Tech Brew Ride Home - Tue. 10/31 – The New M3 Chips And Macs

Episode Date: October 31, 2023

All the headlines for last night’s Apple event, including interesting details about the new M3 chips. One year on from Elon purchasing it, what is X worth now? Is Nvidia gonna have to cancel all those chip orders from China? And are people crying chicken little about the threat of AI in order to do some regulatory capture? Links: Apple ‘Scary Fast’ Mac launch event: the 4 biggest announcements (The Verge) Apple officially unveils M3, M3 Pro, and M3 Max: 3 nanometer, Dynamic Caching GPU, more (9to5Mac) Apple M3 Pro Chip Has 25% Less Memory Bandwidth Than M1/M2 Pro (MacRumors) Nvidia’s $5 Billion of China Orders in Limbo After Latest U.S. Curbs (WSJ) X Says It Is Worth $19 Billion, Down From $44 Billion Last Year (NYTimes) Artists Lose First Round of Copyright Infringement Case Against AI Art Generators (TRH) Google Brain cofounder says Big Tech companies are lying about the risks of AI wiping out humanity because they want to dominate the market (Insider) Learn more about your ad choices. Visit megaphone.fm/adchoices

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 Tech meme right home for Tuesday, October 31st, 2023. I'm Brian McCullough today. All the headlines from last night's Apple event, including interesting details about the new M3 chips. One year on from Elon purchasing it, what is X worth now? Is Nvidia going to have to cancel all those chip orders from China? And are people crying chicken little about the threat of AI in order to do some old-fashioned regulatory capture? Here's what you miss today in the world of tech. The unusually timed Apple event was last night, and there were basically four big announcements. As I said to Alex Cantorwitz on the bonus episode, it was mostly about the new M3 chips,

but because of those, that allowed for the announcement of a new 24-inch iMac, new MacBook Pros, and an entry-level MacBook Pro that finally sends the touchbar to the great design cul-de-sac in the sky, quoting the verge. As expected, Apple's M3 chips took the spotlight during this month's event. The new lineup includes the M3, M3 Pro, and M3 Max chips, which Apple says mark the, quote, first personal computer chips made using the more efficient three nanometer process. In addition to offering a faster and more efficient CPU, the trio of chips comes with an updated GPU that supports ray tracing, mesh shading, and dynamic caching, a feature that optimizes the amount of memory the device uses during tasks. Apple's M3 chips offer up to 128 gigabytes of unified memory

with the most powerful M3 Max chip coming up with 92 billion transistors, a 40-core GPU and a 16-core CPU. The new 24-inch IMac is getting an M-3-flavored upgrade that Apple says offers two-time faster performance than its M-1-equipped predecessor. Along with the new chip, the refreshed IMA features a 4.5K retina display with more than 1 billion colors, support for Wi-Fi 6E, and a 1080P webcam. The iMac also offers up to 24 gigabytes of unified memory and comes in seven colors, green, yellow, orange, pink, purple, blue, and silver. There are also color-matched accessories that come with the iMac, but they still feature lightning connectors.

The 24-inch iMac costs $1,29 with an 8-core CPU or 1499 with a 10-core chip. You can pre-order it starting today with availability starting on November 7th. The iMac isn't the only device getting an M-3-powered upgrade. Apple also announced a new pair of 14- and 16-inch MacBook Pro models that come with either the M3 Pro chip or the higher-end M-3 Macs. Both laptops feature a mini-l-D display, a 1080P camera, a six-speaker surround system, 22 hours of battery life, and up to 128 gigabytes of RAM. They're also available in a space black finish with a new coating that's supposed to help reduce fingerprints, as well as a silver option. While the 14-inch MacBook Pro with an M-3,

3 Pro chip starts at $1,99. The 16-inch M-3-Pro model starts at 2499. These MacBook Pro models are available to pre-order today and will become available on November 7th. Alongside the M3 Pro and M-3-Max-equipped MacBook Pro models, Apple is releasing a cheaper 14-inch MacBook Pro that comes with the base M3 chip and starts at 1599. The device replaces the 13-inch MacBook Pro with an M2 chip that Apple released last year and offers performance that's up to 60% faster. The touchbar model is being discontinued, which means it's all

physical keys from here out. There are some drawbacks to this entry-level model, though. It features a meager 8-gibytes of RAM and comes in just silver and space gray variations. The black color is exclusive to the higher-end MacBook Pros. The device is available to pre-order today and officially launches on November 7th, end quote. More on those M3 chips. As Apple says, they're the first 3 nanometer chips in consumer computers, the first for Apple. They have a dynamically caching GPU, 15% faster performance cores over the M2, and 30% faster efficiency cores than the M1. And that's interesting. As Mac Rumors points out, Apple repeatedly compared the M3 to the M1 and not the M2 chip. Quote, based on the latest 3 nanometer technology and featuring all new GPU architecture,

the M3 series of chips is said to represent the fastest and most power. efficient evolution of Apple Silicon thus far. For example, the 14-inch and 16-inch MacBook Pro with M3 Pro chip is up to 40% faster than the 16-inch model with M1 Pro, according to Apple. However, according to Apple's own hardware specifications, the M3 Pro system on a chip features 150 GBS memory bandwidth compared to 200 GBS on the earlier M1 Pro and M2 Pro. As for M3 Max, Apple says it's capable of, quote, up to 400 GBS. which is because the scaled-down M-3-max with 14-core CPU and 30-core GPU has only 300 GBS of memory bandwidth, whereas the equivalent scaled-down M-2-max with 12-core CPU and 30-core GPU featured 400-GBS bandwidth,

just like its more powerful 12-core CPU, 38-core GPU iteration. Notably, Apple has also changed the core ratios of the higher-tier M-3-pro chip compared to its direct predecessor. The M3 Pro with 12-core CPU has six performance cores versus eight performance cores on the 12-core M2 Pro, and six efficiency cores versus four efficiency cores on the 12-core M2 Pro, while the GPU has 18 cores versus 19 on the equivalent M2 Pro chip. Additionally, while the M-3 chip's 16-core neural engine has the same number of cores as the one Apple featured in the 3-nometer-based A-17 Pro chip that debuted in the iPhone 15 Pro series in September, it's comparatively weaker on paper in terms of maximum achievable throughput,

which is measured in trillions of operations per second or tops. According to Apple, the M3 neural engine is capable of 18 tops, whereas the A17 Pro Neural Engine is capable of 35 tops. It's hard to say for certain, but it is possible that the iPhone 15 Pro requires a higher-performing neural engine for features like computational photography and face ID, whereas the M3 can compensate in other areas like machine learning by utilizing its additional GPU cores.

Taken together, it's presently unclear what real-world difference these changes make to M3 performance when pitted against Apple's equivalent precursor chips in various usage scenarios, especially given that the latest processors include the new dynamic caching memory allocation technology, which ensures that only the exact amount of memory needed is used for each task. This opaqueness is not helped by the fact that Apple advertises the power of the new M3 Pro and M3 Max chips by emphasizing comparisons to the M1 Pro and M1 Max. rather than the more recent M2 variants against which performance gains appear more modest. Hopefully, we will learn more in time when the first thoroughgoing third-party benchmarks become available, end quote.

Quoting Vadim Yuryev on Twitter. Fun fact, because Apple removed two P-cores on the M3 Pro chip and gave it two more E-cores, it's not much faster than the M2 Pro chip. M2 Pro was 16% faster than M1 Pro. Apple says M3 Pro is 20% faster. than M1 Pro, pretty sad, L.O.L. end quote. And quoting the Ginger Bill on Twitter. It looks like the M3 Pro is actually a downgrade from the M2 Pro in many ways for certain workloads. Why would you reduce the number of performance cores and memory bandwidth when it's called the pro

variant, end quote. Sources say U.S. rules may compel Nvidia to cancel those more than $5 billion worth of chip orders from China that we told you about recently, quoting the journal. The U.S. government told Nvidia in a letter last week that the new export controls on the sale of high-end chips to countries including China were instead effective immediately. China's biggest AI and cloud computing companies including Alibaba, TikTok owner ByteDance, and Baidu had made large orders for delivery next year, the people said. Orders from major Chinese companies for 2024 exceeded $5 billion, one of the people said.

A spokesman for Nvidia said the company has been working to allocate its advanced AI computing systems, which use graphics chips affected by the rules to customers in the U.S. and elsewhere, and is pursuing additional supply. These new export controls will not have a meaningful impact in the near term, the spokesperson said, end quote. According to a source X is giving employees restricted stock units at a $45 per share price, which would imply that, well, what used to be known as Twitter is worth less than half of what it was. was a year ago. Quoting the New York Times. X, the company formerly known as Twitter, handed out stock grants to employees on Monday that showed it was worth about $19 billion,

down about 55% from the $44 billion that Elon Musk paid to buy the firm a year ago, according to internal documents seen by the New York Times. Mr. Musk paid $54, $20 a share, to buy Twitter just over a year ago. The tech billionaire has since said he overpaid for the social network. In March, he wrote an email to workers that he believed the company was worth $20 billion, calling it an inverse startup. In the paperwork for the new stock grants, X said the equity would be offered at $45 a share in the form of restricted stock units, which employees can earn over time. Employees will still be paid in cash in the amount of $54.20 per share for any outstanding shares that were granted to them. Under previous management,

the company said, it's unclear why the share price has not dropped by the same percentage as the company's valuation, though X could have altered the amount of shares outstanding. earlier reported on the valuation, end quote. A U.S. judge has dismissed AI copyright infringement claims against mid-jurney and deviant art by three artists, but at the same time allowed a claim against stability AI to proceed, quoting the Hollywood Reporter. Notably, a claim for direct infringement against stability AI was allowed to proceed based on allegations the company used copyrighted images without permission to create stable diffusion. Stability has denied the contention that it's stored and incorporated those images into its AI system. It maintains that

training its model does not include wholesale copying of works, but rather involves development of parameters like lines, colors, shades, and other attributes associated with subjects and concepts from those works that collectively define what things look like. The issue, which may decide the case, remains contested. The litigation revolves around stability's stable diffusion, which is incorporated into the company's AI Image Generator Dream Studio. In this case, the artists will have to establish that their works were used to train the AI system. It's alleged that Deviant Arts Dream Up and Mid-Journey are powered by stable diffusion, a major hurdle artist's face, is that training datasets are largely a black box. In his dismissal of infringement claims,

the judge wrote that the plaintiff's theory is unclear as to whether there are copies of training images stored in stable diffusion that are utilized by Deviant Art and Mid-Jurney. He pointed to the defense's arguments that it's impossible for billions of images, quote, to be compressed into an active program like stable diffusion. The judge questioned whether Mid-Jurney and Deviant Art, which offers use of stable diffusion through their own apps and websites, can be liable for direct infringement if the AI system, quote, contains only algorithms and instructions that can be applied to the creation of images that include only a few elements of a copyrighted work.

The judge stressed the absence of allegations of the companies playing an affirmative role in the alleged infringement. Quote, plaintiffs need to clarify their theory against Mid-Journey. is it based on Mid Journey's use of stable diffusion, on Mid Journey's own independent use of training images to train the Mid Journey product, or both, end quote. Finally today, bringing another trend to your attention, I think I mentioned that at that AI conference earlier this month, the term de jour was open source. Everyone was talking about how they were embracing open source to bring new versions of AI to the four. So let me unpack

this. Let's imagine a scenario where the biggest AI models will always win because they're the biggest, have the most data, or the best trained, are just always the best, the most accurate or whatever. In that scenario, this is already a done deal. The companies owning the biggest models will win this space. But what you have to think about is they're not sharing the secret sauce of these models, the biggest players like Open AI. And because of that, maybe nobody can challenge their supremacy without huge costs because they can't reverse engineer the secret sauce. Thus, the move towards open-sourcing everything in order for a thousand flowers, if you will, to bloom. So what I'm saying is some folks are worried that

there are already incumbents here in this nascent AI space. Open AI, Anthropic, a couple others. This thing could already be an oligopoly before it even really got started. That's one of the reasons why meta is open-sourcing their models. If they're already behind in a field that is rapidly being captured by incumbents, open sourcing models would allow them a wedge to stay competitive. And as I said to some of the founders and engineers I spoke to at the conference, that's also why the VC class is pushing the open source narrative. VCs need a whole ecosystem of startups to rise up and bloom for this to be an investable space. If this new space is already closed off by the first movers, it's dead, at least for investors. Now, there's another angle,

and that is, well, if you open source the models, if anyone can spin up their own AI, then some people fear chaos, everything from terrorists to misinformation merchants to hackers, having access to some pretty powerful new tools. Thus, the push to regulate the models, as we discussed yesterday, with that slew of regulatory announcements. But the angle to that angle is this. Have you noticed that some of the people shouting the loudest for regulation are the so-called incumbents, the open AIs of this world? Being like, please, regulate the space, please regulate us. Well, some folks are worried that this is just another way the newly minted incumbents are trying to pull the drawbridge up on their newly minted moat. This stuff is so dangerous. Governments, you can't trust anybody but us with it. That would shut down any sort of disruption from below. The term for this is called regulatory capture. Now, obviously, I'm an investor in the space, so you should take everything I just laid out with a grain of salt or three, but I am trying to do what I always do on the show and present the contours of what is really happening in the Valley as I see it. So, this is all leading up to the news that

Andrew Eng, who taught Sam Altman at Stanford and co-founded Google Brain, said that big tech is, in his words, lying about AI extinction to trigger heavier regulation of rivals. Quoting Insider, Google Brain was a deep learning AI research team that merged with the deep division earlier this year. Andrew Eng, an adjunct professor at Stanford University who taught OpenAI CEO Sam Altman, told the Australian Financial Review that the biggest tech companies hope to trigger strict regulation with the, quote, bad idea that AI could make us go extinct. There are definitely large tech companies that would rather not have to try to compete with open source, so they're creating fear of AI leading to human extinction, he told the news outlet. It's been a weapon for lobbyists

to argue for legislation that would be very damaging to the open source community, end quote. governments around the world are looking to regulate AI, citing concerns over safety, potential job losses, and even the risk of human extinction. The European Union will likely be the first region to enforce oversight or regulation around generative AI. Eng said that the idea AI could wipe out humanity could lead to policy proposals that require licensing of AI, which risk crushing innovation. Any necessary AI regulation should be created thoughtfully, he added, end quote. This has led to a great deal of back and forth online in the AI community, but again, I present it to you in an attempt to give you a sense of all sides of this evolving debate. Quick request, I know that I have lots of people who work at Apple that listen to this show,

but are there any folks listening right now that work at Apple Podcasts itself? If so, can you get in touch with me at Brian at Techmeme.com? I have a question for you. Thanks in advance. Talk to you tomorrow.