00.910 --> 00:33:06.750 war. If we don't get it right in AI, the risk 00:33:06.750 --> 00:33:12.990 is that there is AI takeoff or an AI weapon and 00:33:12.990 --> 00:33:15.950 it just wipes us out. I mean, I think it's an 00:33:15.950 --> 00:33:19.599 existential risk for the planet. It's not unlike 00:33:19.599 --> 00:33:23.700 global warming. Question on that. So there is, 00:33:23.700 --> 00:33:29.440 I guess, a gap right now between the online space 00:33:29.440 --> 00:33:31.779 and the actual physical capabilities of AI, right? 00:33:32.039 --> 00:33:36.829 We don't have robots yet. Do you think all these 00:33:36.829 --> 00:33:39.009 bad nation states will be able to produce, for 00:33:39.009 --> 00:33:41.529 example, killer robots or have that enough technology 00:33:41.529 --> 00:33:44.309 to actually disrupt this fully or resume very, 00:33:44.309 --> 00:33:46.569 very far away? Because if we look at that company, 00:33:46.589 --> 00:33:49.009 what's it called? Boston Dynamics? Is it Boston 00:33:49.009 --> 00:33:52.569 Dynamics? All those advanced robots? So when 00:33:52.569 --> 00:33:55.109 I think of, okay, let's incorporate AI here, 00:33:55.410 --> 00:33:57.930 it's actually quite powerful. But then I also 00:33:57.930 --> 00:33:59.369 know that the country's not having it. So that 00:33:59.369 --> 00:34:01.730 makes me feel a little bit safer at night. Do 00:34:01.730 --> 00:34:04.190 you think that that transition will be easy or 00:34:04.190 --> 00:34:06.940 not? No, it definitely won't be easy and also 00:34:06.940 --> 00:34:09.619 I think we're I'm getting a little bit away from 00:34:09.619 --> 00:34:13.260 my expertise. So I Most of the work that I've 00:34:13.260 --> 00:34:17.420 been doing has been for Commercial AI, you know 00:34:17.420 --> 00:34:20.280 in the business world and it's mostly been with 00:34:20.280 --> 00:34:23.559 text So I haven't had a lot of personal experience 00:34:23.559 --> 00:34:27.880 with robots for weapons But I can tell you that 00:34:27.880 --> 00:34:33.969 you know Autonomous weapons AI weapons that have 00:34:33.969 --> 00:34:37.309 the, they can actually make the decision to pull 00:34:37.309 --> 00:34:40.170 the trigger. They exist. They, you know, several 00:34:40.170 --> 00:34:42.690 nations have them, including the US. And it's 00:34:42.690 --> 00:34:46.650 a really, it's a really scary thing. And there's, 00:34:46.670 --> 00:34:51.269 there's this, you know, the arms race is pushing 00:34:51.269 --> 00:34:55.329 a lot of the participants to be ever less careful 00:34:55.329 --> 00:34:59.010 and put ever more powerful weapons into the field. 00:34:59.030 --> 00:35:05.980 That's the way it's always been. But Yeah, it's 00:35:05.980 --> 00:35:08.739 a very scary kind of race here and I hope that 00:35:08.739 --> 00:35:13.360 we pull back from Brick. So on that note about 00:35:13.360 --> 00:35:15.619 not arms race or anything like that but about 00:35:15.619 --> 00:35:18.880 AI and technology, I know you're also a mentor 00:35:18.880 --> 00:35:21.659 for Techstars. I was just going to ask you, do 00:35:21.659 --> 00:35:24.739 you think that a startup can still exist without 00:35:24.739 --> 00:35:28.599 AI on something that they do or that's completely 00:35:28.599 --> 00:35:34.099 off? Everything needs to be around AI. I heard 00:35:34.099 --> 00:35:38.340 a recent statement from a venture capitalist. 00:35:38.519 --> 00:35:41.900 I think it was in Sequoia or maybe A16Z or one 00:35:41.900 --> 00:35:45.500 of those large funds. And they said that over 00:35:45.500 --> 00:35:50.340 90 % of incoming pitches are AI related. So I 00:35:50.340 --> 00:35:55.280 guess it's possible, but pretty slim. I think 00:35:55.280 --> 00:35:57.900 part of it is supply and demand. There's a lot 00:35:57.900 --> 00:36:02.130 of capital. Looking for AI ideas and there are 00:36:02.130 --> 00:36:04.150 a lot of entrepreneurs that are trying to be 00:36:04.150 --> 00:36:07.190 where the capital is going Yeah, that makes sense, 00:36:07.190 --> 00:36:10.550 but would you so? Because I know you also teach 00:36:10.550 --> 00:36:13.170 people when they help them out So what advice 00:36:13.170 --> 00:36:15.769 would you give to an entrepreneur just launching 00:36:15.769 --> 00:36:18.789 their startup in terms of incorporating AI into 00:36:18.789 --> 00:36:27.349 their idea? I Think that AI is a tool it helps 00:36:27.349 --> 00:36:31.599 you achieve certain things. It helps you maybe 00:36:31.599 --> 00:36:34.000 get where you're going, but it doesn't help you 00:36:34.000 --> 00:36:38.139 decide where you're going. It's not a destination. 00:36:42.239 --> 00:36:45.300 It doesn't produce value by itself. It helps 00:36:45.300 --> 00:36:47.519 you produce the value that you want to produce. 00:36:47.860 --> 00:36:53.840 So it's a means to an end. It's technology. It's 00:36:53.840 --> 00:36:57.639 not a product. If that makes sense. So my advice 00:36:57.639 --> 00:36:59.900 to the entrepreneur, you know, if the entrepreneur 00:36:59.900 --> 00:37:02.539 tells me, you know, actually, you know, I recently 00:37:02.539 --> 00:37:08.960 had this I had, you know, I had a startup founder 00:37:08.960 --> 00:37:11.480 and told me, look, I, you know, I built this 00:37:11.480 --> 00:37:16.539 AI based product. And it's really cool. It has 00:37:16.539 --> 00:37:19.920 all these technologies in it. And I'm having 00:37:19.920 --> 00:37:21.920 like a lot of trouble selling it, you know, it's 00:37:21.920 --> 00:37:24.699 not really getting the kind of pick up the market 00:37:24.699 --> 00:37:28.340 traction. that I'm looking for. So what should 00:37:28.340 --> 00:37:32.219 I do? Should I maybe raise some money, maybe 00:37:32.219 --> 00:37:36.039 lower my price, maybe fix my ad, maybe I should 00:37:36.039 --> 00:37:42.019 take on a partner? How do I fix it? And the problem 00:37:42.019 --> 00:37:46.000 with that question is that users don't really 00:37:46.000 --> 00:37:48.380 care if you're using AI or not. Maybe investors 00:37:48.380 --> 00:37:52.059 do, but users don't care. Users have certain 00:37:52.059 --> 00:37:55.119 pain points, and as entrepreneurs, it's our job 00:37:55.119 --> 00:37:58.659 to fix those pain points using whatever legal 00:37:58.659 --> 00:38:02.760 means we can come up with. And it just so happens 00:38:02.760 --> 00:38:07.920 that a lot of problems could probably be fixed 00:38:07.920 --> 00:38:10.760 with the application of these new capabilities 00:38:10.760 --> 00:38:13.699 that just got unlocked. But the user, again, 00:38:13.780 --> 00:38:15.860 the user doesn't care whether we're using a new 00:38:15.860 --> 00:38:18.019 way to fix it or an old way to fix it. They just 00:38:18.019 --> 00:38:20.739 care that it's fixed. So the advice of the entrepreneur 00:38:20.739 --> 00:38:24.280 is Before you decide that you want to build something 00:38:24.280 --> 00:38:30.659 with AI, talk to users and make sure that you're 00:38:30.659 --> 00:38:33.079 building something that they would want. And 00:38:33.079 --> 00:38:36.059 then you figure out how to build it. It makes 00:38:36.059 --> 00:38:39.539 a lot of sense. In terms of raising investment, 00:38:39.820 --> 00:38:44.559 because I know you're also in space, how hard 00:38:44.559 --> 00:38:47.099 do you think it is right now to raise investment 00:38:47.099 --> 00:38:49.659 overall? because here for example in UK a lot 00:38:49.659 --> 00:38:51.159 of people are complaining oh my god it's been 00:38:51.159 --> 00:38:53.760 so hard and no one's investing you know money's 00:38:53.760 --> 00:38:56.800 really tight we know for a fact that in the US 00:38:56.800 --> 00:38:59.420 you know rounds are at least five times bigger 00:38:59.420 --> 00:39:02.500 than in the UK which is great you know our minimum 00:39:02.500 --> 00:39:05.539 you know round is 250k and that's really a lot 00:39:05.539 --> 00:39:07.500 of money because in the US it's around a million 00:39:07.500 --> 00:39:10.860 so how is that working what's the current status 00:39:10.860 --> 00:39:14.039 okay so there's two questions there there's the 00:39:14.039 --> 00:39:17.739 US versus UK or US versus other countries And 00:39:17.739 --> 00:39:20.559 there's also how hard is it now versus how hard 00:39:20.559 --> 00:39:26.500 was it before. So for the second question, I 00:39:26.500 --> 00:39:29.619 think that valuations for private companies, 00:39:29.739 --> 00:39:32.019 for startups have generally come down in the 00:39:32.019 --> 00:39:35.519 US over the past like maybe six months, 12 months. 00:39:36.320 --> 00:39:42.559 And you can see that in a lot of investors that 00:39:42.559 --> 00:39:44.900 have sort of mark to market accounting. They 00:39:44.900 --> 00:39:47.739 have taken down the marks of their private investments. 00:39:48.239 --> 00:39:51.099 And it's also, you know, it's more difficult 00:39:51.099 --> 00:39:53.320 to raise money now than it was maybe six or 12 00:39:53.320 --> 00:39:56.780 months ago. Not necessarily a bad thing. It was 00:39:56.780 --> 00:40:01.820 just like such a mania going on. But it's still, 00:40:01.840 --> 00:40:03.699 you know, there's still a lot of VC investments. 00:40:03.820 --> 00:40:06.500 There's still a lot of VCs that have dry powder 00:40:06.500 --> 00:40:09.539 and there's still a lot of capital looking for 00:40:09.539 --> 00:40:14.199 AI opportunities. In terms of the US versus other 00:40:14.199 --> 00:40:17.239 countries, it's a really good question and I 00:40:17.239 --> 00:40:21.800 don't have a lot of expertise there, but there's 00:40:21.800 --> 00:40:24.860 something special about the Bay Area ecosystem 00:40:24.860 --> 00:40:30.099 and what they've created over the past 30, 40, 00:40:30.119 --> 00:40:34.119 50 years. I know that a lot of other countries 00:40:34.119 --> 00:40:36.380 have tried to replicate it and it's been elusive. 00:40:37.500 --> 00:40:39.599 It's not that nobody else has succeeded, but 00:40:39.630 --> 00:40:42.510 the scale is, you know, it's tough to replicate. 00:40:42.929 --> 00:40:45.010 I can't tell you that I know exactly what it 00:40:45.010 --> 00:40:48.489 is, but, you know, something about the physical 00:40:48.489 --> 00:40:51.869 proximity of a couple of universities and all 00:40:51.869 --> 00:40:55.230 of these capital providers, it's like a magnet, 00:40:55.429 --> 00:40:58.829 it's like a self -reinforcing system and it seems 00:40:58.829 --> 00:41:04.210 to work. I also think that, you know, there's 00:41:04.210 --> 00:41:07.679 something in... There's something about cultures 00:41:07.679 --> 00:41:12.780 that are just different. Something about the 00:41:12.780 --> 00:41:16.960 value of the individual versus the value of a 00:41:16.960 --> 00:41:21.260 group and the value of society. The US makes 00:41:21.260 --> 00:41:23.360 different choices than other countries do, and 00:41:23.360 --> 00:41:28.079 those choices have pluses and minuses. Maybe 00:41:28.079 --> 00:41:31.320 the entrepreneurial ecosystem is a plus, but 00:41:31.320 --> 00:41:34.380 hey, we have one of the most expensive healthcare 00:41:34.380 --> 00:41:37.920 systems in the world. to a lot of other people 00:41:37.920 --> 00:41:39.840 who live in a lot of other countries. It's crazy 00:41:39.840 --> 00:41:44.099 that you have to go shopping for dentists and 00:41:44.099 --> 00:41:48.219 you have to like haggle with your pharmacist 00:41:48.219 --> 00:41:51.099 about how much to pay for a medication. But that 00:41:51.099 --> 00:41:53.940 happens in the US all the time and it doesn't 00:41:53.940 --> 00:41:56.380 happen in any other advanced economy. So there's 00:41:56.380 --> 00:41:59.079 definitely trade -offs. That's always a big shock. 00:41:59.659 --> 00:42:01.599 Our second friend of mine who travels to the 00:42:01.599 --> 00:42:06.119 US. He also has family there. We're talking about 00:42:06.119 --> 00:42:08.400 Azomorpozol, which is the medicine you take through 00:42:08.400 --> 00:42:10.500 to the stomach protector. And he was saying I 00:42:10.500 --> 00:42:12.760 had to get some prescription there in the States 00:42:12.760 --> 00:42:16.360 and he was like $700. Whereas in here, I literally 00:42:16.360 --> 00:42:18.579 can go to my supermarket and buy it and it'll 00:42:18.579 --> 00:42:22.099 be two pounds. So very, very cheap. It's a bit 00:42:22.099 --> 00:42:26.239 of a shock. Another question I have is Would 00:42:26.239 --> 00:42:28.760 you advise someone coming in from Europe and 00:42:28.760 --> 00:42:30.880 say, you know what, just go to the US and try 00:42:30.880 --> 00:42:33.219 and raise some funds there? Or do you think automatically 00:42:33.219 --> 00:42:34.780 they'll be shut down and say, oh, where are you 00:42:34.780 --> 00:42:37.179 coming from? The UK, France, whatever, Germany. 00:42:37.699 --> 00:42:39.960 Yeah, we don't tend to invest in companies because 00:42:39.960 --> 00:42:42.800 it's illegal minefield or we don't trust you 00:42:42.800 --> 00:42:47.039 to do things the right way. You know, I don't 00:42:47.039 --> 00:42:49.380 think that switching countries is the right move 00:42:49.380 --> 00:42:55.539 for most entrepreneurs. I think that The biggest 00:42:55.539 --> 00:42:58.960 problems are around building something that people 00:42:58.960 --> 00:43:02.500 want. It's achieving product market fit. If you 00:43:02.500 --> 00:43:05.460 are able to achieve product market fit, the capital 00:43:05.460 --> 00:43:07.960 will come to you, even if it has to come overseas. 00:43:08.719 --> 00:43:14.920 So I think that over the past 20, 30 years, I 00:43:14.920 --> 00:43:17.719 think, you know, even if you look at US capital, 00:43:17.900 --> 00:43:20.619 that there's just a lot more investment overseas 00:43:20.619 --> 00:43:22.659 and, you know, that they're looking for deals 00:43:22.659 --> 00:43:25.889 in much, much bigger places. certainly broader 00:43:25.889 --> 00:43:29.190 than the Bay Area. And I think that concentration 00:43:29.190 --> 00:43:35.289 is decreasing, it's diffusing. I know that there's 00:43:35.289 --> 00:43:37.969 not as much capital available to entrepreneurs 00:43:37.969 --> 00:43:45.869 in more conservative countries. But I think that 00:43:45.869 --> 00:43:51.159 if you, let's say if you're in... If you're in 00:43:51.159 --> 00:43:53.320 France, if you're in Germany, if you're in the 00:43:53.320 --> 00:43:56.119 UK and you have developed a product that has 00:43:56.119 --> 00:43:59.199 good traction that users really want, I think 00:43:59.199 --> 00:44:01.000 that you can raise the capital without moving. 00:44:02.679 --> 00:44:05.980 You may not be able to raise it from local investors 00:44:05.980 --> 00:44:09.920 that are down the street, but I think that it 00:44:09.920 --> 00:44:13.119 is possible to do. This is going to be a very 00:44:13.119 --> 00:44:16.000 tricky question, I realize that. If you had just 00:44:16.000 --> 00:44:18.300 one piece of advice to give to an entrepreneur, 00:44:18.739 --> 00:44:23.719 what would that be? Build something people want 00:44:23.719 --> 00:44:30.360 Yeah, that's that's because I think Sometimes 00:44:30.360 --> 00:44:32.019 you meet a lot of people who are building this 00:44:32.019 --> 00:44:34.519 fantastic idea in their mind and then they realize 00:44:34.519 --> 00:44:36.519 actually no one wants this what I'm gonna do 00:44:36.519 --> 00:44:39.280 with it and the same happens quite often with 00:44:39.280 --> 00:44:41.719 software companies where they build this product 00:44:41.719 --> 00:44:43.380 that they think the market's really gonna need 00:44:43.380 --> 00:44:44.940 and that they didn't hear in the market and the 00:44:44.940 --> 00:44:46.320 market say why did you spend a couple million 00:44:46.320 --> 00:44:49.980 here because that's not for us that's No one's 00:44:49.980 --> 00:44:53.480 going to use it. As we're closing to the end 00:44:53.480 --> 00:44:55.639 of our podcast, if people want to reach out to 00:44:55.639 --> 00:44:58.940 you, how can they do so? Yeah. So the best way 00:44:58.940 --> 00:45:01.739 to reach me is on LinkedIn. It's Alan Bachman, 00:45:01.800 --> 00:45:06.340 A -L -O -N -B -O -C -H -M -E -N. And just drop 00:45:06.340 --> 00:45:11.019 me a note. I respond to everybody that reaches 00:45:11.019 --> 00:45:13.559 out to me. Thank you so much, Alan. Thank you 00:45:13.559 --> 00:45:14.860 so much for your time today. It's been super 00:45:14.860 --> 00:45:19.340 insightful. Thank you, Ricardo. Take care. Let 00:45:19.340 --> 00:45:19.739 me just.