WEBVTT - Tech News: A Bad PR Week for AI 0:00:04.360 --> 0:00:12.280 Welcome to tech Stuff, a production from iHeartRadio. He there, 0:00:12.320 --> 0:00:15.880 and welcome to tech Stuff. I'm your host, Jonathan Strickland. 0:00:15.880 --> 0:00:19.919 I'm an executive producer with iHeartRadio and how the Tech Area. 0:00:20.560 --> 0:00:24.280 It's time for the tech News for Tuesday, February twenty first, 0:00:24.440 --> 0:00:28.000 two twenty three, and we're going to start this episode 0:00:28.040 --> 0:00:30.520 off with a news segment that I would like to 0:00:30.560 --> 0:00:35.159 call Blame the Robots. So the Washington Post published an 0:00:35.280 --> 0:00:39.840 article by pranschu Verma titled AI is starting to pick 0:00:40.000 --> 0:00:45.360 who gets laid off, which is perhaps a bit sensationalized 0:00:45.440 --> 0:00:48.800 once you read the actual story, but maybe only a 0:00:48.920 --> 0:00:53.800 little bit sensationalized. All right, So here's how this all unfolds. 0:00:54.800 --> 0:01:00.400 There are software packages and various services and algorithms that 0:01:00.520 --> 0:01:03.960 some hiring managers rely upon in order to do kind 0:01:03.960 --> 0:01:08.000 of a pass filter across job applicants in order to 0:01:08.080 --> 0:01:11.720 narrow down the search. There are companies that offer that 0:01:11.840 --> 0:01:15.039 very service, and it starts to make sense if you're 0:01:15.360 --> 0:01:17.600 hiring at a company that gets a lot of attention. 0:01:18.280 --> 0:01:22.119 So let's say that you're a popular company, you have 0:01:22.280 --> 0:01:26.480 fairly rare job openings, and you get lots, like maybe 0:01:26.520 --> 0:01:30.640 thousands of applicants per job that you list, Well, you 0:01:30.760 --> 0:01:34.399 probably need some help whittling down the applicant list to 0:01:34.400 --> 0:01:37.920 get to a manageable pool of potential hires, right like 0:01:38.000 --> 0:01:42.160 you need something to separate the cream from everything else. 0:01:42.880 --> 0:01:45.560 It's not easy to do, especially at really high volumes. 0:01:45.560 --> 0:01:51.160 And so these services and software packages essentially reduce applicants 0:01:51.200 --> 0:01:53.880 down to data points and they kind of have to. 0:01:54.240 --> 0:01:57.760 And depending on how tough that filter is, folks can 0:01:57.800 --> 0:02:00.160 get weeded out. Sometimes a lot of folks. May you 0:02:00.160 --> 0:02:03.480 get down to, like less than a dozen applicants out 0:02:03.480 --> 0:02:06.960 of thousands has to be pretty brutal. Well, this article 0:02:07.080 --> 0:02:10.840 postulates that we could see the reverse come into play 0:02:10.919 --> 0:02:15.679 as well, that a company might lean on similar software 0:02:15.720 --> 0:02:20.440 and services to identify the people who contribute the least 0:02:20.520 --> 0:02:23.320 to the company or perform at a level that's considered 0:02:23.360 --> 0:02:26.880 to be below their peers. Therefore, they could be candidates 0:02:26.880 --> 0:02:30.200 for layoffs when the corporate overlords deemed that it is 0:02:30.280 --> 0:02:35.480 time to reduce headcount in these troubling times. And therein 0:02:35.560 --> 0:02:40.480 lies the story, right that an algorithm might determine that 0:02:40.639 --> 0:02:45.920 you are expendable instead of your human boss. And I 0:02:45.960 --> 0:02:48.480 am aware that I'm making an assumption here that your 0:02:48.520 --> 0:02:52.639 boss is in fact human, some program, you know, some 0:02:52.760 --> 0:02:57.440 freaking robot has determined that you're getting laid off. It 0:02:57.560 --> 0:03:02.520 sounds positively dystopia, doesn't it. And with big tech companies 0:03:02.600 --> 0:03:06.320 laying off thousands of folks over this year and the 0:03:06.360 --> 0:03:11.160 previous year, it's easy to imagine managers shrugging off the 0:03:11.240 --> 0:03:14.040 responsibility of telling someone they no longer have a job 0:03:14.639 --> 0:03:17.080 by leaving it to the old zeros and ones of 0:03:17.080 --> 0:03:22.200 a presumably objective and emotionless system. But this in turn 0:03:22.760 --> 0:03:27.559 brings up other problems. As I have mentioned in tons 0:03:27.600 --> 0:03:30.840 of episodes, one of the many problems we have in 0:03:30.960 --> 0:03:35.960 AI systems can come down to unintended bias within the 0:03:36.000 --> 0:03:39.240 system itself. So if the system is biased, it could 0:03:39.320 --> 0:03:44.600 end up targeting employees of specific ethnicities or backgrounds. Now 0:03:44.680 --> 0:03:47.960 Vermat makes this argument in the Washington Post article. He 0:03:47.960 --> 0:03:52.120 says that if the algorithm were to, say, determine that 0:03:52.280 --> 0:03:55.280 people of color have a higher incidence rate of leaving 0:03:55.320 --> 0:03:58.880 their jobs, that a person of color is more likely 0:03:58.920 --> 0:04:03.400 to leave their job than say, they're white colleagues, well, 0:04:03.520 --> 0:04:06.520 then the system might naturally start to target employees who 0:04:06.560 --> 0:04:09.440 happen to be people of color for the purposes of layoffs. 0:04:09.440 --> 0:04:14.240 But then you're getting into very dangerous legal and ethical territory. 0:04:14.680 --> 0:04:18.320 It's as if you're targeting these specific people because of 0:04:18.360 --> 0:04:22.479 their race. Also, I'm not sure how well an algorithm 0:04:22.600 --> 0:04:27.920 can actually judge a person's contributions. Presumably stuff from employee 0:04:27.960 --> 0:04:30.760 reviews and such would play a big part, But in 0:04:31.080 --> 0:04:35.800 highly collaborative work, a person could act as the sort 0:04:35.800 --> 0:04:38.640 of lynch pen that keeps a team working really well together, 0:04:39.200 --> 0:04:42.840 even if they themselves don't have the highest numbers on 0:04:42.920 --> 0:04:47.279 whatever the deliverables are. So in my opinion, relying on 0:04:47.400 --> 0:04:52.960 AI to make or even guide decisions regarding layoffs is 0:04:53.000 --> 0:04:56.599 really a bad move all around. It can make sense 0:04:57.080 --> 0:05:00.840 in the applicant phase, but in layoffs I would say 0:05:01.000 --> 0:05:03.919 avoid it. It doesn't look good for the company. It 0:05:03.960 --> 0:05:07.120 could ultimately lead to choices that will harm the overall 0:05:07.240 --> 0:05:11.320 organization in the long run. In this article, Verma mentions 0:05:11.360 --> 0:05:14.440 that while folks at Google wondered if perhaps they had 0:05:14.480 --> 0:05:16.919 been laid off due to an algorithm choosing them because 0:05:17.560 --> 0:05:19.920 there didn't seem to be much rhyme or reason to 0:05:19.960 --> 0:05:24.000 the layoffs, the company denies making use of anything of 0:05:24.040 --> 0:05:27.719 the sort. There's kind of a distinct lack of cases 0:05:27.760 --> 0:05:31.839 where we know that an algorithm definitively played a part 0:05:32.360 --> 0:05:36.599 in layoffs. However, Verma in the article also cites a 0:05:36.680 --> 0:05:40.800 survey that showed ninety eight percent of HR managers there 0:05:40.800 --> 0:05:44.719 were three hundred of them participating in the survey, had 0:05:44.760 --> 0:05:47.760 said that they plan to rely on software and algorithms 0:05:47.800 --> 0:05:52.360 to help make such decisions about layoffs this year. So 0:05:52.839 --> 0:05:55.080 even if you were to argue it hasn't happened yet, 0:05:55.160 --> 0:05:58.000 it looks like it's going to happen real soon. My 0:05:58.120 --> 0:06:01.560 guess is we'll see some high cases where some company 0:06:01.640 --> 0:06:06.560 relies too heavily on algorithms and it'll come back to 0:06:06.600 --> 0:06:09.560 haunt them, perhaps only in PR, but it will be 0:06:10.680 --> 0:06:14.039 a big blowback, and then maybe then we'll start to 0:06:14.040 --> 0:06:19.800 see people form best practices around the whole thing. I 0:06:19.839 --> 0:06:23.080 still think it feels a bit like shirking responsibility in 0:06:23.120 --> 0:06:27.039 my opinion, If the top brass decides that layoffs are necessary, 0:06:27.320 --> 0:06:30.480 then they are obligated to make each and every layoff 0:06:30.520 --> 0:06:34.919 decision transparent and honest, I think they owe their employees 0:06:34.960 --> 0:06:38.280 as much. And it's really infuriating because you'll see managers 0:06:38.279 --> 0:06:43.680 who get a directive saying you have to apply this 0:06:43.880 --> 0:06:49.479 artificial bell curve to the employees who are reporting to you. 0:06:50.080 --> 0:06:52.040 We heard a story about that just a couple of 0:06:52.040 --> 0:06:55.719 weeks ago, where a director actually essentially was fired for 0:06:55.800 --> 0:07:01.560 refusing to follow that because it arbitrarily requires managers to 0:07:01.600 --> 0:07:04.599 assign people as low performers even if you don't have 0:07:04.640 --> 0:07:08.760 any low performers on your team, and that just again 0:07:08.920 --> 0:07:12.920 seems inherently unfair. I feel like relying on AI to 0:07:12.960 --> 0:07:17.360 make these choices also is inherently unfair and can miss 0:07:17.400 --> 0:07:21.240 some really important factors that may not reduce down to 0:07:21.400 --> 0:07:24.800 pure data. But we've got a lot of other AI 0:07:24.920 --> 0:07:27.840 news to get through today. A lot of it is bad. 0:07:27.880 --> 0:07:30.920 I'm not gonna lie. And our next story comes from 0:07:31.000 --> 0:07:36.400 Vanderbilt University. The Peabody College at Vendorbilt, and that college's 0:07:36.440 --> 0:07:39.640 Office of Equity, Diversity and Inclusion send out a message 0:07:39.680 --> 0:07:42.920 to students in the wake of the terrible shooting at 0:07:42.960 --> 0:07:48.200 Michigan State University. And clearly this was a delicate task 0:07:48.640 --> 0:07:52.520 that needed empathy and support. I needed a message that 0:07:52.600 --> 0:07:57.000 showed that Vanderbilt's staff have students and their welfare at 0:07:57.000 --> 0:08:00.280 the top. Of their priority list, so of course they 0:08:00.360 --> 0:08:04.400 use chat GPT to help craft the message. This pretty 0:08:04.480 --> 0:08:08.600 much sent me spiraling, y'all, because passing the buck to 0:08:08.720 --> 0:08:13.000 AI to handle things that are this important, things that 0:08:13.120 --> 0:08:18.960 intrinsically involve a very human connection, it just feels beyond 0:08:19.040 --> 0:08:22.960 shortsighted and crass to me. At best, you could say 0:08:23.120 --> 0:08:26.120 this was a poor decision, but at worst it implies 0:08:26.160 --> 0:08:29.440 that leadership has little to no regard for students and 0:08:29.560 --> 0:08:32.080 instead we'll just lean on the robots to handle the 0:08:32.120 --> 0:08:35.120 tough stuff. Anyway, the end of the message contained the 0:08:35.200 --> 0:08:41.040 line paraphrased from open ais Chat GBTAI Language Model Personal Communication, 0:08:42.200 --> 0:08:45.000 and at least the word paraphrase indicates that there was 0:08:45.160 --> 0:08:48.880 human involvement in taking the generated message and shaping it 0:08:48.920 --> 0:08:53.000 properly for students. So it was a collaborative effort, you 0:08:53.000 --> 0:08:56.560 could say. But still, the fact that staff tapped AI 0:08:56.720 --> 0:08:59.199 in the first place to help with such a sensitive 0:08:59.240 --> 0:09:04.520 matter doesn't look good. It looks like people who want 0:09:04.600 --> 0:09:09.400 to avoid the hard stuff, and hard stuff isn't the 0:09:09.559 --> 0:09:14.680 human connection stuff, the stuff that has incredible impact on 0:09:14.760 --> 0:09:19.000 emotion and mental health, whether it's layoffs or counseling people 0:09:19.080 --> 0:09:22.800 in the wake of a violent act like the shooting 0:09:22.800 --> 0:09:26.240 at the University of Michigan. That is the wrong way 0:09:26.240 --> 0:09:30.160 to use AI, in my opinion. That is inherently the 0:09:30.160 --> 0:09:36.880 realm of humanity and to off source that to AI's 0:09:38.120 --> 0:09:42.600 it shows such a huge disregard for the people who 0:09:42.600 --> 0:09:48.000 are ultimately the recipients of those messages that I think 0:09:48.000 --> 0:09:53.360 it's unconscionable. Now the Associate Dean and Assistant Dean who 0:09:53.360 --> 0:09:57.439 are part of this process have both stepped back from 0:09:57.600 --> 0:10:00.840 that office of equity, diversity and inclusion, which is probably 0:10:00.880 --> 0:10:05.400 for the best, but yeah, this was a really bad 0:10:05.520 --> 0:10:11.080 use case for AI. Last week, the representatives from around 0:10:11.080 --> 0:10:15.240 the world attended the Summit on Responsible Artificial Intelligence in 0:10:15.240 --> 0:10:20.720 the Military Domain or re aim our AIM and we've 0:10:20.720 --> 0:10:24.440 talked about how incorporating AI into military processes and hardware 0:10:24.559 --> 0:10:31.240 raises really difficult questions regarding safety, accountability, escalation, and more. 0:10:31.920 --> 0:10:36.160 Reps from many countries, including the United States and China, 0:10:36.200 --> 0:10:41.600 but excluding Russia, which wasn't invited and Ukraine did not attend. 0:10:41.760 --> 0:10:44.720 They were invited, but clearly have other things going on 0:10:44.800 --> 0:10:48.160 at the moment. Anyway, these representatives all met to discuss 0:10:48.280 --> 0:10:53.040 the issues of AI in its role in military operations. 0:10:53.559 --> 0:10:56.120 At the conclusion of the summit, all but one of 0:10:56.160 --> 0:10:59.959 the representatives of the countries that attended signed an agree 0:11:00.400 --> 0:11:04.280 to commit to developing AI military applications that quote do 0:11:04.440 --> 0:11:08.960 not undermine international security, stability and accountability end quote. So 0:11:09.200 --> 0:11:13.520 what was the one nation that abstained. That would be Israel. Now, 0:11:13.880 --> 0:11:19.360 don't like heap tons of criticism on Israel, because there 0:11:19.400 --> 0:11:25.200 are critics who say this entire meeting was largely for show, 0:11:25.600 --> 0:11:29.560 because according to critics, there was nothing in the summit 0:11:29.679 --> 0:11:32.920 or the agreement that is legally binding for any of 0:11:32.960 --> 0:11:36.880 the countries involved. So, in other words, the critics are 0:11:36.880 --> 0:11:40.120 saying that the reps are all like, yeah, yeah, totally 0:11:40.160 --> 0:11:42.760 AI killing people would really be bad. Let's totally not 0:11:42.840 --> 0:11:46.600 do that, but they would have no real accountability to 0:11:46.720 --> 0:11:50.880 follow through on that promise. Further, the agreement did not 0:11:51.040 --> 0:11:54.480 include certain AI assisted or controlled systems that are already 0:11:54.480 --> 0:11:58.640 in use, like AI controlled drones, And so there's concern 0:11:58.720 --> 0:12:02.920 that this agreement, while represents little more than just putting 0:12:02.920 --> 0:12:05.480 on a show to say, yes, we're all aware of 0:12:05.520 --> 0:12:09.240 this and it is a bad thing. Now, Honestly, I 0:12:10.640 --> 0:12:15.600 heavily suspect that several countries, including the United States and 0:12:15.679 --> 0:12:21.840 also China, will continue incorporating AI into military applications, including 0:12:22.080 --> 0:12:27.520 weaponized AI. I would be absolutely shocked if they didn't 0:12:27.600 --> 0:12:31.240 continue down that pathway, because there's a very real fear 0:12:31.520 --> 0:12:33.840 that if you don't do it, the other guy will, 0:12:33.880 --> 0:12:36.679 and then there will be an AI gap. Now maybe 0:12:36.760 --> 0:12:38.480 I only think that because I'm a child of the 0:12:38.520 --> 0:12:41.600 seventies and eighties and I saw how this very similar 0:12:41.600 --> 0:12:45.839 scenario played out with nuclear armaments, because boy howdy, was 0:12:45.880 --> 0:12:49.280 that a thing. So I would love for us to 0:12:49.480 --> 0:12:54.080 avoid the mistakes of the past. But I really am 0:12:54.240 --> 0:12:58.600 skeptical that that's going to happen, because again, unless everybody 0:12:59.559 --> 0:13:03.680 is hell accountable and agrees to not further that work, 0:13:04.160 --> 0:13:06.319 someone is going to. And if someone's going to, then 0:13:06.400 --> 0:13:11.120 everyone is going to because otherwise you are at a disadvantage. Okay, 0:13:11.760 --> 0:13:13.320 with all that doom and gloom out of the way, 0:13:13.400 --> 0:13:15.560 let's take a quick break. When we come back, I've 0:13:15.600 --> 0:13:27.760 got some more news items to talk about. We're back now. 0:13:27.760 --> 0:13:29.839 I still have a couple more AI stories, but these 0:13:29.880 --> 0:13:32.280 are not quite as apocalyptic as the ones we started 0:13:32.360 --> 0:13:36.120 off with. The one is that Clark's world magazine, which 0:13:36.120 --> 0:13:39.760 has been publishing fantasy and science fiction stories online since 0:13:39.800 --> 0:13:45.280 two thousand and six, has temporarily stopped accepting submissions. Why 0:13:46.040 --> 0:13:50.920 because apparently the magazine has received too many AI generated submissions, 0:13:50.960 --> 0:13:54.120 and until they have access to better tools to detect 0:13:54.120 --> 0:13:57.320 those kinds of things, they have chosen to hold off 0:13:57.440 --> 0:14:03.720 accepting anymore. That's both understandable and it stinks. Not that 0:14:03.760 --> 0:14:06.400 I think the magazine made the wrong call here. I 0:14:06.400 --> 0:14:10.520 think this is the right call, but rather it stinks 0:14:10.559 --> 0:14:14.440 because there are genuine authors and would be authors out 0:14:14.440 --> 0:14:17.640 there who have great stories to tell, and they're seeing 0:14:17.760 --> 0:14:20.000 an outlet closed off to them, at least for now, 0:14:20.440 --> 0:14:23.680 and it's all thanks to AI generated stories. Now, I 0:14:23.680 --> 0:14:26.440 think in most cases the AI generated stuff has to 0:14:26.480 --> 0:14:31.160 be a collaborator kind of relationship, because in my own 0:14:31.200 --> 0:14:35.800 experience the stories generated by AI they aren't very good. Like, 0:14:36.160 --> 0:14:40.200 grammatically they work, and you know, you get some interesting 0:14:40.280 --> 0:14:43.280 descriptions and stuff, but the actual stories tend to be 0:14:43.440 --> 0:14:49.160 pretty mundane and uninteresting. I imagine that most folks who 0:14:49.160 --> 0:14:52.160 are using AI are leaning on it for stuff like 0:14:52.280 --> 0:14:55.560 generating initial ideas, maybe shaping a certain part of the 0:14:55.680 --> 0:15:01.160 narrative at least any stories that are difficult to determine, Oh, 0:15:01.200 --> 0:15:03.680 this was made by AI, right, if they haven't done 0:15:03.720 --> 0:15:07.240 any massaging, it often is pretty easy to detect that 0:15:07.240 --> 0:15:09.400 it's AI, or the very least, it's easy to detect 0:15:09.400 --> 0:15:12.200 that's not a very good story and it wouldn't pass 0:15:12.240 --> 0:15:16.440 the bar for publication. But still, here's another example of 0:15:16.480 --> 0:15:19.120 how AI can end up harming creative types, whether it's 0:15:19.120 --> 0:15:23.000 from the unauthorized copying of their style or displacing them 0:15:23.040 --> 0:15:26.400 from the creator community. Insider reports something that I think 0:15:26.400 --> 0:15:28.960 most folks already have a pretty good handle on, and 0:15:29.040 --> 0:15:33.359 that is the emergence and reception to chat GPT probably 0:15:33.480 --> 0:15:35.840 means we're going to see a whole bunch of copycats 0:15:35.840 --> 0:15:38.760 in the very near future. And to be clear, chatbots 0:15:38.920 --> 0:15:41.880 have been a thing for years. I'm sure you're all 0:15:41.880 --> 0:15:44.960 aware of that. In fact, someone who is once in 0:15:45.000 --> 0:15:48.120 the business of reporting on tech, a person whom I 0:15:48.200 --> 0:15:51.560 know and respect and like very very much, ended up 0:15:51.560 --> 0:15:55.440 working at a company that developed sophisticated chatbots. But these 0:15:55.440 --> 0:15:59.600 were tools that were intended for narrow use cases, something 0:15:59.600 --> 0:16:02.520 that would work well within the confines of a particular 0:16:02.560 --> 0:16:07.560 company's services and processes. The stuff we're seeing now is 0:16:07.600 --> 0:16:10.920 made to be more general purpose, and with that comes 0:16:10.920 --> 0:16:14.920 the problems of reliability and accuracy as well as transparency. 0:16:15.720 --> 0:16:20.520 It is easier, not easy, mind you, but easier to 0:16:20.560 --> 0:16:24.120 build a reliable and accurate tool that works within an 0:16:24.280 --> 0:16:28.400 enclosed system, like the customer service arm of a consumer 0:16:28.440 --> 0:16:31.720 facing company. But it's another thing when it's just you know, 0:16:31.880 --> 0:16:35.320 free range AI chat bot. And meanwhile, the guy who 0:16:35.400 --> 0:16:39.200 runs the company that made chat GPT has said repeatedly 0:16:40.240 --> 0:16:43.640 that he thinks chat GPT isn't that good, or at 0:16:43.680 --> 0:16:47.960 least it is far from perfect. And yet we're currently 0:16:48.000 --> 0:16:54.840 living through a buzzy, height heightened age of chat GPT 0:16:55.280 --> 0:17:00.320 and its peers like like Barred from Google. Tech Runch 0:17:00.320 --> 0:17:03.480 has a piece titled the AI photo app trend has 0:17:03.480 --> 0:17:06.960 already fizzled, new data shows and you should totally check 0:17:07.000 --> 0:17:09.520 out this article. The author of the piece, Sarah Perez, 0:17:10.000 --> 0:17:13.320 lays out some of the data, including download numbers and revenue, 0:17:13.480 --> 0:17:16.120 and she shows that while the text to image AI 0:17:16.200 --> 0:17:20.160 tools initially made a really big splash when they started 0:17:20.200 --> 0:17:24.840 to emerge, particularly late last year, excitement has dropped off 0:17:24.880 --> 0:17:29.040 considerably since then. There's been a lot of backlash in 0:17:29.080 --> 0:17:32.199 the space, ranging from artists who are understandably upset to 0:17:32.200 --> 0:17:36.360 see their style co opted by AI, to users who 0:17:36.359 --> 0:17:39.880 are concerned that the tools can create inappropriate images far 0:17:39.960 --> 0:17:44.199 too readily, and that any restrictions that are designed to 0:17:44.320 --> 0:17:48.040 limit that sort of stuff aren't always the best. Whether 0:17:48.080 --> 0:17:51.640 those actually played a big part in cooling this trend, 0:17:51.840 --> 0:17:54.439 or maybe it was just that folks were getting tired 0:17:54.440 --> 0:17:57.000 of the shiny new thing and they had already moved on, 0:17:57.720 --> 0:18:01.359 I am uncertain, but my guests is that we're going 0:18:01.400 --> 0:18:04.800 to see the space continue to evolve, perhaps with fewer 0:18:04.840 --> 0:18:08.080 players as this goes on, if some of them find 0:18:08.080 --> 0:18:11.520 it too difficult to cover costs with the declining revenues. 0:18:12.119 --> 0:18:15.040 But I don't think AI generated imagery is going to 0:18:15.200 --> 0:18:18.240 just go away at this point. Now that being said, 0:18:19.280 --> 0:18:22.440 one fun story, or at least in my opinion, it's 0:18:22.480 --> 0:18:26.119 fun that relates to the AI generated imagery involves a 0:18:26.200 --> 0:18:31.040 robot from Carnegie Mellon University. So it's a robot arm 0:18:31.080 --> 0:18:36.159 that has the name Frieda, which yes, is both a 0:18:36.200 --> 0:18:41.800 tribute to Frieda the artist, and is an acronym that 0:18:41.840 --> 0:18:46.440 stands for Framework and Robotics Initiative for developing arts, and 0:18:46.600 --> 0:18:51.040 it also generates images based on text prompts. Only in 0:18:51.080 --> 0:18:54.840 this case, the images it makes are not digital images. 0:18:54.880 --> 0:18:58.680 They're not computer generated images. They are real world paintings. 0:18:59.080 --> 0:19:02.720 You have a robot that paints with actual brushes and 0:19:02.840 --> 0:19:06.679 actual paint It creates works of art based off text 0:19:06.800 --> 0:19:11.000 input and directions. According to tech Spot, it takes about 0:19:11.040 --> 0:19:14.720 an hour from the point where the robot receives input 0:19:14.920 --> 0:19:17.040 in the form of the text to the point where 0:19:17.040 --> 0:19:19.400 it begins to paint, because it actually has to plot 0:19:19.400 --> 0:19:24.000 out how it's going to physically paint this. How are 0:19:24.000 --> 0:19:26.199 the brushstrokes going to go, how long are they going 0:19:26.240 --> 0:19:28.160 to go, how much pressure is going to be used? 0:19:28.960 --> 0:19:34.960 What style is it going to follow? And that's very 0:19:35.040 --> 0:19:39.200 understandable because as we all know, there are different strokes 0:19:39.240 --> 0:19:41.840 for different folks, because the world don't move to the 0:19:41.880 --> 0:19:45.119 beat of just one drum. Shout out to me if 0:19:45.160 --> 0:19:49.399 you get that reference. Anyway, The roboticists and engineers are 0:19:49.480 --> 0:19:53.600 quick to say that Frieda isn't an artist. FRIEDA is 0:19:53.600 --> 0:19:58.120 a collaborator. FRIEDA is not creative. FRIEDA just follows instructions 0:19:58.160 --> 0:20:03.040 as best it can pay the subject of the art 0:20:03.160 --> 0:20:07.919 in the style that was dictated by its collaborator. Anyway, 0:20:07.960 --> 0:20:10.560 I just thought this was a neat take on AI 0:20:10.720 --> 0:20:15.240 generated images. Somehow it feels different that because it's, you know, 0:20:15.280 --> 0:20:19.240 a physical painting. It's something that you could hold or 0:20:19.280 --> 0:20:21.680 put into a frame and hang on a wall, or 0:20:22.080 --> 0:20:24.960 you know, something you could rip apart in rage as 0:20:25.040 --> 0:20:28.920 robots get another art commission and you don't. Now, finally 0:20:29.800 --> 0:20:32.320 we're off of the AI stories and we can get 0:20:32.320 --> 0:20:35.280 to everything else. So next up. Part of the big 0:20:35.280 --> 0:20:37.679 news this week is Meta really shook things up on 0:20:37.720 --> 0:20:42.000 Sunday announcing that the company is introducing a subscription service 0:20:42.040 --> 0:20:46.440 called Meta Verified. It's just in the testing phase now, 0:20:46.840 --> 0:20:49.720 and you know, the plan is to widely deploy it, 0:20:49.760 --> 0:20:52.920 but we'll see if things go poorly in the test 0:20:53.000 --> 0:20:56.320 markets that Meta is trying out at the moment. But essentially, 0:20:57.200 --> 0:21:02.280 this subscription service is a verification tool. Users would have 0:21:02.320 --> 0:21:04.840 to submit proof that they are who they claim to be, 0:21:05.640 --> 0:21:09.760 using government issued ID for example, and in return for 0:21:09.920 --> 0:21:12.840 twelve bucks a month or fifteen bucks a month if 0:21:12.840 --> 0:21:16.040 you're doing it on iOS or Android, because Google and 0:21:16.080 --> 0:21:19.120 Apple take their own cut of the fee. You then 0:21:19.240 --> 0:21:23.680 get a little blue badge on Facebook and or Instagram 0:21:24.000 --> 0:21:28.520 saying you're bona fide. On top of that, badge subscribers 0:21:28.520 --> 0:21:31.439 will also have access to services that are meant to 0:21:31.440 --> 0:21:35.920 protect against imposter accounts. They're supposed to get better customer support, 0:21:36.520 --> 0:21:39.960 and they're supposed to get improved discoverability when folks are 0:21:39.960 --> 0:21:42.920 actually searching for them, which am I opinion, is stuff 0:21:43.000 --> 0:21:46.040 that should really be standard for all users, whether they're 0:21:46.040 --> 0:21:49.600 paying a subscription or not. I think it's kind of 0:21:49.680 --> 0:21:53.359 bullpucky to say one of the benefits to verification is 0:21:53.400 --> 0:21:55.639 that meta will make sure other folks aren't trying to 0:21:55.880 --> 0:21:59.679 impersonate you. I mean, arguably, this is a bigger problem 0:21:59.720 --> 0:22:02.439 for notable folks like celebrities and brands. And I'm not 0:22:02.480 --> 0:22:06.639 talking me, I'm talking real celebrities. I have no illusions 0:22:06.680 --> 0:22:10.199 that I'm a celebrity, but I've still seen plenty of 0:22:10.240 --> 0:22:14.000 instances of friends being impersonated as someone has either gained 0:22:14.040 --> 0:22:17.480 access to their account or created a copy account in 0:22:17.560 --> 0:22:21.359 an attempt to phish for data. Like this is still 0:22:21.400 --> 0:22:25.880 something that affects the average person on these platforms. It's 0:22:25.960 --> 0:22:30.199 not just for the celebrities. But yeah, I'm that's facing 0:22:30.200 --> 0:22:32.360 issues with the revenue for a lot of different reasons, 0:22:32.800 --> 0:22:36.200 so it's not surprising that the company is now introducing 0:22:36.200 --> 0:22:39.240 the subscription feature. It just feels like the quote unquote 0:22:39.240 --> 0:22:43.159 benefits of the service are things that really everyone on 0:22:43.200 --> 0:22:48.040 the platform should have access to by default. Maybe I'm 0:22:48.080 --> 0:22:53.040 just being unreasonable here. Today Microsoft will attempt to defend 0:22:53.160 --> 0:22:56.760 its planned acquisition of Activision Blizzard in the EU in 0:22:56.840 --> 0:23:01.040 a meeting that's behind closed doors and Brussels. Previously, EU 0:23:01.160 --> 0:23:04.760 regulators indicated that they would block the purchase, saying it 0:23:04.760 --> 0:23:07.359 would result in less competition in the video game space 0:23:07.680 --> 0:23:11.520 and allow Microsoft to engage in actively anti competitive practices, 0:23:11.520 --> 0:23:16.040 such as preventing other platforms like Sony PlayStation from having 0:23:16.080 --> 0:23:20.800 access to popular video game franchises like Call of Duty. Earlier, 0:23:20.920 --> 0:23:25.520 Microsoft Reps signed a deal with Nintendo Reps that legally 0:23:25.600 --> 0:23:29.320 binds Microsoft to bring Call of Duty titles to Nintendo 0:23:29.359 --> 0:23:33.520 platforms for ten years, and further that all titles will 0:23:33.520 --> 0:23:37.119 be available on Nintendo platforms the same day that they 0:23:37.160 --> 0:23:41.280 come out for Xbox platforms, with quote full feature and 0:23:41.480 --> 0:23:46.959 content parity end quote between these versions, meaning Nintendo won't 0:23:47.000 --> 0:23:49.359 have to be happy with a watered down version of 0:23:49.400 --> 0:23:51.560 Call of Duty. It's going to get the real thing, 0:23:51.680 --> 0:23:55.000 just like Xboxes. This puts pressure on Sony to make 0:23:55.000 --> 0:23:58.960 a similar agreement or else Microsoft could argue before the 0:23:59.040 --> 0:24:04.040 EU regulators that Microsoft has made attempts to ensure fairness 0:24:04.040 --> 0:24:07.560 between the various console companies, but Sony isn't playing ball 0:24:07.680 --> 0:24:11.880 on purpose in an effort to scuttle the deal, surprisingly, 0:24:11.920 --> 0:24:16.480 at least to me, the Communications Workers of America the CWA, 0:24:16.640 --> 0:24:20.680 a union organization here in the US, has also urged 0:24:20.720 --> 0:24:24.480 the EU to approve the acquisition deal. They say that 0:24:24.560 --> 0:24:28.520 Microsoft has been more receptive to attempts at unionizing than 0:24:28.600 --> 0:24:33.320 Activision Blizzard has, and that without Microsoft's oversight, employees and 0:24:33.440 --> 0:24:39.080 Activision could find themselves facing tough managerial resistance to unionizing. 0:24:39.640 --> 0:24:42.800 By the time you hear this, a decision has probably 0:24:42.840 --> 0:24:45.239 been made one way or the other. But as I 0:24:45.240 --> 0:24:48.800 write this episode, it has yet to be announced, and again, 0:24:48.840 --> 0:24:51.520 the meeting is behind closed doors, so it might be 0:24:51.560 --> 0:24:54.080 a little while before we find out what the results are. 0:24:54.560 --> 0:24:58.359 Corporate employees at Amazon are looking at decreased compensation this 0:24:58.440 --> 0:25:01.640 year like an actual pay now. The reason for that 0:25:01.800 --> 0:25:04.600 is because some of their compensation is tied up in 0:25:04.720 --> 0:25:09.000 stock units, so as part of their salary, Amazon corporate 0:25:09.040 --> 0:25:14.080 workers get stock in Amazon. However, Amazon stock price has 0:25:14.119 --> 0:25:17.320 taken some massive hits over the last year, and that 0:25:17.359 --> 0:25:20.280 means that the stocks awarded to corporate employees are worth 0:25:20.400 --> 0:25:24.760 much less than they were a year earlier. That's particularly 0:25:24.800 --> 0:25:29.000 tough because when Amazon structures its salary deals, they are 0:25:29.119 --> 0:25:31.560 at least partly based on the idea of the stock 0:25:31.720 --> 0:25:34.560 having a value of around one hundred and seventy dollars 0:25:34.560 --> 0:25:37.119 per share. So, in other words, that's part of the 0:25:37.160 --> 0:25:41.040 justification of yes, your salary is X amount of dollars 0:25:41.119 --> 0:25:44.960 instead of why, because you're also being compensated by stock 0:25:45.040 --> 0:25:47.960 units that are considered to be worth one hundred seventy 0:25:47.960 --> 0:25:52.400 dollars per share. However, at the time of recording, Amazon 0:25:52.480 --> 0:25:55.760 stock is currently at ninety four dollars fifty eight cents 0:25:55.800 --> 0:26:00.440 per share, so a little more than half of what 0:26:00.840 --> 0:26:05.320 it was when these salary figures were first calculated. So 0:26:05.480 --> 0:26:07.800 if the cash part of your salary is dependent upon 0:26:07.840 --> 0:26:09.760 the fact that the rest of your compensation is coming 0:26:09.760 --> 0:26:12.280 in the form of stocks that were calculated at one 0:26:12.359 --> 0:26:14.879 hundred and seventy bucks per share, it means you're getting 0:26:14.920 --> 0:26:18.200 significantly less per year on top of that, the company 0:26:18.240 --> 0:26:21.080 has been laying off thousands of employees. I wouldn't be 0:26:21.119 --> 0:26:23.440