WEBVTT - Tech News: Artists Versus Artificial Intelligence 0:00:04.400 --> 0:00:07.800 Welcome to tech Stuff, a production from I Heart Radio. 0:00:11.880 --> 0:00:14.600 Hey there, and welcome to tech Stuff. I'm your host, 0:00:14.760 --> 0:00:18.360 Jonathan Strickland. I'm an executive producer with I Heart Radio 0:00:18.440 --> 0:00:21.520 and how the tech area. It's time for the tech 0:00:21.560 --> 0:00:27.560 news for Tuesday, January twenty twenty three, and let's start 0:00:27.640 --> 0:00:31.120 off with cryptocurrency. Now, I'm sure most of y'all know 0:00:31.440 --> 0:00:34.919 I'm pretty skeptical of crypto in general, and they've got 0:00:34.960 --> 0:00:38.720 a lot of bones to pick with cryptocurrency. But I 0:00:38.760 --> 0:00:40.920 also think it's only fair to report on when the 0:00:40.960 --> 0:00:45.479 market gets some momentum in a positive direction. So Reuter's 0:00:45.479 --> 0:00:48.720 reports that Bitcoin is up twenty six percent since the 0:00:48.800 --> 0:00:51.920 beginning of this year and has climbed above the twenty 0:00:52.000 --> 0:00:55.800 thousand dollars per coin mark for the first time in months. 0:00:56.240 --> 0:00:59.920 Other crypto coins are also performing well, many of them 0:01:00.120 --> 0:01:05.280 kind of you know, following bitcoins footsteps, and some so 0:01:05.400 --> 0:01:10.360 called meme coins, which others would argue our trash coins, 0:01:10.840 --> 0:01:15.360 are going absolutely bonkers. You're seeing like five thousand percent 0:01:15.560 --> 0:01:18.400 increase in value for some of these. Now I should 0:01:18.600 --> 0:01:22.240 add that these meme coins are usually worth like fractions 0:01:22.480 --> 0:01:25.880 of a penny, so even a small amount of growth 0:01:26.360 --> 0:01:29.400 ends up looking huge if you talk about percentage, right, 0:01:30.240 --> 0:01:34.639 If it went from point zero zero zero zero zero 0:01:34.760 --> 0:01:39.960 zero eight to point zero zero zero zero zero one zero, 0:01:40.360 --> 0:01:44.240 that's a huge jump in percentage, but in actual money 0:01:44.240 --> 0:01:48.600 it's not much at all. Right. Also, Reuters rightfully points 0:01:48.600 --> 0:01:51.720 out that these meme coins tend to be even more 0:01:51.880 --> 0:01:55.640 unstable than other types of crypto, and the value can 0:01:55.720 --> 0:01:59.040 drop just as quickly as it can go up. See 0:01:59.080 --> 0:02:02.400 also coin, which was all the rage for about a 0:02:02.400 --> 0:02:04.600 month a couple of years ago before the bottom fell out. 0:02:05.120 --> 0:02:09.399 Reuter's also cautions that this upturn may not sustain itself, 0:02:09.840 --> 0:02:13.920 that the recent gains in cryptocurrency are probably connected to 0:02:14.000 --> 0:02:17.720 an overall positive look at the global economy and people 0:02:17.800 --> 0:02:21.440 making a bet that inflation is at least for the moment, 0:02:21.600 --> 0:02:26.680 kind of done. Obviously, any big change could send values 0:02:26.760 --> 0:02:31.279 down again, and as we all learned with Russia's invasion 0:02:31.400 --> 0:02:37.799 of Ukraine, sometimes it's just an unpredictable global event that 0:02:37.880 --> 0:02:40.880 can really have a huge impact on the economy. So 0:02:41.840 --> 0:02:45.360 nothing is ever certain. The changes are not always predictable. 0:02:45.400 --> 0:02:48.440 So I guess what I'm saying is crypto appears to 0:02:48.480 --> 0:02:51.560 be recovering and I think I said last year that 0:02:51.639 --> 0:02:55.160 I didn't believe that the setbacks we saw in late 0:02:55.240 --> 0:02:59.000 two were the death knell for crypto in general. It 0:02:59.120 --> 0:03:02.840 was just a real reckoning for crypto and blockchain and 0:03:02.919 --> 0:03:05.800 related topics. So I suppose we're gonna have to keep 0:03:05.800 --> 0:03:08.120 an eye on stuff like how governments are going to 0:03:08.240 --> 0:03:12.160 treat the crypto exchange Binance, for example, to see where 0:03:12.280 --> 0:03:16.840 crypto goes in because it's not out of the woods 0:03:16.960 --> 0:03:19.880 right There are plenty of governments around the world that 0:03:19.960 --> 0:03:23.880 are taking a more critical view of cryptocurrency and looking 0:03:23.880 --> 0:03:28.640 at imposing regulations. So while we're seeing some improvement right now, 0:03:28.960 --> 0:03:32.520 that could be just a short term gain. Uh it's 0:03:32.600 --> 0:03:35.880 too early to say, but I don't know. Maybe we'll 0:03:35.920 --> 0:03:37.880 see something kind of akin to what we saw in 0:03:39.240 --> 0:03:44.160 where the value really went up dramatically in the early 0:03:44.240 --> 0:03:48.520 part of the year. Now we've got several stories about AI, 0:03:49.600 --> 0:03:51.320 and when I said I thought AI was going to 0:03:51.400 --> 0:03:54.320 be a big topic in three I didn't necessarily mean 0:03:54.360 --> 0:03:59.440 it would dominate headlines in mid January. Uh. I suspect 0:03:59.440 --> 0:04:03.120 we're gonna see AI coverage calmed down a bit later 0:04:03.160 --> 0:04:05.360 this year. Like I think it's always going to be 0:04:05.440 --> 0:04:08.560 a big part of the news in three, but I 0:04:08.560 --> 0:04:11.440 don't think it's going to be quite as dominant once 0:04:11.960 --> 0:04:14.880 things kind of initially calmed down. Right at the moment, 0:04:14.920 --> 0:04:17.800 it's sort of the scary new thing that has media 0:04:17.800 --> 0:04:20.640 outlets jumping to cover, And I don't really think it's 0:04:20.760 --> 0:04:23.760 it's necessarily that scary. It's just that's how it's being treated. 0:04:24.720 --> 0:04:30.119 But first up, getting Images, which maintains a huge library 0:04:30.160 --> 0:04:33.919 of stock photography and artwork that people can license to 0:04:34.080 --> 0:04:38.520 use for their works, is suing the company Stability AI. 0:04:38.720 --> 0:04:44.080 Getting images accuses Stability AI of scraping the library of 0:04:44.120 --> 0:04:47.880 images in getting images, just in an effort to train 0:04:48.600 --> 0:04:52.799 an art tool. An AI art tool called stable diffusion. 0:04:53.200 --> 0:04:58.880 So quick explanation here. One discipline within artificial intelligence is 0:04:59.240 --> 0:05:01.800 machine learn arning, or you might even argue that it's 0:05:02.160 --> 0:05:05.880 adjacent artificial intelligence. The ven diagram has a lot of 0:05:05.920 --> 0:05:11.000 overlap and machine intelligence, just like AI, can take lots 0:05:11.160 --> 0:05:15.360 of different forms. But one common practice in machine learning 0:05:15.880 --> 0:05:19.279 is to teach a computer model how to do whatever 0:05:19.320 --> 0:05:21.839 it is the model that is supposed to do by 0:05:21.960 --> 0:05:26.040 feeding the model lots and lots of information. For example, 0:05:26.600 --> 0:05:28.840 let's say you wanted to teach an AI system to 0:05:28.960 --> 0:05:32.559 recognize images of cats, well, you would want to feed 0:05:32.600 --> 0:05:35.440 the model tens of thousands of pictures of kiddie cats. 0:05:35.440 --> 0:05:38.680 But you would also want to feed lots of pictures 0:05:38.720 --> 0:05:41.360 that have no cats in them at all to the 0:05:41.400 --> 0:05:44.000 model so that it could tell the difference, and you 0:05:44.040 --> 0:05:46.880 would continuously tweak the model so that it would get 0:05:46.920 --> 0:05:50.920 better and better at distinguishing which images actually have cats 0:05:50.960 --> 0:05:54.240 in them. Well, if you want an AI image generation 0:05:54.360 --> 0:05:59.440 program that makes images that are actually, you know, recognizable 0:05:59.480 --> 0:06:02.359 as the of stuff you prompted the a A model 0:06:02.400 --> 0:06:06.400 to create, you have to feed it lots and lots 0:06:06.400 --> 0:06:09.919 of images, right, otherwise you could just get these random 0:06:10.040 --> 0:06:13.760 shapes and colors that don't look like anything, or that 0:06:13.800 --> 0:06:16.760 look like really disturbing versions of whatever it was you 0:06:16.800 --> 0:06:20.640 wanted to create. So Getty Images has a truly enormous 0:06:20.640 --> 0:06:25.080 supply of photographs. They're stretched across decades. Some of them 0:06:25.240 --> 0:06:29.560 are work for higher images, where photographers have set up, 0:06:29.720 --> 0:06:32.080 you know, like scenes in a studio and they just 0:06:32.520 --> 0:06:35.640 shoot tons and tons and tons of photographs and then 0:06:35.760 --> 0:06:40.320 sell those photos to Getty Images. You get some pretty 0:06:40.320 --> 0:06:43.400 wacky versions of that. If you've ever seen pictures of 0:06:43.440 --> 0:06:46.279 a woman wearing a silvery body suit and a metallic 0:06:46.400 --> 0:06:49.000 visor while she's holding an ear of corn. You know 0:06:49.040 --> 0:06:51.640 what I'm talking about. And yes, that's a real stock image. 0:06:52.160 --> 0:06:54.280 Most of you, I'm sure have seen it somewhere at 0:06:54.320 --> 0:06:59.279 some point. Other images come from like paparazzi who take 0:06:59.360 --> 0:07:02.680 pictures of clebrities and notable figures and then they license 0:07:02.760 --> 0:07:05.320 or sell their photographs to Getty. And what Getty is 0:07:05.320 --> 0:07:10.960 saying is that Stability AI used software to scrape Getty's 0:07:11.000 --> 0:07:14.840 images and copy them without paying for them, without paying 0:07:14.840 --> 0:07:17.720 a license for it, and using those images then to 0:07:17.760 --> 0:07:24.320 train a I that will ultimately you be used to 0:07:24.440 --> 0:07:29.280 create a competing product, right, a competing service, Like instead 0:07:29.320 --> 0:07:34.400 of going to Getty Images to license a stock photograph 0:07:34.720 --> 0:07:37.960 that represents whatever it is you want, you would go 0:07:38.160 --> 0:07:43.400 to Stable Diffusion and have it generate an image based 0:07:43.400 --> 0:07:46.440 on what you want. But Getty says, well, first of all, 0:07:46.480 --> 0:07:49.960 you're using our images to train your computer to do this, 0:07:50.480 --> 0:07:53.320 then you're going to introduce something that competes against our 0:07:53.360 --> 0:07:57.800 own business. That's an unfair business practice. Now we're starting 0:07:57.840 --> 0:07:59.960 to get into an area where the law is really 0:08:00.160 --> 0:08:02.760 lagging behind. The law is not designed to deal with 0:08:02.800 --> 0:08:08.400 this kind of intellectual property issue, so it does sound 0:08:08.440 --> 0:08:10.520 like this method is creeping in on the violation of 0:08:10.560 --> 0:08:14.600 intellectual property laws, or at least tiptoeing around them. But 0:08:14.680 --> 0:08:18.680 without actual law or court precedents. With AI, it's a 0:08:18.680 --> 0:08:21.600 gray area at best. But my guess is this year 0:08:21.680 --> 0:08:23.120 is going to be one where we start to see 0:08:23.120 --> 0:08:26.920 that gray turn into more black and white sooner rather 0:08:27.000 --> 0:08:31.960 than later, because it is becoming oppressing and uh and 0:08:32.400 --> 0:08:36.560 relevant issue. On a similar note, a group of artists 0:08:36.559 --> 0:08:40.959 have joined to file a class action lawsuit against three companies, 0:08:41.160 --> 0:08:44.880 dev and Art, mid Journey, and our buddy Stability AI 0:08:45.000 --> 0:08:48.760 that we just talked about like Getty. These artists argue 0:08:48.840 --> 0:08:52.439 that the companies have made illegal use of copyrighted works 0:08:52.480 --> 0:08:56.600 in order to train their respective AI models to generate images. 0:08:57.040 --> 0:08:59.960 Ours Technica has a great article about this, and they 0:09:00.120 --> 0:09:04.560 site and AI analyst named Alex Champion Dard, who has 0:09:04.600 --> 0:09:09.680 pointed out some potential problems with the lawsuit. For one, 0:09:09.920 --> 0:09:13.960 the lawsuit states that quote, when used to produce images 0:09:14.080 --> 0:09:18.160 from prompts by its users, stable diffusion uses the training 0:09:18.200 --> 0:09:23.720 images to produce seemingly new images through a mathematical software process. 0:09:23.760 --> 0:09:27.120 These quote unquote new images are based entirely on the 0:09:27.160 --> 0:09:31.680 training images and are derivative works of the particular images 0:09:31.800 --> 0:09:36.599 stable diffusion draws from when assembling a given output. Ultimately, 0:09:36.679 --> 0:09:41.280 it is merely a complex collage tool end quote. So 0:09:42.280 --> 0:09:45.760 they're they're really saying that there's nothing original in any 0:09:46.160 --> 0:09:50.040 image that stable diffusion produces, that everything has come from 0:09:50.120 --> 0:09:54.280 one element of one of the images in its massive 0:09:55.080 --> 0:10:00.120 mound of training images. That's not really how genera to 0:10:00.320 --> 0:10:03.760 AI works. The language says, the AI is just taking 0:10:03.800 --> 0:10:06.520 this massive amount of content as a starting point and 0:10:06.520 --> 0:10:09.440 then using that to create a new image, almost like, well, 0:10:09.520 --> 0:10:11.920 let me take the nose from this one, and the 0:10:12.000 --> 0:10:15.880 eyes from this one, and this smile and that hairstyle, 0:10:15.920 --> 0:10:18.040 and then putting them all together to make the quote 0:10:18.120 --> 0:10:21.920 unquote new image. But that's not actually how these models work. 0:10:22.280 --> 0:10:25.959 It's an oversimplification. It's more akin to how a human 0:10:26.080 --> 0:10:29.679 artist would study works made by other people before they 0:10:29.679 --> 0:10:33.240 start producing their own work. They would not necessarily be 0:10:33.360 --> 0:10:35.880 copying someone else's work directly. I mean, that could be 0:10:35.920 --> 0:10:39.000 an exercise to see if you can master the same 0:10:39.040 --> 0:10:42.040 techniques as some other artist. But if you're making your 0:10:42.040 --> 0:10:46.520 own work. You're not trying to copy someone else's work. Instead, 0:10:46.520 --> 0:10:49.720 you're using other people's work as sort of an inspirational 0:10:49.800 --> 0:10:53.360 launching ground on how to proceed. That could include things 0:10:53.400 --> 0:10:59.000 like a color palette, the brush technique, the perspective, all 0:10:59.040 --> 0:11:01.760 these sorts of things. So that's a little bit closer 0:11:01.800 --> 0:11:04.960 to what these AI models are doing. And Champion Dard 0:11:05.000 --> 0:11:07.679 points out that the lawsuit could lead to nothing if 0:11:07.679 --> 0:11:10.520 the defense is able to argue that what they're being 0:11:10.520 --> 0:11:14.000 accused of doing is just not true. But then, as 0:11:14.000 --> 0:11:17.199 I mentioned earlier, the real issue is that there isn't 0:11:17.280 --> 0:11:20.640 law that outlines the parameters to which generative AI can 0:11:20.679 --> 0:11:25.320 rely upon existing copyrighted works as training material. I mean, presumably, 0:11:25.720 --> 0:11:28.800 if we're looking at the written word, most writers have 0:11:29.000 --> 0:11:32.520 read a lot before they start writing. So as long 0:11:32.559 --> 0:11:35.760 as a writer is not plagiarizing some other writer, would 0:11:35.760 --> 0:11:39.199 you argue that a writer owes money to all the 0:11:39.280 --> 0:11:42.040 authors that they themselves have read before they produce their 0:11:42.040 --> 0:11:48.320 own work, Because surely by reading that influences your own style, 0:11:48.800 --> 0:11:53.440 it influences your own sensibilities. So if we if we 0:11:53.520 --> 0:11:55.800 take this to an extreme, you would say, ah, you 0:11:55.880 --> 0:11:59.040 owe money to anyone who has influenced you in the 0:11:59.040 --> 0:12:04.280 production of your own work. That's clearly not realistic or uh, 0:12:04.480 --> 0:12:08.040 you know, reasonable. But when it comes to AI, it 0:12:08.080 --> 0:12:12.200 gets a lot more tricky because you're literally using these 0:12:12.280 --> 0:12:16.400 materials to train up the AI to make its own So, yeah, 0:12:16.559 --> 0:12:18.520 we're gonna have to keep a close eye to see 0:12:18.559 --> 0:12:21.679 how this develops throughout this year. We're gonna take a 0:12:21.760 --> 0:12:24.040 quick break. When we come back, we'll talk about some 0:12:24.080 --> 0:12:36.560 more tech news. We're back, and we're not done with 0:12:36.720 --> 0:12:40.760 AI yet. Several news outlets have reported that the chat 0:12:40.800 --> 0:12:46.600 GPT three point five system successfully passed sections of the 0:12:46.679 --> 0:12:50.160 US Bar Exam. That's the exam that lawyers have to 0:12:50.200 --> 0:12:53.240 pass here in the United States before they can law 0:12:53.360 --> 0:12:56.720 your year or something. I understand that the Bar exam 0:12:56.840 --> 0:12:59.880 is very challenging and that for lots of lawyers in 0:13:00.000 --> 0:13:03.880 may take more than one attempt to pass the Bar exam. Now, 0:13:03.920 --> 0:13:07.760 to be clear, chat GPT did not pass the full exam, 0:13:08.160 --> 0:13:10.960 so it it it didn't pass it, and it is 0:13:11.000 --> 0:13:15.720 certainly not currently allowed to practice law. Rather, it passed 0:13:15.840 --> 0:13:19.720 sections on evidence and torts, and I was disappointed to 0:13:19.720 --> 0:13:23.079 find out torts are not pastries filled with fruit m 0:13:24.440 --> 0:13:28.679 raspberry tort The program fell short when it came to 0:13:28.760 --> 0:13:32.600 taking the full exam. It scored fifty point three on 0:13:32.640 --> 0:13:36.439 a test that sets a minimum passing grade of sixty 0:13:36.200 --> 0:13:38.959 eight percent. And I don't think anyone is expecting the 0:13:39.040 --> 0:13:43.200 US legal system to recognize AI lawyers in the near future, 0:13:43.880 --> 0:13:46.560 but some of the field have already expressed concern about this, 0:13:46.760 --> 0:13:49.720 and then others have said these tools could be really 0:13:49.800 --> 0:13:53.520 useful as a way to augment human lawyers as they 0:13:53.520 --> 0:13:56.400 go about their jobs. For example, this kind of a 0:13:56.480 --> 0:14:00.280 chatbot might be able to generate a first draft aft 0:14:00.640 --> 0:14:04.800 of deposition questions, but a lawyer would then subsequently go 0:14:05.040 --> 0:14:07.800 through this draft and then refine it. So in other words, 0:14:08.120 --> 0:14:11.520 the chat bot could do what we hear advocates say 0:14:11.640 --> 0:14:14.400 AI is good for all the time. It can make 0:14:14.520 --> 0:14:18.320 jobs easier and more efficient, but it doesn't actually replace 0:14:18.840 --> 0:14:22.080 the human behind the job. Of course, for it to 0:14:22.120 --> 0:14:23.640 do that, it has to work well enough to be 0:14:23.720 --> 0:14:27.480 useful rather than distracting or counter productive. And as we've 0:14:27.520 --> 0:14:32.400 seen with chat GPT, it's it's really impressive, but it's 0:14:32.440 --> 0:14:37.000 not there yet, and AI powered chat bots are not 0:14:37.160 --> 0:14:41.000 likely to disappear either. In fact, I'm sure we're gonna 0:14:41.040 --> 0:14:44.280 hear a lot more about them. Microsoft just announced that 0:14:44.360 --> 0:14:48.520 it is increasing access to chat GPT, making it generally 0:14:48.560 --> 0:14:51.880 available to its customers. In an earlier news episode, I 0:14:51.920 --> 0:14:56.480 mentioned how Microsoft is apparently planning a massive investment in 0:14:56.560 --> 0:15:01.240 open Ai that's the startup behind chat g pt, and 0:15:01.440 --> 0:15:04.440 effectively it would mean Microsoft would end up buying just 0:15:04.680 --> 0:15:10.760 a scoch under ownership of the startup. Well, Microsoft has 0:15:10.800 --> 0:15:14.280 not yet confirmed those reports. It has indicated that it 0:15:14.280 --> 0:15:17.280 will let more Microsoft customers make use of chat GPT 0:15:17.440 --> 0:15:21.520 tool via Microsoft Cloud Services. What this means in the 0:15:21.560 --> 0:15:25.040 nearer and long term, I can't really say. I sure 0:15:25.080 --> 0:15:28.960 do hope we'll see, uh that we do not get 0:15:29.000 --> 0:15:32.720 an era of AI generated news and entertainment, because that 0:15:32.760 --> 0:15:34.760 would be kind of the opposite of what we really 0:15:34.760 --> 0:15:37.440 want AI to do, Like we wanted to augment but 0:15:37.520 --> 0:15:40.680 not replace. So I don't want to see it doing 0:15:40.720 --> 0:15:43.400 stuff like what we saw was c net, where we 0:15:43.520 --> 0:15:46.760 just assigned AI these writing jobs and they churn out 0:15:47.560 --> 0:15:51.720 low value articles. Um, I don't really want to see that. 0:15:51.800 --> 0:15:54.080 I mean, obviously, what I want to see is high 0:15:54.120 --> 0:15:59.400 value articles created by talented human writers. But yeah, it 0:15:59.440 --> 0:16:04.120 would be pretty darn hard to be a professional creative 0:16:04.680 --> 0:16:08.280 in a world where we accept the level of creative 0:16:08.280 --> 0:16:14.240 output generated by a I as being desirable. Um, it's 0:16:14.320 --> 0:16:17.600 very hard already to make your living as a creative 0:16:18.080 --> 0:16:20.120 with all the competition that's out there. It would be 0:16:20.160 --> 0:16:23.840 even more difficult if we say, yeah, this AI stuff, 0:16:23.840 --> 0:16:26.320 it's not great, but it gets the job done, so 0:16:26.400 --> 0:16:28.680 let's just go with it. That would make it even harder. 0:16:29.440 --> 0:16:32.240 Musician Nick Cave knows what I'm talking about. A fan 0:16:32.400 --> 0:16:36.280 sent him an AI generated song that mimicked the musician's style. 0:16:36.360 --> 0:16:39.200 And for those of you not familiar with Nick Cave 0:16:39.720 --> 0:16:42.400 of Nick Cave and the Bad Seeds fame, he wrote 0:16:42.440 --> 0:16:45.400 the song Red Right Hand, which is featured heavily in 0:16:45.440 --> 0:16:48.080 the Scream franchise. It's also in a lot of other media. 0:16:48.720 --> 0:16:53.560 He's known for dark, gothic and often melodramatic ballads and 0:16:53.600 --> 0:16:57.080 other style songs. And I recommend listening to him. Depending 0:16:57.120 --> 0:16:59.800 upon your mood and the song you've picked, you'll either 0:17:00.040 --> 0:17:06.840 find him really intriguing or very very silly, or perhaps both. Anyway, 0:17:06.920 --> 0:17:10.240 Nick Cave very much did not like the AI generated piece. 0:17:10.280 --> 0:17:14.120 He said, quote songs arise out of suffering, by which 0:17:14.160 --> 0:17:17.840 I mean they are predicated upon the complex internal human 0:17:17.920 --> 0:17:21.320 struggle of creation. And well, as far as I know, 0:17:21.880 --> 0:17:26.480 algorithms don't feel end quote. He also said the song 0:17:26.600 --> 0:17:29.680 was quote a grotesque mockery of what it is to 0:17:29.720 --> 0:17:33.199 be human, and well, I don't much like it end quote. 0:17:33.600 --> 0:17:36.200 And you know, I find it difficult to disagree with him. 0:17:36.240 --> 0:17:39.760 I think it's pretty true. The song had lyrics like quote, 0:17:40.040 --> 0:17:42.360 in the depths of the night, I hear a call, 0:17:42.680 --> 0:17:46.040 a voice that echoes through the hall. It's a sirens 0:17:46.080 --> 0:17:48.879 song that pulls me in, takes me to a place 0:17:49.040 --> 0:17:53.040 where I can't begin end quote. It's not exactly meaningful. 0:17:53.520 --> 0:17:56.399 I think I mentioned in an earlier news episode that 0:17:56.560 --> 0:17:59.600 actually tried to have chat GPT create a high kup 0:18:00.040 --> 0:18:04.120 and the program produced something that's superficially looked like a hiku, 0:18:04.560 --> 0:18:08.359 but it lacked any poetic value, and further, it didn't 0:18:08.400 --> 0:18:12.960 adhere to the structure of a hiku poem. And generally 0:18:13.000 --> 0:18:15.880 that's what I found when I've used chat GPT. Honestly, 0:18:15.920 --> 0:18:18.200 I find that a little surprising, Like I would think 0:18:18.240 --> 0:18:21.680 the rules part would be something that a program would 0:18:21.680 --> 0:18:25.960 be better at handling. Clearly that's not what chat GPT 0:18:26.119 --> 0:18:28.399 was intended for, So I mean, I gotta cut it 0:18:28.480 --> 0:18:31.080 some slack in that sense. But when it comes to 0:18:31.640 --> 0:18:34.880 two works of art that have specific form and structure 0:18:34.920 --> 0:18:37.800 that you're supposed to follow, I would think computer programs 0:18:37.840 --> 0:18:41.120 would be better at doing that. It wouldn't necessarily mean 0:18:41.160 --> 0:18:43.960 the stuff created would be any good, but it would 0:18:43.960 --> 0:18:47.600 at least adhere to the rules of the art form. So, 0:18:47.640 --> 0:18:51.479 for example, sonnets, that it would adhere to the number 0:18:51.680 --> 0:18:56.439 of of versus and the uh the rhythm of them 0:18:56.560 --> 0:19:01.600 and the rhyme scheme. But it doesn't. So oh yeah, 0:19:01.640 --> 0:19:03.679 it's kind of surprising to me on that. But except 0:19:03.680 --> 0:19:05.760 for the fact that I do know, chat gpt wasn't 0:19:05.800 --> 0:19:11.320 exactly designed to be a sonnet producing machine. Uh but yeah, like, 0:19:11.720 --> 0:19:15.320 we're not at the point where these programs are capable 0:19:15.520 --> 0:19:21.399 of generating artistic content that has real merit to It 0:19:21.400 --> 0:19:24.000 doesn't mean we won't get there, and doesn't mean that, 0:19:24.119 --> 0:19:28.600 you know, it can't at least mimic the lower lower 0:19:28.680 --> 0:19:32.520 forms of pop art. And I don't mean to dismiss 0:19:32.600 --> 0:19:34.760 pop art because I love a lot of pop art, 0:19:35.480 --> 0:19:40.440 but uh, it's hard to argue that certain popular songs 0:19:40.520 --> 0:19:47.880 aren't trivial and simple when they very much are. But yeah, 0:19:47.920 --> 0:19:50.680 I'm I'm with Nick Cave on this one. Okay, moving 0:19:50.720 --> 0:19:54.119 away from AI, Meta is in the news again, and 0:19:54.160 --> 0:19:58.080 for once, Meta is the one doing the suing. So 0:19:58.119 --> 0:20:01.320 in this case, Meta is suing a company called Voyager Labs, 0:20:01.840 --> 0:20:05.240 and Meta says that Voyager Labs created thirty eight thousand 0:20:05.320 --> 0:20:09.280 fake accounts as part of an effort to scrape Facebook 0:20:09.359 --> 0:20:13.520 and Instagram and other social networks of user data. So 0:20:13.760 --> 0:20:16.680 Meta says that the effort pulled data from more than 0:20:16.760 --> 0:20:20.679 half a million Facebook pages and Facebook groups, including stuff 0:20:20.720 --> 0:20:25.400 like the posts people were putting onto Facebook, their photos, 0:20:25.520 --> 0:20:29.240 their friends lists, and any other information that was set 0:20:29.280 --> 0:20:33.480 to be publicly viewable. And well, people who may have 0:20:33.520 --> 0:20:36.280 set their profiles so that only their friends can view it, 0:20:36.600 --> 0:20:39.000 you have to remember they could still end up popping 0:20:39.119 --> 0:20:42.920 up in other pages. So even if you are someone 0:20:42.960 --> 0:20:45.480 who isn't on Facebook at all, some of your information 0:20:45.680 --> 0:20:48.399 maybe up there just because some friends of yours shared 0:20:48.480 --> 0:20:52.160 some stuff. So Voyager Labs is a company that's based 0:20:52.200 --> 0:20:55.640 in Israel, and it describes itself as an AI powered 0:20:55.760 --> 0:21:01.840 investigation company, and it mostly works with agencies like law 0:21:01.920 --> 0:21:07.240 enforcement or military organizations, and one of those is apparently 0:21:07.280 --> 0:21:09.880 the Los Angeles Police Department. But it's not just here 0:21:09.880 --> 0:21:12.000 in the United States. It's all around the world that 0:21:12.080 --> 0:21:17.480 this company does business. So Facebook has long maintained that 0:21:17.800 --> 0:21:22.560 scraping its sites for data violates Facebook's policies that you 0:21:22.600 --> 0:21:25.960 are not allowed to do that. You cannot create tools 0:21:26.000 --> 0:21:29.600 that are meant to go across all of Facebook and 0:21:29.680 --> 0:21:32.560 just pull down as much data as you possibly can. 0:21:33.200 --> 0:21:37.159 Only Facebook is allowed to collect personal information on that 0:21:37.280 --> 0:21:41.480 kind of scale and then exploit it. I guess what 0:21:41.520 --> 0:21:44.320 I'm saying is I'm just grouchy. I hate all this 0:21:44.520 --> 0:21:47.600 data collection stuff. It doesn't matter to me who ends 0:21:47.680 --> 0:21:51.720 up using it. I just think it's a bad thing 0:21:51.760 --> 0:21:55.919 that we've allowed to have happened in general. Uh Now, admittedly, 0:21:56.600 --> 0:22:00.199 it is way more scary when the agency that's maying 0:22:00.359 --> 0:22:03.720 use of this information is say, law enforcement or the military, 0:22:03.800 --> 0:22:08.400 because we've got very long histories of those kinds of 0:22:08.520 --> 0:22:15.119 institutions disproportionately harming specific populations. So I don't mean to 0:22:15.160 --> 0:22:19.240 diminish this. It is definitely scarier that this is being 0:22:19.359 --> 0:22:24.280 used in relation to things like law enforcement but in general, 0:22:24.640 --> 0:22:27.840 I just think this massive amount of data collection and 0:22:27.920 --> 0:22:35.280 analysis is inherently harmful, whether it's a law enforcement agency 0:22:35.320 --> 0:22:39.560 that's relying upon it, or it's the platform itself like 0:22:39.680 --> 0:22:44.520 Facebook or Instagram. Uh, it's just the older I get, 0:22:44.560 --> 0:22:47.119 the less comfortable I am that we allowed this to 0:22:47.200 --> 0:22:50.520 happen and that it has grown to the extent that 0:22:50.600 --> 0:22:55.160 it has. Okay, I don't mean to be a fear mongerer. 0:22:55.440 --> 0:22:57.480 Let's take another quick break. When we come back, we'll 0:22:57.480 --> 0:23:10.040 have some other news stories to talk about. We're back 0:23:10.680 --> 0:23:14.359 over in the UK, Parliament is drafting an amendment to 0:23:14.480 --> 0:23:17.560 the Online Safety Bill. That's a bill that's intended to 0:23:17.960 --> 0:23:23.960 mostly protect children against harmful content that is proliferated upon 0:23:24.000 --> 0:23:28.000 the Internet, and this particular amendment is likely to have 0:23:28.160 --> 0:23:33.159 tech executives sweating. So a group of lawmakers proposed an 0:23:33.200 --> 0:23:37.800 amendment last year that would seek criminal charges against tech 0:23:37.880 --> 0:23:42.200 executives who failed to protect children from harmful content on 0:23:42.400 --> 0:23:46.720 their respective platforms. So, in other words, Mark Zuckerberg could 0:23:46.760 --> 0:23:53.000 be held criminally responsible for allowing harmful material to perpetuate 0:23:53.080 --> 0:23:56.840 on Facebook, and if if the court could show that 0:23:58.040 --> 0:24:01.280 he didn't do enough to protect Chill Drin then he 0:24:01.320 --> 0:24:04.639 could face criminal charges for that in the UK. So 0:24:05.240 --> 0:24:07.680 this whole thing created kind of a lively debate within 0:24:07.760 --> 0:24:13.240 Parliament between the desire to protect children, which is obviously important, 0:24:13.840 --> 0:24:18.480 and protecting freedom of speech, which is also really important. 0:24:18.840 --> 0:24:21.320 In fact, that's why we have these concepts like safe 0:24:21.359 --> 0:24:25.560 Harbor on the Internet, where a platform is not responsible 0:24:25.600 --> 0:24:29.560 necessarily for the content that its users post to that 0:24:29.640 --> 0:24:33.240 platform because the platform didn't generate it. It's just it's 0:24:33.320 --> 0:24:39.520 just a a place where people gather. So this had 0:24:39.600 --> 0:24:43.360 kind of a struggle within Parliament, and ultimately the Prime 0:24:43.400 --> 0:24:47.399 Minister of England has indicated that what they're gonna do 0:24:47.720 --> 0:24:50.240 is they're going to create a similar amendment to the 0:24:50.280 --> 0:24:53.400 proposed one that aims for the same goal, that will 0:24:53.480 --> 0:24:56.760 hold tech executives responsible for a failure to protect children, 0:24:57.640 --> 0:25:00.639 and that will then be drafted and a amended to 0:25:01.040 --> 0:25:05.960 the Online Safety Bill. The original proposed amendment will be withdrawn. 0:25:07.160 --> 0:25:10.720 I can't pretend to fully understand the political process here, 0:25:11.040 --> 0:25:13.080 except that maybe the goal is to have a more 0:25:13.160 --> 0:25:16.879 focused amendment put in place. Anyway, this would be a 0:25:16.920 --> 0:25:20.440 really dramatic step if it does in fact go through, 0:25:20.520 --> 0:25:24.080 which it seems like that's the case, and it should 0:25:24.080 --> 0:25:28.320 really worry anyone who is a leader in a social 0:25:28.359 --> 0:25:31.840 network because it could mean that they could potentially be 0:25:31.960 --> 0:25:34.920 charged as a criminal in the UK in the future 0:25:35.000 --> 0:25:38.200 if they failed to live up to the Online Safety 0:25:38.240 --> 0:25:42.720 Bills requirements to protect children. And again I think protecting 0:25:42.800 --> 0:25:47.960 children is absolutely critical. I I I have worried about 0:25:48.400 --> 0:25:56.159 social networks effects on children, even beyond the really obvious stuff. 0:25:57.080 --> 0:25:59.840 But yeah, this is a this is a dramatic step 0:26:00.119 --> 0:26:03.280 and we'll have to see how it plays out. It 0:26:03.359 --> 0:26:07.679 also points to the kind of precarious position that the 0:26:07.800 --> 0:26:12.239