#NewPodcast: What Lies Beneath?

WEBVTT 1 00:00:04.879 --> 00:00:09.429 Welcome back to the new podcast series. It's Kelsey corps with sweet fish media. 2 00:00:09.869 --> 00:00:13.150 Today I'm going to share with you a new podcast by our friends over 3 00:00:13.189 --> 00:00:18.589 at in Taros. It's called what lies beneath. This podcast is for anything 4 00:00:18.629 --> 00:00:25.100 and everything about risk. The host is Jennifer Bisegley and she interviews risk management 5 00:00:25.179 --> 00:00:30.300 professionals from every industry, learning their INS and outs, ups and downs, 6 00:00:30.339 --> 00:00:35.060 about protecting their companies and themselves from risk. If you think you'll find the 7 00:00:35.140 --> 00:00:39.729 show valuable after you check out this quick snippet, just search what lies beneath 8 00:00:39.810 --> 00:00:44.570 and apple or your favorite podcast player makes you subscribe and if you really like 9 00:00:44.729 --> 00:00:48.850 it, don't forget to leave a review. In this episode snippet she talks 10 00:00:48.929 --> 00:00:55.560 to nick by about the human and AI partnership. Let's tune in. Ai 11 00:00:55.719 --> 00:01:00.799 Is the most significant tack advancement that we've seen in the last fifteen years and 12 00:01:00.840 --> 00:01:03.280 even though it's been around for a long time, the increase in productivity we've 13 00:01:03.280 --> 00:01:08.310 seen with the advent of big data, of major internet platforms and available publicly, 14 00:01:08.870 --> 00:01:12.390 has been able to technology to take off. And, in addition, 15 00:01:12.709 --> 00:01:19.340 the technology is advancing in many different dimensions, not just some machine learning in 16 00:01:19.579 --> 00:01:25.420 computer vision and natural language processing, and given that the major fuels for AI, 17 00:01:26.219 --> 00:01:30.299 data processing power and algorithms, are all growing at a very significant rate, 18 00:01:30.379 --> 00:01:33.930 I think we're at the beginning of a very long AI boom. So 19 00:01:34.129 --> 00:01:38.969 I think this is the most significant technology that will likely reshape the economic landscape 20 00:01:40.209 --> 00:01:45.049 in the decade ahead. With regards to the Investment Strategy, I guess that's 21 00:01:45.090 --> 00:01:49.840 either. There are two major things that I'm particularly interested in. The first 22 00:01:49.040 --> 00:01:53.200 is the use of Ai to do something fun, mentally new and valuable, 23 00:01:53.560 --> 00:01:57.439 and I can be used for a lot of things and certainly as a hot 24 00:01:57.560 --> 00:02:01.750 area. It's overmarked at these days, but when AI is combined with significant 25 00:02:01.790 --> 00:02:07.590 and often unique data sets, it can create fundamentally new capabilities that weren't possible 26 00:02:08.030 --> 00:02:10.590 just, you know, three or four years ago, and sometimes not even 27 00:02:10.629 --> 00:02:16.620 conceivable four or five years ago, and those are the types of uses of 28 00:02:16.659 --> 00:02:21.939 Ai that I think the biggest companies will likely be created around. The Second 29 00:02:22.020 --> 00:02:29.300 Tiller is a distinctive data advantage. A lot of AI capabilities will be competized 30 00:02:29.340 --> 00:02:34.050 and offered by large platforms as free services in exchange for use of data. 31 00:02:34.729 --> 00:02:38.409 What's interesting to me is when a company has found a way to get access 32 00:02:38.449 --> 00:02:45.360 to privileged access, exclusive access to very significant data sets and really be able 33 00:02:45.400 --> 00:02:50.919 to get a meaningful head start in an area and build additional motes on top 34 00:02:51.000 --> 00:02:54.680 of that initial data set access. And sometimes that comes from additional customer data, 35 00:02:55.520 --> 00:03:00.469 sometimes it comes through large business development deals with other data providers. But 36 00:03:00.550 --> 00:03:05.030 if a company is able to do something fundamentally new and valuable and they have 37 00:03:05.189 --> 00:03:08.870 a meaningful data advantage and that leads to significant boats, I think is a 38 00:03:08.909 --> 00:03:14.379 chance to build a multi billion dollar company. I think they're all good points 39 00:03:14.580 --> 00:03:16.340 and you know, here at and tear us we see the same thing that 40 00:03:16.539 --> 00:03:21.860 technology and AI of the last ten years of just created for us. It's 41 00:03:21.900 --> 00:03:24.259 been a market. I do think it's a little bit interesting to think about 42 00:03:24.379 --> 00:03:30.129 things that aren't talked about enough when it comes to artificial intelligence, and two 43 00:03:30.210 --> 00:03:32.930 things that come to mind for me immediately. One may because we had lawyers 44 00:03:32.969 --> 00:03:38.250 in the office yesterday, are the business ethics side of ai or even the 45 00:03:38.330 --> 00:03:43.159 unconscious bias that are a lot of folks are talking about. So you know, 46 00:03:43.280 --> 00:03:46.000 what are your thoughts on those? Or do you see other areas that 47 00:03:46.080 --> 00:03:49.080 haven't really been addressed or that are on the horizon. We should be thinking 48 00:03:49.080 --> 00:03:53.919 about great question. I think there are a variety of subjects that aren't talked 49 00:03:53.960 --> 00:03:57.509 about enough. I think the one that I've put at the top of my 50 00:03:57.669 --> 00:04:02.710 list is that AI is very different from human intelligence. It's a natural initial 51 00:04:02.750 --> 00:04:09.270 reaction when a new technology comes out to understand and think creatively about the threats 52 00:04:09.310 --> 00:04:14.300 that it poses. And I think there is an immediate point that many people 53 00:04:14.460 --> 00:04:17.819 need in AI research labs and then the press sort of seized upon and the 54 00:04:17.899 --> 00:04:21.620 early days of our most recent ai boom, which is that this will lead 55 00:04:21.660 --> 00:04:28.449 to a kind of super human form of intelligence, Agi and its strongest form. 56 00:04:28.930 --> 00:04:30.930 I think what we're finding, and is you really big into the details 57 00:04:30.970 --> 00:04:36.329 of how AI is progressing, it's very different than human intelligence. Human intelligence 58 00:04:36.410 --> 00:04:43.839 isn't interesting evolutionary luge of a lot of different capabilities that have merged together to 59 00:04:44.000 --> 00:04:50.000 help us optimize our survival and reproduction, and to think that computer intelligence will 60 00:04:50.040 --> 00:04:55.189 follow the same general path or that is representative of the same types of general 61 00:04:55.230 --> 00:05:00.509 understandings, I think is is an interesting leap. We may both be subsets 62 00:05:00.550 --> 00:05:04.350 of a rawder whole, but a lot remains to be proven. And specifically 63 00:05:04.589 --> 00:05:10.699 what a as the very good at to date is solving very narrow pattern recognition 64 00:05:10.740 --> 00:05:17.620 problems extremely well, so image recognition, language recognition or paradigmatic use cases. 65 00:05:18.339 --> 00:05:21.730 But when it comes to the field of human judgment, so far, and 66 00:05:21.850 --> 00:05:26.490 we're so early in this revolution, so there's a long way ways to go, 67 00:05:27.529 --> 00:05:30.569 the results so far have been pretty disappointing. Ai has not been able 68 00:05:30.730 --> 00:05:38.279 to combine different understandings of context and be able to form a kind of reasoning 69 00:05:38.439 --> 00:05:42.360 that we would see as legitimate for complex human decisions. So I think there's 70 00:05:42.360 --> 00:05:46.959 a lot to be done and and the immediate anthropomorphism of saying Oh ai is 71 00:05:47.000 --> 00:05:50.110 going to look a lot like us, I think is giving rise to a 72 00:05:50.269 --> 00:05:55.790 more complex reality. And many people have said, I think in my view, 73 00:05:55.870 --> 00:06:00.790 rightly, that AI has made its biggest jumps by mimicking the human mind 74 00:06:00.829 --> 00:06:03.779 and it may need to do more of that in particular and building in more 75 00:06:03.899 --> 00:06:10.579 innate structure, sort of prewiring to do certain things very well like we do. 76 00:06:10.699 --> 00:06:15.180 We have a predisposition to language that's extraordinary and it's made us very effective. 77 00:06:15.259 --> 00:06:17.689 There are many other examples as well. So I guess at the top 78 00:06:17.769 --> 00:06:21.930 of my list of things that are not discussed enough. I would I would 79 00:06:21.930 --> 00:06:27.730 rank that very high. On ethics, I think it's a critical area for 80 00:06:27.850 --> 00:06:31.370 AI to be broadly used and broadly trusted, and particularly when you get to 81 00:06:31.449 --> 00:06:36.519 the level of making human like judgments, for people will ask, why should 82 00:06:36.519 --> 00:06:41.399 I trust that? I know who who said so and why can you explain 83 00:06:41.439 --> 00:06:45.519 this to me? We need to be able to trust machines with thing with 84 00:06:46.000 --> 00:06:49.110 the type of logic that would count in a human explanation, and we're we 85 00:06:49.269 --> 00:06:53.910 have a ways to go in that. I think there is a part of 86 00:06:53.990 --> 00:06:59.509 that problem is a bias of data sets. You know one, one fascinating 87 00:06:59.670 --> 00:07:02.540 bias of data sets and you see this in science, he's you see in 88 00:07:02.660 --> 00:07:06.060 public safety, you see it in a lot of other fields. Relates to 89 00:07:06.259 --> 00:07:12.339 gender by my wife is a genomics entrepreneur in the field of women's health and 90 00:07:12.500 --> 00:07:18.009 she's really taught me about this and there's some powerful example things like the development 91 00:07:18.050 --> 00:07:26.170 of seat belts. Seat belts were made by men and optimize for men and 92 00:07:26.290 --> 00:07:32.399 using data relating to men and as a result, seat belts historically works very 93 00:07:32.439 --> 00:07:35.360 well for men but worked much less well for women. And there are a 94 00:07:35.399 --> 00:07:39.879 lot more fatilities and car accidents for women, just because of the way that 95 00:07:40.040 --> 00:07:43.839 sort of the data that they were built upon. In a lot of fields 96 00:07:43.920 --> 00:07:46.870 and health, the initial, you know, discoveries and the choice of what 97 00:07:46.990 --> 00:07:53.149 subjects to pursue were made by men, and the theories that followed them the 98 00:07:53.310 --> 00:07:57.230 men. The you know, products that came out of them were developed by 99 00:07:57.269 --> 00:07:59.459 men, and so, as you can imagine, there was very little focus 100 00:07:59.540 --> 00:08:05.980 on women at all and I think an understanding of the significance of data. 101 00:08:07.100 --> 00:08:11.139 That is it need not be biased in the narrow sense. It could do 102 00:08:11.379 --> 00:08:16.610 just be not representative broadly, but understanding that when you're developing models, I 103 00:08:16.689 --> 00:08:22.810 think is critical. Such an interesting topic by Nick over at Van Rock and 104 00:08:22.970 --> 00:08:26.970 wow. Jennifer is going to be a phenomenal host to this podcast again. 105 00:08:26.089 --> 00:08:31.720 To find this show, just search what lies beneath an apple podcast or wherever 106 00:08:31.879 --> 00:08:35.080 you do. You're listening, subscribe and leave a review if you like it. 107 00:08:35.600 --> 00:08:37.519 Maybe this podcast isn't for you, but you know someone who would like 108 00:08:37.639 --> 00:08:41.870 it, well, don't forget to tell them so you in next week for 109 00:08:41.950 --> 00:08:48.190 an extra special episode. Until then, I hate it when podcasts incessantly ask 110 00:08:48.269 --> 00:08:52.710 their listeners for reviews, but I get why they do it, because reviews 111 00:08:52.750 --> 00:08:56.539 are enormously helpful when you're trying to grow a podcast audience. So here's what 112 00:08:56.620 --> 00:08:58.460 we decided to do. If you leave a review for me to be growth 113 00:08:58.539 --> 00:09:03.019 in apple podcasts and email me a screenshot of the review to James at Sweet 114 00:09:03.019 --> 00:09:07.779 Fish Mediacom, I'll send you a signed copy of my new book, content 115 00:09:07.860 --> 00:09:11.529 based networking, how to instantly connect with anyone you want to know. We 116 00:09:11.649 --> 00:09:13.330 get a review, you get a free book. We both win.