Tech Brew Ride Home - Tue. 08/22 – Is Elon Tanking Things On Purpose?

Episode Date: August 22, 2023

Microsoft tries to appease His Majesty’s Regulators. The Arm IPO is a go. Why Nvidia continues to be huge even in China. A new AI translation model from Meta. I continue to wonder if Elon is tanking things on purpose. And a new social media platform built on top of X? Links: Microsoft to sell off Activision cloud gaming rights to Ubisoft in bid for UK approval (The Verge) Arm files for Nasdaq listing, as SoftBank aims to sell shares in chip designer it bought for $32 billion (CNBC) Why China remains hungry for AI chips despite US restrictions (FT) Meta releases an AI model that can transcribe and translate close to 100 languages (TechCrunch) X is planning to hide headlines from news links for ‘improved aesthetics’ (TechCrunch) Social Platform Friend.tech Gains 100K Users in Days Even in Depths of a Bear Market (CoinDesk) 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, August 22nd, 2023. I'm Brian McCullough today. Microsoft tries to appease his majesty's regulators. The arm IPO is a go. Why, Nvidia continues to be huge, even in China, a new AI translation model from meta. I continue to wonder if Elon is tanking things on purpose and a new social media platform built on top of X.

Here's what you miss today in the world of tech. In an attempt to appease UK regulators, Microsoft is restructuring its. Activision deal by agreeing to sell off Activision Cloud Gaming rights to Ubisoft. This has triggered a new UK-CMA investigation that could go until October 18th. Quoting the Verge, this restructured deal means that if Microsoft does close its proposed acquisition, then it will not be able to release Activision Blizzard games exclusively on Xbox Cloud Gaming. Microsoft won't be able to exclusively control the licensing terms of Activision Blizzard games on rival services either. Instead, Ubisoft will control the streaming rights to Activision Blizzard games outside of the EU

and licensed titles back to Microsoft to be included in Xbox Cloud Gaming. Ubisoft will also add Activision Blizzard games to its Ubisoft Plus multi-access subscription, which is available across PC Xbox, Amazon Luna, and on PlayStation via Ubisoft Plus Classics. The restructured transaction won't affect Microsoft's obligations to the European Commission, though. Microsoft has made several cloud gaming deals and EU regulators approved the Activision Blizzard deal thanks to a free license to consumers in EU countries that would allow them to stream via quote any cloud game streaming service of their choice, all current and future Activision Blizzard PC and console games that they have a license for. The CMA will now assess the reworked deal

over the coming weeks to deliver a decision by the October 18th deadline. This is not a green light. We will carefully and objectively assess the details of the the restructured deal and its impact on competition, including in light of third-party comments, says Sarah Cardell, chief executive of the CMA. Our goal has not changed. Any future decision on this new deal will ensure that the growing cloud gaming market continues to benefit from open and effective competition, driving innovation, and choice, end quote. So, is the deal finally going to happen? Is there an opening here, as Kellan Browning tweeted, the CMA focused its concerns around Microsoft attaining dominance in cloud gaming, so this could appease it. But also don't know

how significant this is in the overall scope of the acquisition. It really depends on if cloud gaming ends up taking off right now. It's negligible, end quote. And as our friend Martin SFP Bryant tweeted, quote, under Microsoft and Activision Blizzard's new deal to appease the UK CMA, Activision's cloud gaming rights outside the European economic area will be sold to Ubisoft. Well, that will make cloud gaming more complicated and confusing for consumers, end quote. It's official, SoftBanks Arm has filed to list on the NASDAQ and in their prospectus reported $524 million in net income on $2.68 billion in fiscal year, 23 revenue, down 1% year year over year. The company wants to go with the ticker Arm, ARM, quoting CNBC.

Arm with just under 6,000 employees, plays a pivotal role in the world of of consumer electronics designing the architecture of chips that are found in 99% of all smartphones, making it a key provider of technology to Apple, Google, and Qualcomm. The company was founded in 1990 as a joint venture between several companies and Apple to create a low power processor for battery-powered devices. It first went public in 1998 before being taken private in 2016 by SoftBank. But the company is also facing headwinds from a slowdown in demand for products like smartphones, which has hit chip firms across the board.

Arm's net sales fell 4.6% year-on-year in the second quarter, while the unit swung to a loss, according to SoftBank's earnings release. SoftBank's beleaguered Vision Fund, meanwhile, has racked up billions of dollars in losses of late due to tech bets that soured in a high-interest rate environment. In its filing, Arm made the case that its technology would be essential for AI applications, although it focuses on central processors, not the graphics processors that are required for creating big AI models. The CPU is vital in all AI systems, whether it is handling the AI workload entirely or in combination with a co-processor, such as a GPU or an NPU, Arm said in the filing. Arm identified X86, the instruction set used in Intel and AMD processors, as well as Risk 5,

an open-source instruction set backed by several big tech companies as sources of competition. The company said that its three largest customers accounted for 44% of the company's total revenue. The company's largest customer, Arm China, an independent entity, accounted for 24% of sales. Arm also said that Qualcomm, which it is currently suing over a licensing violation, accounted for 11% of sales, end quote. Indeed, Arm spent more than 3,500 words in its 330-page IPO prospectus explaining the risks to its business that it sees in China. Speaking of the chip situation vis-à-vis China, I marveled yesterday at the catbird seat in Vydea finds themselves in as sort of a global geopolitical linchpin, if you believe AI is the next big thing, which these nation states seem to believe it will be. But this is even true in China,

where Nvidia's A-800 and H-800 chips hobbled for the Chinese market to meet U.S. rules have seen huge demand, because they're more powerful than anything else the Chinese can get their hands on right now, quoting the Financial Times. The U.S. acted aggressive. last year to limit China's ability to develop artificial intelligence for military purposes, blocking the sale there of the most advanced U.S. chips used to train AI systems. But big advances in the chips used to develop generative AI have meant that the latest U.S. technology on sale in China is more powerful than anything available before. That is, despite the fact that the chips have been deliberately hobbled for the Chinese market to limit their capabilities, making them

less effective than products available elsewhere in the world. The result has been soaring Chinese orders for the latest advanced U.S. processors. China's leading internet companies have placed orders for $5 billion worth of chips from Nvidia, whose graphical processing units have become the workhorse for training large AI models. The impact of soaring global demand for Nvidia's products is likely to underpin the chipmaker's second quarter financial results due to be announced on Wednesday. Washington set a cap on the maximum processing speed of chips that could be sold in China, as well as the rate at which the chips can transfer data. A critical factor when it comes to training large AI models, a data-intensive job that requires connecting large numbers of chips together.

Nvidia responded by cutting the data transfer rate on its A-100 processors at the time its top-of-the-line GPUs, creating a new product for China called the A-800 that satisfied the export controls. This year, it is followed with data transfer limits on its H-100, a new and far more powerful processor that was specially designed to train large language models, creating a version called the H-800 for the Chinese market. Chipmaker has not disclosed the technical capabilities of the made-for-China processors, but computer makers have been open about the details. Lenovo, for instance, advertises servers containing H-800 chips that it says are identical in every way to H-100 sold elsewhere in the world, except that

they have a transfer rate of only 400 gigabytes per second. The longer training times raise costs since chips will need to consume more power, one of the biggest expenses with large models. However, even with these limits, the H-800-chips on sale and chips, China are more powerful than anything available anywhere else before this year, leading to the huge demand. The H-800 chips are five times faster than the A-100 chips that had been in Vida's most powerful GPUs, according to Patrick Moorhead, a U.S. chip analyst at More Insights and Strategy, end quote.

Meta has released Seamless M4T, an AI model that can translate and transcribe, nearly 100 languages across text and speech, as well as seamless align a translation data set. Quoting TechCrunch, meta claims that Seamless M4T represents a significant breakthrough in the field of AI-powered speech-to-speech and speech-to-text. Our single model provides on-demand translations that enable people who speak different languages to communicate more effectively. Meta writes on a blog post shared with TechCrunch, Seamless M4T implicitly recognizes the source languages without the need for a separate language identification model, end quote. Seamless M4T is something of a spiritual successor to meta's no language. left behind, a text-to-text machine translation model, and universal speech translator, one of the few direct speech-to-speech translation systems to support the Hokian language,

and it builds on massively multilingual speech, meta's framework that provides speech recognition, language identification, and speech synthesis tech across more than 1,100 languages. Meta isn't the only one, investing resources in developing sophisticated AI translation and transcription tools. Beyond the wealth of commercial services and open-source, model is already available from Amazon, Microsoft, OpenAI, and a number of startups. Google is creating what it calls the universal speech model, a part of the tech giant's larger effort to build a model that can understand the world's 1,000 most spoken languages. Mozilla, meanwhile, spearheaded common voice, one of the largest multi-language collections of voices for training automatic speech

recognition algorithms. But seamless M4T is among the more ambitious efforts to date to combine translation and transcription capabilities into a single model. In developing it, Meta said that it scraped publicly available text in the order of tens of billions of sentences and speech, four million hours worth from the web. In an interview with TechRunch, Juan Pino, a research scientist at Meta's AI Research Division and a contributor on the project, wouldn't reveal the exact sources of the data saying only that there was a variety of them. But Meta claims that the data it mined, which might contain personally identifiable information, the company admits, wasn't copyrighted and came primarily from open source or licensed sources. Whatever the case, Meta used the scrape

text and speech to create the training dataset for Seamless M4T called Seameless Align. Researchers aligned 443,000 hours of speech with texts and created 29,000 hours of speech-to-speech alignments, which taught Seamless M4T how to transcribe speech to text, translate text, generate speech from text and even translate words spoken in one language into words in another language. Meta claims that on an internal benchmark, Seamless M4T performed better against background noises and speaker variations in speech-to-text tasks compared to the current state-of-the-art speech transcription model. It attributes this to the rich combination of speech and text data in the training dataset, which Meta believes gives Seamless M4T a leg up over speech-only and text-only

models, end quote. Elon Musk says X will remove headlines. from link previews in posts with news articles so that the previews display only the lead image and the news sites URL, quoting TechCrunch. Currently, a Twitter card for a news article or a blog post shows the headline and summary text, just on the web, along with the header image in the preview card of a post. However, if the proposed change comes through, X will only show the image with a link in a post. That means if a publication or a blog doesn't post any accompanying text with the link, users will only see the link and the image for that article. A source told fortune that the update aims to reduce the height of a post to fit more posts on one screen.

They also said that Musk thinks that removing headlines from the preview card will reduce clickbait. However, with no preview card, publications, or blogs can simply write any text in citing users to click on the link, end quote. Again, I'm kind of joking about Elon trying everything he can think of to get people to quit X, but at this point, only, kind of kidding? Think about the thing where Elon suggested removing the block functionality. Let's say you're someone who has considered quitting X, but you've stuck around because you spent five years developing a block list, and now you feel fairly secure in terms of the defenses you built on X and you don't want to go somewhere else where you'd have to start from scratch. Well,

Elon wants to do away with that sort of sunk costs, which would free you up to leave. And let's say you're a news organization who maybe feels jerked around by the Musk regime's editorial decisions, but you've stayed on Twitter all this time because no alternative has similar traffic yet. You're still there because that's where the people are. Well, if Elon makes it harder for you to gin up traffic to your articles anyway, then what would be the point of sticking around? Finally today, friend. Tech, which lets people use crypto to buy shares of X accounts, has quickly hit 100,000 users per a database of wallet addresses, which, by the way, has raised privacy concerns, but, quoting CoinDesk.

The database listed briefly on GitHub and now withdrawn, apparently listed the crypto wallets, addresses, and linked X accounts. The growth correlates with the huge revenue made by the platform since its August 10th launch, with over 25 million in fees generated according to Defi Lama. A Dune Analytics dashboard corroborates the finding, pegging the number of unique users at at least 80,000 with 15,000 users added since Sunday. A slew of personalities outside of crypto circles on X, joined friends. friend. Tech, Richard Fays Banks Bank's Benxton, the second, co-founder of the influential e-sports community, Fays Clan, joined the platform late Sunday and saw his share price quickly

become among the most expensive. Shares of NBA player Grayson Allen also surged in the hours after joining the shares grant the holder's privileges, such as the ability to send private messages to the sellers. Meanwhile, friend.com is quickly going viral. The platform has made over 1.04 million in fees, set at 5% of the value of each transaction. the past 24 hours alone. That's banked the platform some $709,000 worth of ether in revenue, what the platform takes after paying out gas fees and other costs, data from Defi Lama shows, end quote. I needed a new heel for my shoes, so I decided to go to Morganville, which is what they called Shelbyville in those days. So I tied an onion to my belt, which was the style at the time.

Now to take the ferry cost to nickel, and in those days, Nichols had pictures of bumblebees on them. Give me five bees for a quarter, you'd say. Now, where were we? Oh yeah. The important thing was that I had an onion on my belt, which was the style at the time. Maybe instead of song lyrics, when I have nothing to say at the end of a show, because I never remember to save song lyrics for that purpose when I need them, maybe I'll just give you Simpsons quotes instead. My kids watch nothing but The Simpsons these days. They've been through all of the seasons twice at this point.

Talk to you tomorrow.