www-wheresyoured-at-ntkfj5.txt (32969B)
1 [2] Ed Zitron's Where's Your Ed At 2 3 • [3]Home 4 • [4]About 5 6 [6]Log In [7]Subscribe 7 [8] Sign up [9] Sign in 8 9 • [12]Home 10 • [13]About 11 12 • [14]Sign up 13 14 [15] Log in [16] Subscribe 15 16 Subprime Intelligence 17 18 [17]Edward Zitron Feb 19, 2024 15 min read 19 20 Please scroll to the bottom for news on my next big project, Better Offline, 21 coming this Wednesday! 22 23 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24 25 Last week,[18] Sam Altman debuted OpenAI's "Sora," a text-to-video AI model 26 that turns strings of text into full-blown videos, much like how[19] OpenAI's 27 DALL-E turns text into images. These videos — which are usually no more than 60 28 seconds long — can at times seem impressive, until you notice a little detail 29 that breaks the entire facade, like[20] in this video where a cat wakes up its 30 owner, but the owner's arm appears to be part of the cushion and the cat's paw 31 explodes out of its arm like an amoeba. Reactions to Sora's AI generated videos 32 — and, indeed, the existence of the model itself — have ranged from breathless 33 hype to outright fear that this will be used to replace video producers, in 34 that it can created reality-adjacent videos that for a few seconds seem 35 remarkably real, especially in the case of[21] some of OpenAI's demo videos. 36 37 However, even in OpenAI's own hand-picked Sora outputs you'll find weird little 38 things that shatter the illusion, where[22] a woman's legs awkwardly shuffle 39 then somehow switch sides as she walks (30 seconds) or[23] blobs of people 40 merge into each other. These are, on some level, remarkable technological 41 achievements, until you consider what they are for and what they might do — a 42 problem that seems to run through the fabric of AI. 43 44 We're just over a year into the existence (and proliferation) of ChatGPT, 45 DALL-E, and other image generators, and despite the obvious (and reasonable) 46 fear that these products will continue to erode the foundations of the already 47 unstable economies of the creative arts, we keep running into the problem that 48 these things are interesting, surprising, but not particularly useful for 49 anything. 50 51 Sora's outputs can mimic real-life objects in a genuinely chilling way, but its 52 outputs — like DALL-E, like ChatGPT — are marred by the fact that these models 53 do not actually know anything.[24] They do not know how many arms a monkey has, 54 as these models do not "know" anything. Sora generates responses based on the 55 data that it has been trained upon, which results in content that is reality- 56 adjacent, but not actually realistic. This is why, despite shoveling billions 57 of dollars and likely petabytes of data into their models, generative AI models 58 still fail to get the basic details of images right,[25] like fingers or eyes, 59 or tools. 60 61 These models are not saying "I shall now draw a monkey," they are saying "I 62 have been asked for something called a monkey, I will now draw on my dataset to 63 generate what is most likely a monkey." These things are not "learning," or 64 "understanding," or even "intelligent" — they're giant math machines that, 65 while impressive at first, can never assail the limits of a technology that 66 doesn't actually know anything. 67 68 Despite what fantasists may tell you, these are not "kinks" to work out of 69 artificial intelligence models — these are the hard limits, the restraints that 70 come when you try to mimic knowledge with mathematics. You cannot "fix" 71 hallucinations (the times when a model authoritatively tells you something that 72 isn't true, or creates a picture of something that isn't right), because these 73 models are predicting things based off of tags in a dataset, which it might be 74 able to do well but can never do so flawlessly or reliably. 75 76 This is a problem that dramatically limits how much one can rely on generative 77 AI, and it's one that compounds severely with the complexity of what you're 78 asking it to do. Words can be copy-pasted and edited, and citations can be 79 checked. Images, however, are much tougher to edit, and videos are an entirely 80 different beast, especially if you're generating lifelike humans or animals. 81 While Sora is interesting and potentially quite scary to filmmakers, it's 82 important to consider some practical questions, like "how can someone actually 83 make something useful out of this?" and "how do I get this model to do the same 84 thing every time without fail?" While an error in a 30-second-long clip might 85 be something you might miss, once you see one of these strange visual 86 hallucinations it's impossible to ignore them. The assumption is that audiences 87 are stupid, and ignorant, and "just won't care," and I firmly disagree — I 88 think regular people will find this stuff deeply offensive. 89 90 I believe artificial intelligence companies deeply underestimate how perfect 91 the things around us are, and how deeply we base our understanding and 92 acceptance of the world on knowledge and context. People generally have four 93 fingers and a thumb on each hand, hammers have a handle made of wood and a head 94 made of metal, and monkeys have two legs and two arms. The text on the sign of 95 a store generally has a name and a series of words that describe it, or perhaps 96 its address and phone number. 97 98 These are simple concepts that we learn from the people and places we see as we 99 grow up, and what's very, very important to remember is that these are not 100 concepts that artificial intelligence models are aware of. When they see 20,000 101 pictures with signs in them, they understand that signs look a certain way, and 102 have some stuff on them, and then generate what's on the sign based on a user's 103 request and their dataset's tags that match that request. Even when a model is 104 fed exactly how a sign should be spelled out, it doesn't actually understand 105 what that information means or how it should be used, because the instructions 106 you are giving are based on your knowledge of signs and their contents, and the 107 model has no knowledge of any kind. 108 109 [26]AI fanatics are currently fantasizing over a world where they can put a few 110 sentences into a prompt and create an entire series of TV, unable to realize 111 that we are rapidly approaching the top of generative AI's[27] S-curve, where 112 after a period of rapid growth things begin to slow down dramatically. While 113 Sora and[28] other video generators like Pika may seem like the future (and are 114 capable of some impressive magic tricks), they are not particularly adept — 115 much like a lot of generative AI — at performing a particular task. Once you 116 get past the idea that you can now generate an almost-useful video that lasts 117 roughly a minute, one must consider the practical applications of this kind of 118 product. Even Microsoft struggled to find compelling use cases for their $7m AI 119 Superbowl commercial, and these use cases are even narrower once you realize 120 that generative video is so much more restrained by its hallucinations. Where 121 will Sora be useful? 122 123 Even if the costs weren't prohibitive, one cannot make a watchable movie, TV 124 show, or even commercial out of outputs that aren't consistent from clip to 125 clip, as even the smallest errors are outright repulsive to viewers. And as 126 I've suggested above, while these models might "improve," the billions of 127 dollars burned by OpenAI, Anthropic and Stability AI's models have found few 128 ways to mitigate the restrictions of an artificial intelligence that doesn't 129 have an intellect. I am also completely out of patience when it comes to being 130 told what it "will do" in the future. 131 132 Generative AI's greatest threat is that it is capable of creating a certain 133 kind of bland, generic content very quickly and cheaply. As I discussed in my 134 last newsletter, media entities are increasingly normalizing their content to 135 please search engine algorithms, and the jobs that involve pooling affiliate 136 links and answering where you can watch the Super Bowl are very much at risk. 137 The normalization of journalism — the consistent point to which many outlets 138 decide to write about the exact same thing — is a weak point that makes every 139 outlet "[29]exploring AI" that bit more scary, but the inevitable outcome is 140 that these models are not reliable enough to actually replace anyone, and those 141 that have experimented with doing so[30] have found themselves deeply 142 embarrassed. 143 144 Despite the frothy tales and visions of how generative artificial intelligence 145 will automate our entire existence, there's a distinct lack of practical 146 outputs that suggest that it is even capable of doing so. ChatGPT can spin up 147 piles of anodyne business copy, yet its outputs always require enough editing 148 that it's questionable how much time you've actually saved. Generative image 149 models are capable of creating cool-looking images that can replace generic 150 images that you might use in a project, but no matter how many different 151 prompts you use, they all kind of look the same, and that's even before you 152 notice how the minute details look off. Is a product that can only 153 sort-of-kind-of do something[31] really going to create trillions of dollars of 154 economic value? 155 156 I don't argue it will, at least not in such a way that anybody's lives will be 157 improved. 158 159 Shell Games 160 161 I believe we're reaching the upper limits about what generative AI can do[32] 162 and how accurate its outputs can be, and I believe that once reality catches up 163 with artificial intelligence's marketing, there will be a dramatic knock-on 164 effect that savages the entire tech industry.[33] A Wall Street Journal article 165 from mid-February told a worrying tale of OpenAI and Anthropic — the two 166 largest AI companies — racing to sell their generative AI systems despite the 167 prevalence of hallucinations, and how few answers they had for applications 168 that were highly regulated or dealt with highly sensitive data. When pressed on 169 the issue at a conference, Anthropic's Chief Science Officer Jared Kaplan was 170 only able to come up with one idea — that it would make a model capable of 171 saying "I don't know" to an answer, which in turn would create a situation 172 where the AI would err on the side of caution, restricting its willingness to 173 answer prompts at all. 174 175 The Journal seems unalarmed about multi-billion-dollar companies having very 176 few answers about the critical problem with their core product, but I'd argue 177 that a generative AI's inability to reliably generate stuff is an existential 178 threat that should have smothered these companies early in their lives. 179 180 And there are so many stories about how unreliable this technology is.[34] 181 British delivery firm DPD recently had to shut down their generative support 182 chatbot after a customer convinced it to write an insulting poem about the 183 company.[35] A Chevy dealership's ChatGPT-powered virtual assistant ended up 184 offering to sell a user a car for a dollar, and wrote a python script for 185 another.[36] Fortune reported a researcher's study into Large Language Models' 186 ability to understand SEC filings and found that many of them were regularly 187 either unable to answer or hallucinating incorrect information, with Meta's 188 Llama2 model getting 70% of the study's questions wrong.[37] A deeply foolish 189 lawyer relied on ChatGPT to cite cases in a motion, only to find that it cited 190 several non-existent pieces of case law. That lawyer — Steven A. Schwartz — was 191 fined $5,000 and ordered to i[38]nform each judge incorrectly cited as the 192 author of a non-existent verdict in the motion. In June of last year, OpenAI 193 was [39]sued for defamation in Georgia by a radio host who claimed that ChatGPT 194 generated a false legal complaint that accused him of embezzling money. 195 Microsoft destroyed MSN.com — a page that gets nearly two billion viewers a 196 month — by replacing its human staff with an artificial intelligence that[40] 197 posts made up stories about bigfoot and[41] stealing other outlets' stories and 198 still getting the details wrong. 199 200 It's also fair to question how many organizations are actually using it.[42] A 201 McKinsey report from August 2023 says that 55% of respondents' organizations 202 have adopted AI, yet only 23% of said respondents said that more than 5% of 203 their Earnings Before Interest (EBIT) was attributable to to their use of AI — 204 a similar number to their 2022 report, one which was published before 205 generative AI was widely available. In plain English, this means that while 206 generative AI is being shoved into plenty of places, it doesn't seem to be 207 generating organizations money. 208 209 There are indications that consumers have also lost interest. As [43]pointed 210 out by Alex Kantrowitz’ Big Technology newsletter, traffic to ChatGPT on both 211 mobile and web has started to stagnate, if not decline. In January 2024, 212 ChatGPT had 1.6 billion visits — 11% below the all-time peak of 1.8 billion. 213 This makes it only modestly more popular than Bing, which had 1.3 billion 214 unique visits during that period. On the mobile front, ChatGPT has an estimated 215 6.3 million US users — or 1.7 times less than the total of new Snapchat users 216 added during Q4 2023. 217 218 Tech's largest cash cow since the cloud computing boom of the 2000s is based on 219 a technology that is impossibly unreliable, a technology with a potent inverted 220 Midas touch that burns far more money than it makes.[44] According to The 221 Information, OpenAI made around $1.6 billion in revenue in 2023, and[45] 222 competitor Anthropic made $100 million, with the expectation they'd make $850 223 million in 2024. What these stories don't seem to discuss are whether these 224 companies are making a profit, likely because generative AI is a deeply 225 unprofitable product, demanding massive amounts of cloud computing power to the 226 point that OpenAI CEO Sam Altman is trying to raise[46] seven trillion dollars 227 to build chips to bring the costs down — though reports suggest that "the 228 figure represents the sum total of investments that participants in such a 229 venture round would need to make," which is basically the same thing. It’s 230 also, incidentally, a greater sum than the GDPs of France and the United 231 Kingdom combined. 232 233 While it's hard to tell precisely how much it’s losing, The Information[47] 234 reported in mid-2023 that OpenAI's losses "doubled" in 2022 to $540 million as 235 it developed ChatGPT, at a time when it wasn’t quite so demanding of cloud 236 computing resources.[48] Reports suggest that artificial intelligence companies 237 have worse margins than most software startups due to the vast cost of building 238 and maintaining their models, with gross margins in the 50-55% range — meaning 239 the money that it actually makes after incurring direct costs like power and 240 cloud compute. This figure is way below the 75-90% that modern software 241 companies have. In practical terms, this means that the raw infrastructure 242 firms — the companies that allow startups to integrate AI in the first place — 243 are not particularly healthy businesses, and they're taking home far less of 244 their money as actual revenue. 245 246 Luckily for them, Anthropic and OpenAI aren't really at risk, because they've 247 taken on an important part of the tech ecosystem — they're the tail of a very 248 hungry snake. 249 250 Turning On The Screw 251 252 During the imaginary panic of Sam Altman's ouster from OpenAI last year,[49] 253 Semafor reported that Microsoft's $10 billion investment was largely made up of 254 credits for their Azure cloud computing platform. In essence, Microsoft 255 "invested" $10 billion in money that OpenAI had to spend on Microsoft's 256 services, meaning that OpenAI would have to use Microsoft's "Azure" cloud 257 computing service to run ChatGPT.[50] When Google invested $2 billion in OpenAI 258 competitor Anthropic, it did so in tranches — $500 million up front and an 259 additional $1.5 billion over a non-specific period of time. Coincidentally, 260 this funding round took place only a few months after[51] Anthropic signed a 261 multi-year deal with Google Cloud worth $3 billion, locking them into Google's 262 compute platform in the process.[52] Amazon also invested $4 billion in 263 Anthropic, who agreed to a "long-term commitment" to provide Amazon Web 264 Services (Amazon's competitor to Microsoft Azure and Google Cloud) with early 265 access to their models — and Anthropic access to Amazon's AI-focused chips. 266 267 While Microsoft, Amazon, and Google have technically "invested" in these 268 companies, they've really created guaranteed revenue streams, investing money 269 to create customers that are effectively obliged to spend their investment 270 dollars on their own services. As the use of artificial intelligence grows, so 271 do these revenue streams, forcing almost every single dollar spent on AI into 272 the hands of a few trillion-dollar tech firms. 273 274 It's a contrived process with a fairly simple revenue stream. 275 276 In the case of an AI company (or a business that has jumped upon the AI 277 bandwagon), their website or app is integrated with OpenAI's ChatGPT or 278 Anthropic's Claude via their APIs. The company pays on a[53] per-token basis 279 for each input (request they make through their software) and output (thing 280 that the model does as a result). When these requests are made, ChatGPT, 281 Claude, or whatever model has to compute the result, which it does using 282 massive amounts of cloud computing — which is bought from the cloud provider 283 (say, Microsoft Azure or Google Cloud). As a result, every interaction with 284 ChatGPT or Claude is, on some level, guaranteed revenue for one of the big tech 285 firms. These were investments in the sense that money changed hands, but while 286 it did so, big tech put giant handcuffs on the wrists of the AI companies that 287 every startup has to use. 288 289 Admittedly, you could argue that the same situation is true for the 290 conventional Internet. Most websites are hosted by a third-party cloud 291 provider. If you visit a site that uses an external company to implement 292 functionality that would otherwise be too complicated to build themselves (like 293 auth, or payment processing, or banking integrations), it’s a sure bet those 294 companies are using Amazon, Microsoft, or Google for hosting. And so, without 295 even realizing it, our online activity benefits a handful of already-powerful 296 companies. The key difference is that, for the most part, people aren’t 297 locked-in and can walk, either to one of the other big players, or to a smaller 298 vendor like Rackspace or Linode. Moreover, the scale is different, and serving 299 a webpage will always cost less than processing a request sent to a generative 300 AI model. 301 302 These golden handcuffs have already led to massive swells of revenue for 303 Microsoft,[54] increasing by 30% in the last quarter alone thanks to the 304 increased usage of graphics processing units (GPUs) which have become essential 305 to the power-hungry demands of AI applications. Google's investment in 306 Anthropic was made in the hopes that it’d see a similar revenue multiplier, and 307 I'd argue Amazon's was made in the same vein — though it was too late to force 308 Anthropic to use AWS as their preferred vendor. 309 310 Big tech has turned the startup ecosystem into a giant goldmine, one that 311 guarantees that almost every dollar spent on any AI product is eventually 312 shared with one of a few multi-trillion dollar tech firms. And on some level, 313 it's become the savior of an ecosystem that hasn't had a new revenue-driving 314 industrial boondoggle this exciting since the Software-As-A-Service boom of the 315 2010s. Some might argue this is a situation where everybody wins — startups get 316 funded because they're able to do new things, venture capitalists make money 317 because their startups can actually get acquired or go public, and big tech 318 makes money because everybody is forced to pay them even more money by proxy. 319 320 I, however, have grave concerns. 321 322 As it stands, generative AI (and AI in general) may have some use. Yet even 323 with thousands of headlines, billions of dollars of investment, and trillions 324 of tokens run through various large language models, there are no essential 325 artificial intelligence use cases, and no killer apps outside of[55] 326 non-generative assistants like Alexa that are now having generative AI forced 327 into them for no apparent reason. I consider myself relatively tuned into the 328 tech ecosystem, and I read every single tech publication regularly, yet I'm 329 struggling to point to anything that generative AI has done other than reignite 330 the flames of venture capital. There are cool little app integrations,[56] 331 interesting things like live translation in Samsung devices, but these are 332 features, not applications. And if there are true industry-changing 333 possibilities waiting for us on the other side, I am yet to hear them outside 334 of the fan fiction of Silicon Valley hucksters. 335 336 This entire hype cycle feels specious, though not quite as specious as the 337 metaverse or cryptocurrency boom. Public companies are pumping their valuations 338 and executive salaries off the back of artificial intelligence hype, yet nobody 339 is saying the blatantly obvious — that this industry is deeply unprofitable and 340 yet to prove its worth.[57] Artificial intelligence is so demanding of 341 computing power that it may need as much electricity as an entire country,[58] 342 Microsoft and[59] Amazon are both investing billions to build even more data 343 centers to capture demand for an unproven product, and[60] Sam Altman of OpenAI 344 has said that the future of AI relies on an "energy breakthrough." 345 346 This industry is money-hungry, energy-hungry, and compute-hungry, yet it 347 doesn't seem to be doing anything to sustain these otherworldly financial and 348 infrastructural demands, other than the fact that people keep saying that 349 "artificial intelligence is the future." And[61] while some claim that AI can 350 help fight climate change, it's impossible to argue that "suddenly using more 351 and more power for a negligible return" is good for the environment. 352 353 And if this wasn't already worrying enough, one has to wonder what happens if 354 we face another economic panic, or if the hype dies down before OpenAI or 355 Anthropic discover a way to make a profit. As it stands, OpenAI and Anthropic 356 are heavily dependent on companies believing that they have to integrate AI 357 into their products, which will require these companies to be able to find ways 358 to integrate AI that users actually care about. And even if they manage to do 359 that, will they do so in a way that actually turns a profit? 360 361 If AI startups — by which I mean those companies integrating these models into 362 their apps — begin to falter, so will the only real revenue stream that these 363 companies have, making them more dependent on big tech to keep them alive. This 364 situation is only made more problematic by the fact that these models are 365 unprofitable, and Altman's desperation for a new chip company or energy 366 breakthrough suggests that they'll only become more unprofitable as they 367 generate more revenue. 368 369 I hope I am wrong. I hope that the bottom doesn't fall out of AI, and that the 370 startup ecosystem grows, and that this all becomes profitable and that 371 everything will be fine. 372 373 As it stands, I am terrified by how unstable this situation is and astonished 374 at how brazenly money and energy is being burned in pursuit of an unsustainable 375 future where big tech exerts more power over fledgling companies, and how 376 despite multiple industry collapses hinged upon unsustainable and unprofitable 377 businesses, Silicon Valley seems incapable of learning a single lesson. 378 379 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 380 381 Thanks for reading the newsletter. 382 383 This Wednesday - 2/21 - I'll be launching my iHeartRadio Podcast "Better 384 Offline," a weekly show exploring the tech industry’s growing influence over 385 society, and how startups, venture capitalists and big tech firms are looking 386 to change the future - for better or for worse. 387 388 I'd be so grateful if you'd subscribe. Here're the links: 389 390 Apple Podcasts:[62] https://podcasts.apple.com/us/podcast/better-offline/ 391 id1730587238 392 393 Spotify:[63] https://open.spotify.com/show/2dBPt1j2DoNij1kVdx8Ig6?si= 394 LY06yZufT7-syqE2OyHTYg 395 396 Pandora:[64] https://www.pandora.com/podcast/better-offline/PC:1001084695[65] 397 https://music.amazon.com/podcasts/a27a4803-938a-4aae-ab45-c28801d4722b/ 398 better-offline 399 400 Overcast:[66] https://overcast.fm/+BGz69vFSlo 401 402 iHeartRadio:[67] https://www.iheart.com/podcast/139-better-offline-150284547? 403 cmp=ios_share&sc=ios_social_share&pr=false&autoplay=true 404 405 Share 406 [68] [69] [70] [71] 407 About the author 408 [73] Edward Zitron 409 410 [74]Edward Zitron 411 412 [75]View all 413 Comments 414 415 Welcome to Where's Your Ed At! 416 417 Subscribe today. It's free. Please. 418 419 [76][ ] Subscribe 420 Great! Check your inbox and click the link. 421 Sorry, something went wrong. Please try again. 422 Ed Zitron's Where's Your Ed At 423 [78] [79] 424 425 • [80]Home 426 • [81]About 427 428 • [82]Sign up 429 430 ©2024 [83]Ed Zitron's Where's Your Ed At. Published with [84]Ghost & [85]Tuuli. 431 [86][Light ] 432 433 Great! You’ve successfully signed up. 434 435 Welcome back! You've successfully signed in. 436 437 You've successfully subscribed to Ed Zitron's Where's Your Ed At. 438 439 Your link has expired. 440 441 Success! Check your email for magic link to sign-in. 442 443 Success! Your billing info has been updated. 444 445 Your billing was not updated. 446 447 448 References: 449 450 [2] https://www.wheresyoured.at/ 451 [3] https://www.wheresyoured.at/ 452 [4] https://www.wheresyoured.at/about/ 453 [6] https://www.wheresyoured.at/signin/ 454 [7] https://www.wheresyoured.at/signup/ 455 [8] https://www.wheresyoured.at/signup/ 456 [9] https://www.wheresyoured.at/signin/ 457 [12] https://www.wheresyoured.at/ 458 [13] https://www.wheresyoured.at/about/ 459 [14] https://www.wheresyoured.at/sam-altman-fried/#/portal/ 460 [15] https://www.wheresyoured.at/signin/ 461 [16] https://www.wheresyoured.at/signup/ 462 [17] https://www.wheresyoured.at/author/edward/ 463 [18] https://www.nbcnews.com/tech/tech-news/openai-sora-video-artificial-intelligence-unveiled-rcna139065?ref=wheresyoured.at 464 [19] https://openai.com/dall-e-2?ref=wheresyoured.at 465 [20] https://twitter.com/tomwarren/status/1758203473881956689?ref=wheresyoured.at 466 [21] https://twitter.com/OpenAI/status/1758192961496760376?ref=wheresyoured.at 467 [22] https://x.com/OpenAI/status/1758192965703647443?s=20&ref=wheresyoured.at 468 [23] https://x.com/OpenAI/status/1758192957386342435?s=20&ref=wheresyoured.at 469 [24] https://x.com/edzitron/status/1758356234233840105?s=20&ref=wheresyoured.at 470 [25] https://www.digitaltrends.com/computing/5-things-ai-image-generators-still-struggle-with/?ref=wheresyoured.at 471 [26] https://twitter.com/mattturck/status/1758269761211777077?ref=wheresyoured.at 472 [27] https://www.npr.org/2021/10/06/1043822817/technology-brought-to-you-by-the-s-curve?ref=wheresyoured.at 473 [28] https://techcrunch.com/2023/11/28/pika-labs-which-is-building-ai-tools-to-generate-and-edit-videos-raises-55m/?ref=wheresyoured.at 474 [29] https://www.theverge.com/2024/1/30/24055718/new-york-times-generative-ai-machine-learning?ref=wheresyoured.at 475 [30] https://www.vox.com/technology/2023/7/18/23798164/gizmodo-ai-g-o-bot-stories-jalopnik-av-club-peter-kafka-media-column?ref=wheresyoured.at 476 [31] https://www.pbs.org/newshour/science/chatbots-can-make-things-up-can-we-fix-ais-hallucination-problem?ref=wheresyoured.at#:~:text=A%20lot%20is,some%20language%20component. 477 [32] https://www.scientificamerican.com/article/yes-ai-models-can-get-worse-over-time/?ref=wheresyoured.at 478 [33] https://www.wsj.com/articles/google-and-anthropic-are-selling-generative-ai-to-businesses-even-as-they-address-its-shortcomings-ff90d83d?ref=wheresyoured.at 479 [34] https://www.techradar.com/pro/a-customer-managed-to-get-the-dpd-ai-chatbot-to-swear-at-them-and-it-wasnt-even-that-hard?ref=wheresyoured.at 480 [35] https://www.businessinsider.com/car-dealership-chevrolet-chatbot-chatgpt-pranks-chevy-2023-12?ref=wheresyoured.at 481 [36] https://fortune.com/2023/12/21/chatgpt-understand-sec-filings-anthropic-meta-llama2-openai-finance-ai/?ref=wheresyoured.at 482 [37] https://www.nytimes.com/2023/06/08/nyregion/lawyer-chatgpt-sanctions.html?ref=wheresyoured.at 483 [38] https://www.cnbc.com/2023/06/22/judge-sanctions-lawyers-whose-ai-written-filing-contained-fake-citations.html?ref=wheresyoured.at 484 [39] https://www.theverge.com/2023/6/9/23755057/openai-chatgpt-false-information-defamation-lawsuit?ref=wheresyoured.at 485 [40] https://futurism.com/msn-is-publishing-more-fake-news?ref=wheresyoured.at 486 [41] https://futurism.com/msn-deletes-plagiarized-incoherent-ai-generated-articles-but-continues-publishing-more?ref=wheresyoured.at 487 [42] https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-in-2023-generative-ais-breakout-year?ref=wheresyoured.at 488 [43] https://www.bigtechnology.com/p/chatgpts-growth-is-flatlining?ref=wheresyoured.at 489 [44] https://www.reuters.com/technology/openai-annualized-revenue-tops-16-billion-information-2023-12-30/?ref=wheresyoured.at 490 [45] https://www.pymnts.com/artificial-intelligence-2/2023/report-anthropic-2024-revenue-could-approach-1-billion/?ref=wheresyoured.at#:~:text=Artificial%20intelligence%20startup%20Anthropic%20has,(AI)%20company's%20financial%20outlook. 491 [46] https://www.theinformation.com/articles/no-sam-altman-isnt-raising-trillions-of-dollars-for-chips?rc=kz8jh3&ref=wheresyoured.at 492 [47] https://www.businessinsider.com/openai-2022-losses-hit-540-million-as-chatgpt-costs-soared-2023-5?ref=wheresyoured.at 493 [48] https://techcrunch.com/2024/01/23/ai-startups-margins-low-valuations/?ref=wheresyoured.at 494 [49] https://www.semafor.com/article/11/18/2023/openai-has-received-just-a-fraction-of-microsofts-10-billion-investment?ref=wheresyoured.at 495 [50] https://www.cnbc.com/2023/10/27/google-commits-to-invest-2-billion-in-openai-competitor-anthropic.html?ref=wheresyoured.at 496 [51] https://www.wsj.com/tech/ai/google-commits-2-billion-in-funding-to-ai-startup-anthropic-db4d4c50?ref=wheresyoured.at#:~:text=Anthropic%20has%20also%20signed%20a%20multiyear%20deal%20with%20Google%20Cloud%20worth%20more%20than%20%243%20billion%2C%20said%20one%20person%20familiar%20with%20the%20matter.%20The%20contract%20was%20signed%20a%20few%20months%20before%20the%20new%20investment%2C%20the%20person%20said. 497 [52] https://techcrunch.com/2023/09/25/amazon-to-invest-up-to-4-billion-in-ai-startup-anthropic/?ref=wheresyoured.at 498 [53] https://openai.com/pricing?ref=wheresyoured.at 499 [54] https://www.cnbc.com/2024/02/12/microsoft-ai-growth-helping-azure-cloud-chip-away-at-amazons-lead.html?ref=wheresyoured.at 500 [55] https://www.vox.com/2023/9/23/23886163/google-microsoft-amazon-generative-ai-assistants?ref=wheresyoured.at 501 [56] https://www.zdnet.com/article/galaxy-ai-features-including-live-translation-are-headed-to-galaxy-buds/?ref=wheresyoured.at 502 [57] https://www.nytimes.com/2023/10/10/climate/ai-could-soon-need-as-much-electricity-as-an-entire-country.html?ref=wheresyoured.at 503 [58] https://datacentremagazine.com/technology-and-ai/microsoft-plans-to-invest-billions-into-ai-data-centres?ref=wheresyoured.at 504 [59] https://www.sdxcentral.com/articles/feature/aws-google-cloud-invest-in-data-center-expansion-sustainability/2024/01/?ref=wheresyoured.at 505 [60] https://www.reuters.com/technology/openai-ceo-altman-says-davos-future-ai-depends-energy-breakthrough-2024-01-16/?ref=wheresyoured.at 506 [61] https://www.bloomberg.com/news/articles/2023-12-14/ai-is-a-double-edged-sword-for-climate-change?ref=wheresyoured.at 507 [62] https://podcasts.apple.com/us/podcast/better-offline/id1730587238?ref=wheresyoured.at 508 [63] https://open.spotify.com/show/2dBPt1j2DoNij1kVdx8Ig6?si=LY06yZufT7-syqE2OyHTYg&ref=wheresyoured.at 509 [64] https://www.pandora.com/podcast/better-offline/PC:1001084695?ref=wheresyoured.at 510 [65] https://music.amazon.com/podcasts/a27a4803-938a-4aae-ab45-c28801d4722b/better-offline?ref=wheresyoured.at 511 [66] https://overcast.fm/+BGz69vFSlo?ref=wheresyoured.at 512 [67] https://www.iheart.com/podcast/139-better-offline-150284547?cmp=ios_share&sc=ios_social_share&pr=false&autoplay=true&ref=wheresyoured.at 513 [68] https://twitter.com/share?text=Subprime%20Intelligence&url=https://www.wheresyoured.at/sam-altman-fried/ 514 [69] https://www.facebook.com/sharer.php?u=https://www.wheresyoured.at/sam-altman-fried/ 515 [70] https://www.linkedin.com/shareArticle?mini=true&url=https://www.wheresyoured.at/sam-altman-fried/&title=Subprime%20Intelligence&summary=Subprime%20Intelligence 516 [71] mailto:?subject=Subprime%20Intelligence&body=https://www.wheresyoured.at/sam-altman-fried/%20Subprime%20Intelligence 517 [73] https://www.wheresyoured.at/author/edward/ 518 [74] https://www.wheresyoured.at/author/edward/ 519 [75] https://www.wheresyoured.at/author/edward/ 520 [78] https://twitter.com/edzitron 521 [79] https://www.wheresyoured.at/rss 522 [80] https://www.wheresyoured.at/ 523 [81] https://www.wheresyoured.at/about/ 524 [82] https://www.wheresyoured.at/sam-altman-fried/#/portal/ 525 [83] https://www.wheresyoured.at/ 526 [84] https://ghost.org/ 527 [85] https://brightthemes.com/themes/tuuli/