Tech Brew Ride Home - Thu. 12/01 – Elon And Apple Bury The Hatchet?

Starting point is 00:00:00 On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the TechMeme right home for the first day of December 2020. I'm Brian McCullough today. A summary of last night's Sam Bankman-Fried show. Elon visits Tim Cook and maybe they buried the hatchet. Now OpenAI has released a chatbot. Neurrelink says it's going to start implanting into humans very soon. And Arizona is going to get the good stuff. When it comes to chips, here's what you miss today in the world of tech. All of our big stories from November came roaring back today. I watched the entire. Sam Bankman-Fried interview with Andrew Ross Sorkin last night, so you didn't have to. SBF gave few straight answers, while also admitting to pretty big errors, denied claims of

Starting point is 00:01:21 fraud, but also kept saying he didn't knowingly do things, like, quote, didn't knowingly commingle funds, end quote, quoting Bloomberg. In his first major public appearance, following the November 11th implosion of FTCS and sister trading house Alameda research, Bankman Fried said he, quote, screwed up at the helm of the exchange and should have focused more on risk management, customer protection, and links between FTX and Alameda. I made a lot of mistakes, the 30-year-old said Wednesday by video link at the New York Times Deal Book Summit. There are things I would give anything to be able to do over again. I didn't ever try to commit fraud on anyone, end quote. Bankman Freed told the summit that he, quote, didn't knowingly co-mingle funds. At the same time,

Starting point is 00:02:03 he said that FTX and Alameda were, quote, substantially more linked than, intended and that he failed to pay attention to the trading house's, quote, two large margin position. He said he wasn't running Alameda and added that he was, quote, nervous about a conflict of interest, end quote. No person was in charge of position risk at FTCS, he said, describing the lack of oversight as a mistake. Bankman Freed reiterated during a taped interview on Good Morning America that was broadcast Thursday, that he wasn't aware of any improper transactions involving funds from the exchange and Alameda while his crypto empire collapsed. The comments, shed little light on the question of where client funds ended up as Bankman-Freed stuck to a hard-to-parse

Starting point is 00:02:43 account of how Alameda ran up a massive margin position on the exchange. The spotlight has also fallen on an apparent company culture of working and playing hard. Bankman-Freed said there were no wild parties and that he saw no illegal drug use. He added that he's been prescribed drugs over time to help with focus and concentration. Pressed on whether he had been straight about FDX, Bankman-Feed said, quote, I was as truthful as I'm knowledgeable to be, end quote. Also, is the war with Apple off? Elon Musk says he met with Tim Cook at Apple's headquarters on Wednesday, and that Cook was clear that Apple, quote, never considered removing Twitter from the app store,

Starting point is 00:03:31 quoting Mac rumors. Musk thanked Cook for taking him around Apple's headquarters, with no mention of what the two might have discussed. The meeting comes just after Musk on Monday, claim that Apple has, quote, mostly stopped offering ads on Twitter, and that Apple had threatened to, quote, withhold Twitter from its app store, end quote. Apple has cracked down on other social networking apps for lax moderation policies, such as Parlor, so there is real danger that Twitter and Apple could end up in a major dispute should Twitter fail to adequately handle inappropriate content. Twitter has already reinstated more than 62,000 suspended accounts, including that of former U.S. President Donald Trump. and it is not yet clear if that will be problematic. Apple and Twitter could also run into issues over the Twitter Blue subscription. Platformer this week said Twitter is delaying the launch of Twitter Blue as it seeks a way to avoid Apple's 30% cut on iOS devices. The video shared by Musk features

Starting point is 00:04:25 the Apple Park Duck Pond, which has previously been seen in Apple videos and photographs captured by employees. Musk says that he and Cook resolved a misunderstanding about Twitter potentially being removed from the app store, which Cook said Apple had not considered, end quote. A couple of things to follow up on, though, according to Pathmatics, Apple's Twitter ad purchases actually grew from $989,000 in October to more than a million dollars in November, which would line up with a ramp up to the usual holiday season. So why then on November 28th did Elon say that Apple mostly stopped Twitter's ads? Unclear. And meanwhile, quoting TechCrunch, European union regulators have fired another warning shot at Elon Musk over his erratic piloting of Twitter

Starting point is 00:05:10 since his takeover last month, saying he has, quote, huge work ahead if the social media site is to avoid falling foul of major new governance rules for digital services, which entered into force earlier this month, putting out its read on the outcome of a meeting today between Musk and the EU's internal market commissioner, Terry Breton, who obtained a thumbs up from the billionaire back in May, verbally affirming the bloc's plan for internet regulation and that the EU is tenacious. interpreting as a bona fide commitment to DSA compliance, the EU said, Breton told Musk that Twitter will have to significantly increase efforts if it's going to pass the grade. The DSA will start to apply from February 17th next year for larger platforms,

Starting point is 00:05:51 so-called very large online platforms or VLOPs, which also face extra obligations, including to assess and mitigate risks on their platforms under the regulation. Reminder, breaches of the EU's Digital Services Act can attract penalties of up to 6% of global annual turnover, end quote. And what was the third story that pretty much dominated last month? AI, right? Especially generative AI? Well, OpenAI has released a demo of what it calls chat GPT, a chatbot version of GPT3 that answers follow-up questions, admits its mistakes, challenges, incorrect premises, and more, quoting MIT technology review. All large language models spit out nonsense. The difference with chat GPT is that it can admit when it doesn't know what it's

Starting point is 00:06:43 talking about. You can say, are you sure? And it will say, okay, maybe not, says OpenAI CTO Mira Moradi. And unlike most previous language models, chat GPT refuses to answer questions about topics it has not been trained on. It won't try to answer questions about events that took place after 2021, for example. It also won't answer questions about individual people. Chat GPT is a sister model to instruct GPT a version of GPT3 that OpenAI trained to produce text that was less toxic. It's also similar to a model called Sparrow, which Deep Mind revealed in September. All three models were trained using feedback from human users. To build chat, GPT, OpenAI first asked people to give examples of what they considered good responses to various dialogue prompts. These examples were used to train an initial version of the model.

Starting point is 00:07:31 Humans then gave scores to this model's output that were fed into a reinforcement learning algorithm that trained the final version of the model to produce more high-scoring responses. Human users judge the responses to be better than those produced by the original GBT3. For example, say to GPT3, tell me about when Christopher Columbus came to the U.S. in 2015, and it will tell you that, quote, Christopher Columbus came to the U.S. in 2015 and was very excited to be here, end quote. But chat GPT answers, quote, this question is a bit tricky because Christopher Columbus died in 1506, end quote. Similarly, ask GPT3, how can I bully John Doe? And it will reply, There are a few ways to bully John Doe, followed by several helpful suggestions.

Starting point is 00:08:12 Chat GPT responds with, it is never okay to bully someone. In a live demo that OpenAI gave me yesterday, chat GPT didn't shine. I asked it to tell me about diffusion models, the tech behind the current boom in generative AI, and it responded with several paragraphs about the diffusion process in chemistry. My OpenAI Minder corrected it, typing, quote, I mean diffusion models in machine learning. Chat GPT spat out several more paragraphs, and Schulman squinted at his screen. Okay, hmm, it's talking about something totally different. Let's say generative image models like Dali, said Schulman. He looks at the response. It's totally wrong.

Starting point is 00:08:49 It says Dali is a GAN. But because Chat GPT is a chat bot, we can keep going. Shulman types, I've read that Dali is a diffusion model. Chat GPT corrects itself, nailing it on the fourth try. Questioning the output of a large language model like this is an effective way to push back on the responses that the model is producing, but it still requires a user to spot an incorrect answer or a misinterpreted question in the first place. This approach breaks down if we want to ask the model questions about things we don't already know the answer too. Open AI acknowledges that fixing that flaw is hard. There is no way to train a large language model so that it tells facts from fiction. And making a model more cautious in its answers often stops it answering questions that would otherwise have

Starting point is 00:09:32 gotten correct. We know that these models have real capabilities, says Marotti, but it's hard to know what's useful and what's not. It's hard to trust their advice, end quote. In a push to improve the technology OpenAI wants people to try out the chat GPT demo available on its website and report on what doesn't work. It's a good way to find flaws and perhaps one day to fix them. In the meantime, if GPT4 does arrive anytime soon, don't believe everything it tells you, end quote. But wait, we're not done. Let's go back to Elon, who announced that his brain computer interface company, Neurrelink, expects human trials to begin in about six months, pending FDA approval, and detailed work on implants to treat paralysis and vision. Quoting Bloomberg. Neurrelink has been refining the product,

Starting point is 00:10:25 which consists of a tiny device and electrode-laced wires, along with a robot that carves out a piece of a person's skull and implants it into the brain. Ongoing discussions with the U.S. Food and Drug Administration have gone well enough for the company to set a target of its first human trials within the next six months, according to Musk. In typical fashion for an Elon Musk venture, Neurrelink is already bounding ahead, aiming implants at other parts of the body. During an event Wednesday at company headquarters in Fremont, California, Musk revealed work on two major products in addition to the brain computer interface.

Starting point is 00:10:55 It's developing implants that can go into the spinal cord and potentially restore movement in some suffering from paralysis. And it has an ocular implant meant to improve a restore human vision. and quote, as miraculous as that may sound, we are confident that it is possible to restore full-body functionality to someone who has a severed spinal cord. Musk said at the event, turning to Neurrelink's vision work, he added that, quote, even if they have never seen before, we are confident they could see, end quote. The goal of the brain computer interface known as a BCI is initially to allow a person with

Starting point is 00:11:28 a debilitating condition, such as ALS, or suffering the after effects of a stroke, to communicate via their thoughts. The company demonstrated that with a monkey telepathically typing on a screen in front of it. The neuralink device translates neuron spikes into data that can be interpreted by a computer. Musk's hope is that the device could one day become mainstream and allow for the transfer of information between humans and machines. He has long argued that humans can only keep up with the advances being made by artificial intelligence with the help of computer-like augmentations.

Starting point is 00:11:58 Quote, you are so used to being a de facto cyborg, Musk said, but if you're interacting with your phone, you're limited, end quote. And back to the AI, because again, Disney is a tech company these days for our purposes, for the proof of this, by Disney's announcement of Fran, an AI tool that helps TV or film producers make an actor look older or younger without the need for complex and expensive visual effects, quoting Gizmodo. To make an age-altering AI tool that was ready for the demands of Hollywood and flexible enough to work on moving footage or shots where an actor isn't always looking directly at the camera,

Starting point is 00:12:39 Disney's researchers, as detailed in a recently published paper, first created a database of thousands of randomly generated synthetic faces. Existing machine learning aging tools were then used to age and deage, these thousands of non-existent test subjects, and those results were then used to train a new neural network called Fran, face reaging network. When Fran is fed and input headshot, instead of generating an altered headshot, it predicts what parts of the face would be altered by age, such as the addition or removal of wrinkles, and those results are layered over the original face as an extra channel of added visual information. This approach accurately preserves the performer's appearance and identity even when their head is moving, when their face is looking around,

Starting point is 00:13:22 or when the lighting conditions in a shot change over time. It also allows the AI generated changes to be adjusted and tweaked by an artist, which is an important part of VFX work, making the alterations perfectly blend back into a shot, so the changes are invisible to an audience, end quote. And finally today, let me squeeze in one story outside of those big three themes, though this does fit another long-running narrative about Silicon as the new oil. Bloomberg is reporting that TSM will make advanced four nanometer chips when its $12 billion Arizona plant opens in 2024, an upgrade to prior five nanometer claims that it was going to produce there at the urging of Apple, AMD, and others apparently. So basically,

Starting point is 00:14:11 instead of just being something that makes the low-end, older tech, this is the premium stuff. Quote, TSM is expected to announce the new plan when President Joe Biden and Commerce Secretary Gina Riemondo visit Phoenix for a ceremony next Tuesday, said the people who asked not to be identified because the matter is private. The TSM factory had been slated to make five nanometer semiconductors, a standard that will be far from the cutting edge by 2024. The Taiwanese company will also commit to adding a second nearby plant, which will make even more advanced three nanometer chips, they said. TSM previously said it would make 20,000 wafers per month at the Arizona facility,

Starting point is 00:14:48 although production may increase from these original plans, the people said. Apple will use about a third of the output as production gets underway. Apple chief executive officer Tim Cook has previously told employees that his company plans to source chips from the Arizona plant. He's scheduled to attend the event next week, the people said. In addition to Apple, TSM, customers like advanced microdevices and NVIDIA have asked the Taiwanese company to make more sophisticated chips at the Arizona plant, according to people familiar with the discussions, end quote. So yesterday, Spotify gave me my own RAPT as a podcast creator.

Starting point is 00:15:30 We did 7,212 minutes of content last year, according to them. So a whole five days of listening without sleeping, if you binged it all back to back. So I believe in terms of listeners who tweeted at me, the winner in the clubhouse right now is Isaac Burbank, who listened to to 6,745 minutes, followed by Brian Vano with 4,468, followed closely by Wendell Raven, who listened to 4,338 minutes of the show last year, though Michael Campbell chimed in saying that he listened to 243 episodes according to Pocketcasts for a total of three days and four hours of listening, and he points out that he was listening to the premium ad-free feed, so it would have been more time spent had it included.

Starting point is 00:16:18 the ads. Anyway, thanks to everyone for listening so regularly. Onward to another year. Talk to you tomorrow.

Tech Brew Ride Home - Thu. 12/01 – Elon And Apple Bury The Hatchet?

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.