WEBVTT - Week in Tech: The Music Industry Strikes Back 0:00:12.119 --> 0:00:16.120 From Kaleidoscope and iHeart podcasts. This is tech Stuff. I'm 0:00:16.120 --> 0:00:18.880 mons Valashan, and I'm delighted that Cara Price is back 0:00:18.920 --> 0:00:22.040 with us. Today we'll be bringing you the emotional highs 0:00:22.079 --> 0:00:25.680 and lows of Silicon Valley, from a tech Titan's gilded 0:00:25.760 --> 0:00:30.360 nuctuals to the soulless writing of checchipt. Then, on a 0:00:30.360 --> 0:00:32.600 more serious note, we'll be looking at the conflict between 0:00:32.600 --> 0:00:34.839 Iran and Israel in cyberspace. 0:00:35.240 --> 0:00:39.440 Cyber attacks are often viewed as a much less costly 0:00:39.840 --> 0:00:42.960 in terms of funding, but also in terms of lives, 0:00:43.440 --> 0:00:47.080 way to critically wound an enemy. 0:00:48.120 --> 0:00:51.000 All of that on the weekend tech It's Friday, June 0:00:51.000 --> 0:00:54.680 twenty seventh. Welcome back, Cara. 0:00:54.840 --> 0:00:55.760 Thank you as. 0:00:56.000 --> 0:00:58.640 Back to tech Stuff, just in time to cover Jeff 0:00:58.680 --> 0:01:00.160 and Lauren's wedding out. 0:01:00.160 --> 0:01:03.160 A coincidence at all, I wouldn't have come back otherwise. 0:01:03.400 --> 0:01:05.920 You do love a Selim spectacle, and this one's going 0:01:05.959 --> 0:01:11.440 to be pretty spectacular, perhaps unsurprisingly given Jeff Bezos's fourtune 0:01:11.480 --> 0:01:13.880 or two hundred and thirty billion dollars and seems to 0:01:13.920 --> 0:01:17.039 be thirsty for a big hype bash this weekend. 0:01:17.280 --> 0:01:19.840 And what do you do when you're the world's richest man, 0:01:20.480 --> 0:01:22.880 you rent out Venice. Now, ven I think that's what 0:01:22.920 --> 0:01:25.920 you do. We don't know much more about the ceremony 0:01:25.959 --> 0:01:29.560 or the parties for that matter. I can guess that 0:01:29.640 --> 0:01:33.480 some of the guests will be Kim Kardashian, who was 0:01:33.520 --> 0:01:37.479 on Lauren Sanchez's bachelorette and then some others who were 0:01:37.520 --> 0:01:41.280 on the Blue Origin spaceflight, which was not her bachelorette party. 0:01:41.280 --> 0:01:46.240 By the way, so the Virgin spaceflight was Kim Kardashian. No, No, Katy. 0:01:45.959 --> 0:01:48.320 Perry, Katy Perry, Gail King. 0:01:48.720 --> 0:01:51.360 I remember when Katy Perry came down from space. There 0:01:51.560 --> 0:01:55.320 was a article in any Mail being like, it's confirmed 0:01:55.400 --> 0:01:56.800 Katy Perry's career is over. 0:01:57.680 --> 0:01:59.120 Oh, I thought they were going to say it's confirmed 0:01:59.160 --> 0:02:00.520 Katy Perry is an alien. 0:02:01.560 --> 0:02:04.720 I wonder which of the if the tech Titans will 0:02:04.800 --> 0:02:08.359 be there, zak Elon, Tim Cook from Apple or Tim 0:02:08.400 --> 0:02:12.480 Apples he's otherwise known, Larian Serge from Google. I would 0:02:12.520 --> 0:02:14.560 guess there'll be fewer at the wedding than they were 0:02:14.600 --> 0:02:15.960 at Trump's inauguration. 0:02:17.440 --> 0:02:18.400 Maybe maybe not. 0:02:18.639 --> 0:02:21.640 The details are murky, but there is a big banner 0:02:21.720 --> 0:02:24.359 flying in Venice at the moment saying if you can 0:02:24.360 --> 0:02:26.880 rent Venice for your wedding, you can pay more tax. 0:02:27.320 --> 0:02:30.200 You know, this wedding is so secret. I was making 0:02:30.240 --> 0:02:32.280 a joke about renting out Venice, but it is hard 0:02:32.320 --> 0:02:36.040 to tell. The party organizers of the wedding are denying it. 0:02:36.080 --> 0:02:38.760 They actually released a statement saying it's not a city 0:02:38.800 --> 0:02:42.320 takeover and the couple is committed to this is like, 0:02:42.360 --> 0:02:45.880 this is insane. The couple is committed to minimizing disruption, 0:02:46.320 --> 0:02:50.280 and Bezos donated three million dollars to various local organizations. 0:02:50.600 --> 0:02:54.520 In the face of criticism, local Venice officials chimed in 0:02:54.560 --> 0:02:57.720 and said the couple has only booked about thirty water 0:02:57.800 --> 0:03:00.400 taxis for the event. I just live for this. 0:03:00.800 --> 0:03:02.600 I do too. And you pointed me to a piece 0:03:02.600 --> 0:03:04.440 in the Wall Street Journal that was the best one. 0:03:04.639 --> 0:03:07.400 It had a headline promising to spill the secrets and 0:03:07.480 --> 0:03:09.680 another the fact, when there's a big, big wedding, make 0:03:09.720 --> 0:03:12.040 this even the Wall Street Journal will slip into the Daily. 0:03:12.080 --> 0:03:13.840 Mom My God, I spill the secrets. 0:03:13.919 --> 0:03:15.520 They were an excited to do so. 0:03:16.440 --> 0:03:18.959 I was a bit skeptical about opening the show with 0:03:19.080 --> 0:03:23.040 talking about the Sanchez Bezos wedding, perhaps because I'm a 0:03:23.040 --> 0:03:25.160 brit and we've had quite a lot of royal wedding 0:03:25.160 --> 0:03:27.720 action in the last decade. But you pointed me to 0:03:27.720 --> 0:03:30.000 something in the Wall Street Journal piece that just felt 0:03:30.040 --> 0:03:32.320 so tech stuff. I couldn't wait him more about it. 0:03:32.480 --> 0:03:34.720 And this is so me. This is where the sort 0:03:34.760 --> 0:03:38.360 of intersection between my nerdiness and my interest in celebrity 0:03:38.360 --> 0:03:40.640 gossip intersects. I was reading the piece and I'm like, 0:03:40.640 --> 0:03:43.320 this is a great piece, But there was a certain 0:03:44.480 --> 0:03:47.920 thing that was mentioned. This luxury event designer said he 0:03:47.960 --> 0:03:53.280 had recently commissioned hold your Hat, a five hundred thousand, 0:03:53.640 --> 0:03:58.160 half a million dollar holograph of a bride's deceased grandfather 0:03:58.480 --> 0:04:01.640 to share well wishes to the that's the cost of 0:04:01.920 --> 0:04:05.360 like no percent of weddings is half a million dollars. 0:04:05.200 --> 0:04:06.560 Or no percent of Besos' wedding. 0:04:06.640 --> 0:04:07.560 That's right, that's right. 0:04:07.960 --> 0:04:11.280 You know it's funny because obviously there's this trend where 0:04:11.440 --> 0:04:14.400 rich people want to live forever. I guess the like 0:04:14.640 --> 0:04:18.320 this follow up trend to that is reanimating your your 0:04:18.360 --> 0:04:20.719 dead relatives for the for the cool sum of half 0:04:20.720 --> 0:04:23.520 a million dollars to deliver a toast at the wedding. 0:04:23.760 --> 0:04:25.599 It does remind me. You remember of that movie Mulan the 0:04:25.640 --> 0:04:29.640 Disney Words Drive when the ancestors reanimated to wish her 0:04:29.640 --> 0:04:32.000 well in her quest. Yes, this is like being if 0:04:32.000 --> 0:04:34.280 you're a billionaire, you can you can your wedding can 0:04:34.360 --> 0:04:36.520 feel like the opening sequence of Mulan. 0:04:36.640 --> 0:04:38.440 When you're a billionaire, you can watch a movie like 0:04:38.480 --> 0:04:40.000 Mulan and say I want that at my wedding. 0:04:40.080 --> 0:04:42.840 Please bring out her to us all remember that? Yes, 0:04:42.960 --> 0:04:46.200 of course, so you've now been subjected to my singing, 0:04:46.440 --> 0:04:50.440 which is appropriate. Turn to our next story, which is 0:04:50.680 --> 0:04:54.719 about the music industry and how the industry is building 0:04:54.760 --> 0:04:57.240 tech to make AI music more traceable. 0:04:57.839 --> 0:05:00.200 All right, so this is interesting because I do you 0:05:00.279 --> 0:05:04.200 think music production is one of the most evolved uses 0:05:04.200 --> 0:05:08.080 of AI. It duped me in twenty twenty three with 0:05:08.200 --> 0:05:09.040 fake Drake. 0:05:09.160 --> 0:05:12.120 Fake Drake. I was actually hoping to tell you about 0:05:12.160 --> 0:05:15.400 fake Drake and play that classic twenty twenty three banger 0:05:15.440 --> 0:05:17.080 and test you on whether or not you knew it 0:05:17.160 --> 0:05:21.120 was a fake or not. But let's play it nonetheless. 0:05:22.520 --> 0:05:26.520 Like it's fake question. 0:05:27.120 --> 0:05:30.360 But that was good. I came when my eggs like 0:05:30.520 --> 0:05:31.160 the lean. 0:05:31.240 --> 0:05:35.040 Honestly though it's very good. 0:05:34.720 --> 0:05:37.160 I love she need. 0:05:39.520 --> 0:05:41.640 Given. You know, if I hadn't known it was AI, 0:05:41.800 --> 0:05:44.120 I probably wouldn't have second guessed it at all. 0:05:44.360 --> 0:05:47.760 It really spooked people, especially the music industry, because if 0:05:47.800 --> 0:05:50.520 people didn't know, and worst of all, if people didn't 0:05:50.520 --> 0:05:53.680 care that this was a fake track, what could that 0:05:53.720 --> 0:05:54.719 mean for the whole business? 0:05:55.120 --> 0:05:57.840 Yeah, you know, the music business has already been battered 0:05:57.920 --> 0:06:02.200 so much in reaction first to piracy and illegal downloads, 0:06:02.240 --> 0:06:05.599 and now the popularity of streaming platforms like Spotify. 0:06:05.839 --> 0:06:09.000 Yeah, and this is another disruption, and potentially an unmanageable one, 0:06:09.200 --> 0:06:12.760 because if AI music is uploaded to streaming platforms and 0:06:12.800 --> 0:06:15.760 it's not labeled as being made by AI, the streaming 0:06:15.760 --> 0:06:19.240 platforms themselves and of course listeners might not be any 0:06:19.320 --> 0:06:19.799 the wiser. 0:06:20.040 --> 0:06:21.839 But is this actually like a problem? 0:06:22.120 --> 0:06:22.360 You know? 0:06:22.520 --> 0:06:25.320 I wonder how much AI music is actually being like 0:06:25.480 --> 0:06:27.600 dumped onto streaming platforms. 0:06:28.200 --> 0:06:33.680 You wonder, And I searched the web. According to the Verge, 0:06:33.680 --> 0:06:36.960 who spoke to the French streaming platform Deza Desa, said 0:06:37.000 --> 0:06:39.600 that as of this April, roughly twenty percent of new 0:06:39.680 --> 0:06:43.000 uploads every day were fully AI generated, so about twenty 0:06:43.080 --> 0:06:45.600 thousand tracks a day and I think the other important 0:06:45.640 --> 0:06:48.280 point here is that AI is obviously trained on what 0:06:48.320 --> 0:06:51.240 you feed it, so these tracks probably more often than 0:06:51.279 --> 0:06:55.520 not imitating musical ideas that perhaps should be licensed before 0:06:55.600 --> 0:06:57.400 the song is widely distributed. 0:06:57.640 --> 0:06:59.080 So it's kind of a money problem. 0:06:59.160 --> 0:07:02.839 Well, yes, the money problem. It's also a fundamental human 0:07:02.920 --> 0:07:06.360 creativity in the age of AI problem, but the money 0:07:06.360 --> 0:07:09.560 problem makes it more urgent capitalism to the rescue of 0:07:09.560 --> 0:07:12.560 the music business. Basically, there are multiple services now being 0:07:12.600 --> 0:07:15.960 developed that can be integrated into streaming platform structures to 0:07:16.040 --> 0:07:20.480 analyze one if uploaded tracks are AI, and two if 0:07:20.480 --> 0:07:22.960 they contain so called protected elements. 0:07:23.200 --> 0:07:24.520 So how does that work. 0:07:24.720 --> 0:07:26.400 There's one product in particular I want to tell you about, 0:07:26.400 --> 0:07:29.360 which is called trace ID, and it's marketed as an 0:07:29.440 --> 0:07:33.800 AI rights management platform. Basically, the software breaks songs into 0:07:33.840 --> 0:07:37.600 stems from vocal tone to melodic phrasing in order to 0:07:37.600 --> 0:07:40.720 better detect mimicry, and that means that the rights holders 0:07:40.800 --> 0:07:43.400 or the platforms can then know if a track needs 0:07:43.440 --> 0:07:46.200 to be licensed and paid for before it's released. 0:07:46.640 --> 0:07:50.520 So that is great for the industry, But what about me, Like, 0:07:51.160 --> 0:07:53.880 you know, I care about infringement, but I'm not like 0:07:54.000 --> 0:07:57.200 obsessed with it, and the thing that I'm really concerned 0:07:57.240 --> 0:07:58.520 with is, like, I want to know if the song 0:07:58.520 --> 0:07:59.640 that I'm listening to is real. 0:08:00.200 --> 0:08:02.920 Taking a step back and not wanting to get too 0:08:03.040 --> 0:08:06.280 philosophical here, but it kind of raises these other questions 0:08:06.320 --> 0:08:10.720 about what AI music actually is because obviously you have 0:08:10.800 --> 0:08:14.520 like fully generative AI generated tracks that are basically have 0:08:14.640 --> 0:08:17.320 no human input. But then a lot of like normal 0:08:17.400 --> 0:08:21.080 musicians as part of their production workflow, use AI tools 0:08:21.080 --> 0:08:23.080 in fact, just as we do right like we use 0:08:23.120 --> 0:08:25.600 AII editing software, we sort of use it in research. 0:08:25.680 --> 0:08:28.720 So like there's a kind of philosophical conundrum about what 0:08:28.760 --> 0:08:31.840 AI music actually is. But these tools we're talking about 0:08:31.880 --> 0:08:36.720 today are really about detecting like fully synthesized full AI tracks. 0:08:37.160 --> 0:08:41.000 In addition to these externally developed products like trace id, 0:08:41.120 --> 0:08:45.280 these streaming platforms are also internally developing tools to scan 0:08:45.480 --> 0:08:48.680 uploaded music and then if they detect a concentration of 0:08:48.679 --> 0:08:52.920 synthetic elements, they can reduce the visibility of AI generated 0:08:52.960 --> 0:08:57.199 tracks in both their algorithmic and also their editorial recommendations. 0:08:57.480 --> 0:09:00.000 Yeah, I've actually seen people on the internet can play 0:09:00.360 --> 0:09:04.000 that their release Radar playlist on Spotify is filled with 0:09:04.040 --> 0:09:06.600 what they suspect to be AI music. This has not 0:09:06.840 --> 0:09:09.880 happened to me personally, but like, it does take a 0:09:09.880 --> 0:09:12.720 lot of work to then go to like an artist 0:09:12.760 --> 0:09:16.679 page and see that someone has no listeners or followers 0:09:16.760 --> 0:09:18.439 or only one or two tracks, you know, And I 0:09:18.760 --> 0:09:21.880 guess there's hope that the actual places that are hosting 0:09:21.880 --> 0:09:24.880 this music would would label it, label it. Yeah. 0:09:24.960 --> 0:09:27.240 Well, the other thing is they might have an incentive 0:09:27.280 --> 0:09:29.800 to do so, because there's this emerging body of research 0:09:29.880 --> 0:09:34.440 that suggests when people feel that interacting with AI generated content, 0:09:34.880 --> 0:09:38.280 that should become less engaged. There's this marketing publication called 0:09:38.320 --> 0:09:41.120 The Drum which reported that more than fifty percent of 0:09:41.120 --> 0:09:44.319 people check out if they believe content is AI generated. 0:09:44.679 --> 0:09:46.959 So again there's another business intentive for the platforms to 0:09:47.000 --> 0:09:47.800 solve this problem. 0:09:48.280 --> 0:09:52.240 And speaking of AI generated content and interesting research, you 0:09:52.360 --> 0:09:55.359 must have seen this the recent MIT study on CHATGPT 0:09:55.559 --> 0:10:00.120 and critical thinking. I actually really liked the terminology that 0:10:00.160 --> 0:10:03.920 they used in this research. The paper was titled your 0:10:04.000 --> 0:10:07.000 Brain on Chat GPT accumulation. And this is what I 0:10:07.000 --> 0:10:11.360 love of cognitive debt. When using an AI assistant for 0:10:11.520 --> 0:10:12.920 essay writing task. 0:10:12.880 --> 0:10:15.679 I love the idea of accumulating cognitive debt. That that 0:10:15.800 --> 0:10:18.840 is such a good phrase and so familiar. I mean, 0:10:18.880 --> 0:10:21.520 I feel like scrolling on my phone. I mean most 0:10:21.520 --> 0:10:23.360 of what I do is accumulate cognitive debt. 0:10:23.480 --> 0:10:26.600 We are just sacks of cognitive debt that we really 0:10:26.600 --> 0:10:29.040 are right now. I mean people have to worry about, 0:10:29.320 --> 0:10:31.240 you know, student debt, They have to worry about other 0:10:31.280 --> 0:10:34.520 kinds of debt. Now we have to worry about cognitive debt. 0:10:34.720 --> 0:10:37.400 I saw a lot of action about this research all 0:10:37.440 --> 0:10:38.559 over my LinkedIn. 0:10:38.400 --> 0:10:40.800 The first place I actually encountered it was in like 0:10:40.920 --> 0:10:43.720 meme format, Like it was like a slideshow. 0:10:43.880 --> 0:10:45.560 There are pictures of their brains, right do lighting up 0:10:45.600 --> 0:10:46.520 in des exactly like. 0:10:46.559 --> 0:10:49.319 I certainly did not see this on like a verified 0:10:49.880 --> 0:10:52.600 news platform. I saw it like on a meme account 0:10:52.640 --> 0:10:56.040 that was like, yeah, your brain's getting worse using chat GPT. 0:10:56.720 --> 0:10:59.680 Before I explain the experiment, there are a few caveats 0:10:59.720 --> 0:11:04.080 that the researchers themselves are eager to share. First, the 0:11:04.120 --> 0:11:08.040 study only has fifty four subjects, which is a relatively 0:11:08.040 --> 0:11:11.600 small sample. And second, and I think the least surprising, 0:11:11.679 --> 0:11:13.760 is that this study has yet to be peer reviewed 0:11:14.160 --> 0:11:17.760 by anybody other than our peers on LinkedIn and Instagram 0:11:17.760 --> 0:11:18.360 and every. 0:11:18.320 --> 0:11:20.880 The interinet had. 0:11:21.160 --> 0:11:24.280 They really did. You know, it's a very buzzy concept, 0:11:24.559 --> 0:11:28.040 this tool chat GBT that we all are using, maybe 0:11:28.120 --> 0:11:31.920 eroding our own ability to think critically. And that's especially 0:11:31.960 --> 0:11:33.920 alarming when you think of what this could mean for 0:11:34.720 --> 0:11:38.800 not my brain, but developing brains. Mine has been developed, 0:11:38.520 --> 0:11:42.040 the trade has left the station. But you know, as 0:11:42.080 --> 0:11:44.800 we've talked about many many times on this podcast, AI 0:11:44.880 --> 0:11:48.320 companies are really marketing themselves and I mean chat GBT, 0:11:48.480 --> 0:11:53.520 especially to college students, and students are using chatbots like quite. 0:11:53.240 --> 0:11:55.800 A bit, quite a bit to assist. 0:11:55.440 --> 0:11:57.760 With or even do their homework. You know. The paper's 0:11:57.800 --> 0:12:01.000 main author felt like her findings were alarming enough and 0:12:01.080 --> 0:12:03.840 people are adapting to life with AI so fast. It 0:12:04.040 --> 0:12:07.120 wasn't peer reviewed because waiting six to eight months might 0:12:07.160 --> 0:12:07.800 be too late. 0:12:08.280 --> 0:12:10.920 The paper's outh theory is a bone hype beast, that's right. 0:12:11.000 --> 0:12:12.719 This is so she's like, it's got to get out 0:12:12.800 --> 0:12:14.640 right now or it's going to be too late. 0:12:15.160 --> 0:12:17.840 But I mean, to be fair, it does sound pretty lombing. 0:12:18.240 --> 0:12:22.040 Yes, there are some silver linings. This is what happened 0:12:22.360 --> 0:12:26.280 fifty four people ages eighteen to thirty nine from the 0:12:26.320 --> 0:12:30.240 Boston area were separated into three groups. They were all 0:12:30.280 --> 0:12:34.920 asked to write sat essays while an electro encephalogram or 0:12:35.000 --> 0:12:39.720 EEG measured their brain activity or more specifically, the tiny 0:12:39.800 --> 0:12:43.160 electrical signals produced by brain cells when they communicate. 0:12:43.600 --> 0:12:46.400 And so there were three groups. What was the difference 0:12:46.400 --> 0:12:47.120 between each group? 0:12:47.280 --> 0:12:51.280 So each group was writing these twenty minute satsays using 0:12:51.400 --> 0:12:55.400 slightly different tools. One group used open AI's chat GPT, 0:12:56.240 --> 0:13:00.040 another Google search engine, and the last group didn't use. 0:12:59.920 --> 0:13:01.800 It anything at all, nothing at all, i e. Just 0:13:01.880 --> 0:13:04.839 the human brain, just their brain for them. 0:13:04.960 --> 0:13:09.280 Unsurprisingly, the group that used chat GPT all delivered. This 0:13:09.360 --> 0:13:13.319 is my favorite part similar essays. English teachers were consulted 0:13:13.360 --> 0:13:17.679 and called them largely soulless. That is a sad indictment, 0:13:17.960 --> 0:13:19.720 it is, and that's actually a good word for it. 0:13:19.720 --> 0:13:23.000 Like if you read stuff that's written by chat GPT, 0:13:23.880 --> 0:13:28.079 you're like, there's something, there's someone here, that's what the 0:13:28.160 --> 0:13:28.720 lights are up. 0:13:28.840 --> 0:13:32.360 Fifty two percent of consumers per the drum check out. 0:13:32.480 --> 0:13:34.880 That feels like a low percent. But coming back to 0:13:34.880 --> 0:13:38.000 the group we were using chat GPT, the EEG picked 0:13:38.040 --> 0:13:41.440 up low executive control and attentional engagement in that group. 0:13:41.760 --> 0:13:46.600 That's contrasted with the brain only group, which showed shocker 0:13:46.840 --> 0:13:51.640 highest neural connectivity, especially in regions associated with creativity and memory. 0:13:51.920 --> 0:13:54.160 And what about the Google Search group, of whom I 0:13:54.320 --> 0:13:56.200 count myself a member of that cohort. 0:13:56.280 --> 0:14:00.280 They're what we'd call mid Their brains were definitely more 0:14:00.320 --> 0:14:03.280 active than the chat GPT groups. So they did this 0:14:03.360 --> 0:14:06.600 experiment three times, and then the researcher switched it up 0:14:06.600 --> 0:14:09.440 a bit. Each person was asked to rewrite one of 0:14:09.480 --> 0:14:13.440 their previous essays, but the chat GPT group could only 0:14:13.600 --> 0:14:17.880 use their brains, while the initial brain only group got 0:14:17.920 --> 0:14:21.360 access to chat GPT. The people who had started with 0:14:21.440 --> 0:14:25.640 chat GPT hardly remembered their own essays. 0:14:26.640 --> 0:14:27.560 Remember that correct. 0:14:27.800 --> 0:14:30.080 This reminds me of once I cheated on a physics exam. 0:14:30.200 --> 0:14:31.880 I cheated off a friend and I just wrote the 0:14:31.920 --> 0:14:34.040 answer without showing my work on a test, and my 0:14:34.080 --> 0:14:36.040 teacher was just like, how did you get that answer? 0:14:36.080 --> 0:14:37.560 It's right, but how'd you get it? And I said, 0:14:37.680 --> 0:14:43.720 I don't know it of course, So the people who 0:14:43.720 --> 0:14:46.520 had started with chat GPT hardly remembered their own essays. 0:14:47.240 --> 0:14:51.200 The EEG confirmed that barely any aspects of the writing 0:14:51.240 --> 0:14:56.520 process had integrated into people's memory networks. The brain only group, however, 0:14:57.400 --> 0:15:02.160 exhibited a significant increase in brain connected across all EEG 0:15:02.320 --> 0:15:05.680 frequency bands when they took a second stab at their 0:15:05.800 --> 0:15:09.960 essays using chat GBT to help, meaning people who use 0:15:10.040 --> 0:15:13.640 their brain displayed the most cognitive activity. 0:15:13.280 --> 0:15:15.000 But not just people use their brain, people who use 0:15:15.040 --> 0:15:17.840 their brain. And then a second time round pad it 0:15:17.880 --> 0:15:18.320 with chat. 0:15:18.240 --> 0:15:20.040 GPT people who use their brain first. 0:15:20.080 --> 0:15:23.720 And that's pretty interesting. I mean, it kind of stands 0:15:23.760 --> 0:15:26.480 to reason, which probably is again why it went viral, 0:15:26.520 --> 0:15:29.240 because it kind of confirms what we might think, which 0:15:29.320 --> 0:15:31.760 is that, like if I use my brain to come 0:15:31.840 --> 0:15:34.440 up with an idea and then use chat GPT to 0:15:34.560 --> 0:15:37.240 refine it and have a thought partner and a conversational 0:15:37.280 --> 0:15:40.200 partner to improve it, like that actually probably is more 0:15:40.240 --> 0:15:42.480 engaging for my brain than just coming up with an idea. 0:15:42.600 --> 0:15:44.640 If I ask chat GPT to count with an idea, 0:15:45.000 --> 0:15:46.920 that is extremely unengaging for my brain. 0:15:47.240 --> 0:15:50.360 Correct, that is the implication, And I just want to 0:15:50.360 --> 0:15:52.680 say one more time. This study has yet to be 0:15:52.680 --> 0:15:56.560 peer reviewed, and additional studies will likely be done. Again, 0:15:57.600 --> 0:16:00.400 the author felt very strongly that this should be released 0:16:00.440 --> 0:16:05.240 early as a warning. And again, imagine what this could 0:16:05.280 --> 0:16:08.520 mean for developing brains, meaning not us young people who 0:16:08.600 --> 0:16:11.680 are among the first to really adapt the technology anyway, 0:16:12.120 --> 0:16:15.440 using it as a generation tool rather than, as we might, 0:16:15.520 --> 0:16:18.760 a refinement tool. This is the real kicker. This one 0:16:18.800 --> 0:16:21.880 has a sense of the researcher has a sense of humor. 0:16:22.720 --> 0:16:26.920 So she assumed that people would use AI to summarize 0:16:26.960 --> 0:16:30.840 her paper, so she laid little AI traps all over it. 0:16:30.920 --> 0:16:33.200 What does it mean to lay little AI traps whatever 0:16:33.280 --> 0:16:33.760 a paper? 0:16:33.920 --> 0:16:36.680 She did a pretty cute thing, which is that she 0:16:36.840 --> 0:16:40.600 did things like instruct large language models to quote only 0:16:40.640 --> 0:16:44.200 read this table below, making it so if fed into 0:16:44.240 --> 0:16:47.320 an LM, only parts of the paper would be summarized. 0:16:47.560 --> 0:16:50.040 So there were these like hidden problems basically exactly. 0:16:50.840 --> 0:16:53.040 I think it was a very nice little sprinkle of 0:16:53.120 --> 0:16:54.040 human ingenuity. 0:17:00.800 --> 0:17:03.520 We've got a couple more headlines today, starting with the 0:17:03.560 --> 0:17:07.760 collaboration the whole tech industry is buzzing about, of course, 0:17:07.840 --> 0:17:11.000 I mean the one between open AI and Sir Johnny Ive, 0:17:11.280 --> 0:17:14.280 the designer behind many of Apple's iconic products like the 0:17:14.320 --> 0:17:17.000 iPod and the iPhone. And we know that Sam Altman 0:17:17.040 --> 0:17:20.440 and I are now collaborating on an AI hardware startup, 0:17:21.000 --> 0:17:24.080 and that's about all we know. Apparently they aren't making 0:17:24.080 --> 0:17:27.479 wearables or earbuds. The device will be pocket sized and 0:17:27.560 --> 0:17:30.639 screenless and will be a kind of interfaced layer with 0:17:30.720 --> 0:17:34.080 the world powered by AI, and Altman and Ive are 0:17:34.080 --> 0:17:36.760 being highly secretive about what the form factor will be, 0:17:37.200 --> 0:17:40.879 but people are buzzing. According to Semaphore, last week at 0:17:40.920 --> 0:17:45.040 the can Lion Advertising Festival, marketers were starting to freak 0:17:45.080 --> 0:17:47.679 out about where they would show video ads in a 0:17:47.720 --> 0:17:51.840 screenless world. But in the meantime, the highly anticipated project 0:17:51.880 --> 0:17:55.119 seems to have hit a roadblock. The startup was called 0:17:55.400 --> 0:17:58.639 Io the two letters I and O, and they've been 0:17:58.640 --> 0:18:01.720 pushing out marketing materials over the last few weeks. But 0:18:01.880 --> 0:18:04.640 now all mentions of Io have been scrubbed from Open 0:18:04.680 --> 0:18:08.080 AI's website and social media channels because it turns out 0:18:08.080 --> 0:18:12.160 there's a trademark lawsuit with another company called Io, which 0:18:12.200 --> 0:18:17.160 is spelled IYO, that's working on voice controlled AI devices. 0:18:17.640 --> 0:18:20.840 Sam Altman is called the lawsuit silly, but it's certainly 0:18:20.920 --> 0:18:24.960 drama in the valley, unlikely ultimately to derail him and 0:18:25.080 --> 0:18:27.280 Johnny Ive. Whatever they may end up with. 0:18:27.600 --> 0:18:30.120 I would watch Drama in the Valley on Bravo. It's 0:18:30.160 --> 0:18:33.240 the pitch yep here we go if you are a 0:18:33.359 --> 0:18:37.560 sucker for post apocalyptic content like me. There's now a 0:18:37.600 --> 0:18:40.000 follow up to the two thousand and one film twenty 0:18:40.000 --> 0:18:42.399 eight Days Later, which is out this month called twenty 0:18:42.400 --> 0:18:46.920 eight Years Later. The original film used lightweight, low resolution 0:18:47.080 --> 0:18:50.119 Canon digital cameras, a cutting edge technology back in the 0:18:50.160 --> 0:18:53.040 early arts, and for the follow up, director Danny Boyle, 0:18:53.080 --> 0:18:56.399 another brit Yeah, chose to stay small and nimble with 0:18:56.640 --> 0:19:00.240 the iPhone. He told Wired that the Apple device was 0:19:00.280 --> 0:19:03.680 the principal camera for the film, with some caveats. Boyle 0:19:03.720 --> 0:19:07.560 and his team ended up overriding the user friendly camera software. 0:19:08.119 --> 0:19:12.080 The iPhone's camera automatically focuses on whatever it assumes is 0:19:12.119 --> 0:19:14.840 the focus of your photo or video, but that's not 0:19:14.880 --> 0:19:17.560 always what you want in a movie, so they essentially 0:19:17.720 --> 0:19:22.840 hacked the iPhones to remove the auto focus. Also, most 0:19:22.840 --> 0:19:25.960 of the time it wasn't just a cinematographer holding an iPhone. 0:19:26.080 --> 0:19:29.920 The production used a massive rig that supported twenty iPhone, 0:19:29.960 --> 0:19:34.280 fifteen Promax cameras my dream all with special accessories. So 0:19:34.320 --> 0:19:37.719 that's twenty different angles on the action being filmed. 0:19:48.119 --> 0:19:49.560 We're going to take a quick break, but when we 0:19:49.600 --> 0:19:53.560 come back, there is a tentative ceasefire between Iran and Israel, 0:19:54.119 --> 0:20:10.439 but does that include cyberwarfare? Stay with us, Welcome back 0:20:10.440 --> 0:20:12.680 to tech Stuff. We want to spend some time talking 0:20:12.720 --> 0:20:16.119 about the conflict between Israel and Iran. Much of the 0:20:16.119 --> 0:20:19.320 battle has played out in public. Missile and drone attacks 0:20:19.359 --> 0:20:23.440 have caused mass casualties in major cities and hospitals. Military 0:20:23.480 --> 0:20:27.199 bases and nuclear sites have all been targeted. As of 0:20:27.240 --> 0:20:30.840 this taping on Wednesday morning, a ceasefire seems to be holding, 0:20:31.640 --> 0:20:34.000 but today we want to shed light on a murkier, 0:20:34.440 --> 0:20:38.119 often invisible act of warfare, and one that's likely to 0:20:38.160 --> 0:20:42.800 continue well after the missiles cease Hittelbos understand how the 0:20:42.840 --> 0:20:45.760 conflict between Israel and Iran is playing out in cyberspace 0:20:46.119 --> 0:20:49.280 and how it might ultimately affect the US is Maggie Miller, 0:20:49.400 --> 0:20:52.720 a cyber security reporter for Politico. Maggie, welcome to tech Stuff. 0:20:52.720 --> 0:20:54.760 Thank you so much for having me, either of. 0:20:54.680 --> 0:20:56.919 This taping or Wednesday morning. The ceasefire is an effect, 0:20:56.960 --> 0:20:59.200 but of course by Friday morning things may be different. 0:21:00.040 --> 0:21:02.440 Give US a bit of the background on both Israel 0:21:02.600 --> 0:21:05.760 and Iran, what their cyber capabilities are, and how much 0:21:05.800 --> 0:21:07.879 attention you were paying to those two countries in the 0:21:07.880 --> 0:21:10.160 cyber realm before this conflict started. 0:21:10.680 --> 0:21:14.400 So Israel has seen pretty widely as one of the 0:21:14.440 --> 0:21:16.560 most advanced in the world in terms. 0:21:16.359 --> 0:21:18.160 Of the cyber capabilities as a. 0:21:18.119 --> 0:21:21.879 Government, but also in terms of the industry experts that 0:21:21.880 --> 0:21:24.240 they have in their country. Tel Aviv is a hub 0:21:24.280 --> 0:21:27.640 of a lot of cybersecurity companies, so they have very 0:21:27.680 --> 0:21:31.200 formidable cyber capabilities. We've also seen them brought to bear. 0:21:31.320 --> 0:21:34.240 There was a cyber element involved in the explosion of 0:21:34.240 --> 0:21:38.679 the pagers used by Hesbolah operatives in Lebanon in recent 0:21:38.720 --> 0:21:42.320