Tech Brew Ride Home - Tue. 01/25 – Is The Nvidia/Arm Deal Off?

Episode Date: January 25, 2022

That already fraught tie-up between Nvidia and Arm increasingly looks like it’s running out of time. Google kills its cookie-killer. Twitter is developing a close friends feature. Meta’s new Machine Learning supercomputer. And why the situation in Ukraine might be a black swan event that is nigh. Sponsors: Do.co/trh Wix.com Links: Nvidia Quietly Prepares to Abandon $40 Billion Arm Bid (Bloomberg) Google kills off FLoC, replaces it with Topics (TechCrunch) Twitter's new Flock feature takes aim at Instagram Close Friends (Input) Meta has built an AI supercomputer it says will be world’s fastest by end of 2022 (The Verge) Hactivists say they hacked Belarus rail system to stop Russian military buildup (ArsTechnica) DHS warns of Russian cyberattack on US if it responds to Ukraine invasion (ABCNews) 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, January 25th, 2022. I'm Brian McCullough today. That already fraught tie-up between Nvidia and Arm increasingly looks like it's running out of time. Google kills its cookie killer. Twitter is developing a close friends feature, meta's new machine learning supercomputer,

and why the situation in Ukraine might be a Black Swan event that is nigh. Here's what you miss today in the world of tech. surprised, not surprised, sources are telling Bloomberg that Nvidia is planning to abandon its arm acquisition as regulators scrutinize the deal. Sources say that SoftBank is preparing an IPO for Arm if the deal doesn't happen, quote. Nvidia has told partners that it doesn't expect a transaction to close, according to one person who asked not to be identified because the discussions are private. SoftBank, meanwhile, is stepping up preparations for an arm initial public offering, As an alternative to the NVIDIA takeover, another person said,

the purchase poised to become the biggest semiconductor deal in history when it was announced in September 2020, has drawn a fierce backlash from regulators and the chip industry, including arms-owned customers. The U.S. Federal Trade Commission sued to stop the transaction in December, arguing that NVIDIA would become too powerful if it gained control over arms chip designs. The acquisition also faces resistance in China

where authorities are inclined to block the takeover if it wins approvals elsewhere, according to one person. But they don't expect it to get that far. Both Nvidia and Arms leadership are still pleading their cases to regulators, according to the people, and no final decision has been made. And through it all, the companies have publicly maintained their commitment to the purchase. If Nvidia manages to get the deal over the line, it would be a massive coup for Chief Executive Officer Jensen Huang, who has built a graphics card business into a chip-making empire. Already, he's sitting atop the most valuable U.S. company in the semiconductor industry,

with a market capitalization of more than half a trillion dollars. But it will be an uphill fight. Qualcomm pulled the plug on its $44 billion takeover of NXP Semiconductors NV in 2018 after nearly two years of regulatory hurdles. The sale of arm is under heavy scrutiny because its chip designs are used and everything from phones to cars to factory equipment, making neutrality the foundation of its business model. The world's biggest tech companies rely on arm technology

and they fear they could lose unfettered access under Nvidia. The ordeal has created divisions within Nvidia. Some people at the company are resigned to the acquisitions defeat, but others think management could use the FTC trial to demonstrate the merits of the transaction. Within SoftBank, there are factions that want to let the process play out, especially since a gain in Nvidia's stock price has made the transaction more valuable. Even after a recent tumble, Nvidia's shares have nearly doubled since the ARM deal was announced. That's added tens of billions of dollars to the initial $40 billion price tag.

Others at SoftBank would prefer to pursue an IPO for arms sooner, while the chip industry is still considered attractive to investors, already concerns about a slowdown are growing, end quote. Yeah, don't think that the recent stock market performance hasn't affected the urgency of what is going on here, but what I'm also hearing is that it is the pressure from the Chinese authorities that might be the real fly in the ointment here. If China's going to say no, no matter what happens, then why go through all of the rigmarole? The initial deal is, expires September 13th, two years after being announced. And the thing is, the FTC only sued to stop the transaction in December, so that trial hasn't even started. The EU still needs to weigh in. Trade tensions

with China are already high, so again, surprised, not surprised, time is not on the side of this deal. This was maybe inevitable, too. Google is shuddering flock. It's controversial replacement for cookies and has instead announced topics, which organizes browsing into more than 300 topics for the purposes of advertising buckets, if you will, quoting TechCrunch. Flock, Federated Learning of Cohorts, Google's controversial project for replacing cookies for interest-based advertising by instead grouping users into groups of users with comparable interests is dead. In its place, Google today announced a new proposal. Topics. The idea here is that your browser will look. learn about your interests as you move around the web. It'll keep data for the last three weeks of

your browsing history, and as of now, Google is restricting the number of topics to 300 with plans to extend this over time. Google notes that these topics will not include any sensitive categories like gender or race. To figure out your interest, Google categorizes the sites you visit based on one of these 300 topics. For sites that it hasn't categorized before, a lightweight machine learning algorithm in the browser will take over and provide an estimated topic based on the name of the domain. When you hit on a site that supports the Topics API for ad purposes, the browser will share three topics you are interested in, one for each of the three last weeks, selected randomly from your top five topics for each week. The site can then share this with its

advertising partners to decide which ads to show you. Ideally, this would make for a more private method of deciding which ad to show you, and Google notes that it also provides users with far greater control and transparency than what's currently the standard. Users will be be able to review and remove topics from their lists and turn off the entire Topics API too. Google's Ben Galbraith noted that Google has spoken to a number of parties to gather feedback for this new proposal, but today marks the start of the company's process of collaborating with the ecosystem. It'll remain to be seen if other browser vendors will be interested in adding the Topics API since they all quickly turned a cold shoulder to flock. I'm somewhat

skeptical that they will want to adopt the Topics API, but it'll be interesting to watch how the ecosystem reacts. It's also worth noting that for advertisers, topics is only one potential signal to decide which ad to show a given user. In some ways, it just becomes another signal for them that can be augmented with data about the article a user is currently reading, for example, or other contextual data about the user, end quote. And yet, here's a flock that lives. Twitter code shows that it is developing flock, this one spelled F-L-O-C, unlike the F-L-O-C of the previous segment, which will let users create tweets for up to 150 close friends after first teasing the feature back in July of 2021.

Quoting input. Much like Instagram's close friends, Flock will allow users to add followers to a list and later send out tweets exclusively to those people. Right now, Twitter is playing around with a list cap of 150 people, plenty of people to send your nudes out to in a less public fashion, I guess, RIP fleets. Twitter has yet to confirm any details about the Flock feature except for July's T's. At that point, we got a quick look at the UI, which includes a drop-down menu at the top of a new tweet to choose between audiences. That UI also included an option to have tweets from trusted friends appear at the top of your timeline. New details from a recent reverse engineering of the feature include the 150-person list limit and the fact that people on your list

will not be notified if you remove them. The backend code also reveals what appears to be a notification you'll receive when someone adds you to their flock. What that notification will actually look like is unclear. Adding a close friends-esque feature makes perfect sense for Twitter. Lots of users maintain separate accounts to keep certain tweets private. Why not just build a feature that allows you to maintain two modes of communication on a single account? Will this, like Instagram's close friends, be used for posting nudes in a not-so public manner? You bet it will, especially given that Twitter is much more accepting of nudity than Instagram is. We also wouldn't be surprised if this ended up being misused for bullying purposes because what is Twitter for, if not

taking screenshots and then dunking on them, end quote. Meta yesterday unveiled what it called the RSC for AI Research Supercluster, a supercomputer for machine learning training, claiming that this project will have 16,000 GPUs and be the world's fastest upon completion in mid-2020. Quoting the verge. The news demonstrates the news demonstrates the the absolute centrality of AI research to companies like Meta. Rivals like Microsoft and NVIDIA have already announced their own AI supercomputers, which are slightly different from what we think of as regular supercomputers. RSC will be used to train a range of systems across meta's businesses, from content moderation algorithms used to detect hate speech on Facebook and

Instagram, to augmented reality features that will one day be available in the company's future AR hardware. And yes, Meta says RSC will be used to design experiences for the Metaverse, the company's insistent branding for an interconnected series of virtual spaces from offices to online arenas. RSC will help meta's AI researchers build new and better AI models that can learn from trillions of examples, work across hundreds of different languages, seamlessly analyze text, images, and video together, develop new augmented reality tools and much more, wrote meta-engineers Kevin Lee and Shubo Sangupta in a blog post outlining the news. We hope RSC will help us build entirely new AI systems that can, for example, power

real-time voice translations to large groups of people, each speaking a different language so they can seamlessly collaborate on research projects or play AR games together, end quote. Work on RSC began a year and a half ago, with Meta's engineers designing the machine's various systems, cooling, power, networking, and cabling entirely from scratch. Phase one of RSC is already up and running and consists of 760 NVIDIA GGX A100 systems containing 6,080 connected GPUs, a type of processor that's particularly good at tackling machine learning problems. Meta says it's already providing up to 20 times improved performance on its standard machine vision research tasks. Before the end of 2022, though, phase two of RSC will be complete. At that

point, it'll contain some 16,000 total GPUs and will be able to train AI systems, quote, with more than a trillion parameters on data sets as large as an exabyte, end quote. This raw number of GPUs only provides a narrow metric for a system's overall performance, but for comparison's sake, Microsoft's AI supercomputer built with Research Lab OpenAI, is built from 10,000 GPUs. These numbers are all very impressive, but they do invite the question, what is an AI supercomputer anyway? And how does it compare to what we usually think of as supercomputers? Vast machines deployed by universities and governments to crunch numbers in complex domains like space, nuclear physics, and climate change. The two types of systems known,

as high-performance computers or HPCs are certainly more similar than they are different. Both are closer to data centers than individual computers in size and appearance and rely on large numbers of interconnected processors to exchange data at blisteringly fast speeds. But there are key differences between the two, as HPC analyst Bob Sorenson of Hyperion Research explains to the verge, quote, AI-based HPCs live in a somewhat different world than their traditional HPC counterparts, says Sorensen, and the big distinction is all about accuracy. The brief explanation is that machine learning requires less accuracy than the tasks put to traditional supercomputers, and so AI supercomputers, a bit of recent branding, can carry out more calculations per second

than their regular brethren using the same hardware. That means when meta says it's built the world's fastest AI supercomputer, it's not necessarily a direct comparison to the supercomputers you often see in the news, rankings of which are compiled by the independent. Top 500.org and published twice a year to explain this a little more, you need to know that both supercomputers and AI supercomputers make calculations using what is known as floating point arithmetic, a mathematical shorthand that's extremely useful for making calculations using very large and very small numbers. The floating point in question is the decimal point which floats between significant figures. The degree of accuracy deployed in floating point calculations can be

adjusted based on different formats, and the speed of most supercomputers is calculated using what are known as 64-bit floating-point operations per second or flops. However, because AI calculations require less accuracy, AI supercomputers are often measured in 32-bit or even 16-bit flops. That's why comparing the two types of systems is not necessarily apples to apples, though this caveat doesn't diminish the incredible power and capacity of AI supercomputers, end quote. And I want to mention this real quick with the strong caveat that I'm no geopolitical expert. I have no idea what could or could not happen militarily, politically, or anything involving this next story, but I do want to put something important on your radar at the end. Hackers say they hit a Belarusian rail system with

ransomware to stop Russia's military buildup against Ukraine. Many Belarus Railway services appear to be offline, quoting from Ars Technica. Referring to the Belarus Railway, a group calling itself cyber partisans wrote on telegram, Bell ZHD at the command of the terrorist Lukashenko these days allows the occupying troops to enter our land. As part of the Pecklo cyber campaign, we encrypted the bulk of the servers, databases, and workstations of the Bell ZHD in order to slow down and disrupt the operations of the road. The backups have been destroyed. Dozens of databases have been cyberattacked. Automation and security systems were deliberately not affected by a cyber attack in order to avoid emergency situations. The group also announced the attack on Twitter. A representative from the group said

in a direct message that the Peklo Cyber campaign targets specific entities and government-run companies with the goal of pressuring the Belarus government to release political prisoners and stop Russian troops from entering Belarus to use its ground for the attacks on Ukraine, end quote. So again, underlining the fact that I have no professional understanding of this conflict, but over the weekend, I was hearing a lot that cyber experts are increasingly coming to believe that any Ukraine conflict, should it flare up, could lead to essentially our first taste of all-out cyber war. Cyber war that could spill out globally. Basically, they believe if things go badly or Russia feels threatened in the case of hostilities, it will not hesitate to unleash cyber attacks broadly,

quoting ABC News. As tensions rise in the standoff over Ukraine, the Department of Homeland, Homeland Security has warned that the U.S. response to a possible Russian invasion could result in a cyber attack launched against the U.S. by the Russian government or its proxies. We assess that Russia could consider initiating a cyber attack against the homeland if it perceives a U.S. or NATO response to a possible Russian invasion of Ukraine threatened its long-term national security, a DHS intelligence and analysis bulletin sent to law enforcement agencies around the country and obtained by ABC News said. The bulletin was dated January 23rd, 2020. We assess that Russia's threshold for conducting disruptive or destructive cyber attacks in the homeland

probably remains very high, and we have not observed Moscow directly employ these types of cyber attacks against U.S. critical infrastructure, notwithstanding cyber espionage and potential pre-positioning operations in the past, the bulletin said. Last year, cybercriminals based in Russia caused two of the most destructive cyber attacks in recent memory. The U.S. has said colonial pipeline was the victim of a ransomware attack in May 2021, shutting down operations and causing widespread outages across the country, and meat supplier JBS, had its operations shut down due to Russian-backed hackers. Russia is also responsible for the solar winds breach in late 2020. The U.S. has said where the U.S. says Russian-backed cybercriminals gain access to 10 U.S. government agencies, including the Department of Homeland

Security and Department of Commerce, end quote. So if you did not have massive cyber attacks crippling U.S. infrastructure on your 2022 worry list, here you go. Also, over the weekend, the thing that I kept hearing was that Silicon Valley and tech companies could be a primary target as well. So, sometimes Black Swan events aren't rare. Sometimes, frighteningly, they can seem likely. Here's part of a Twitter thread from Cisco's Director of Threat Intelligence for Cisco's Talos Cyber Intelligence Wing. Matthew Olney, listened to the end, quote, So CISA is warning CI that Russia may act in cyberspace if the West response threatens Russia's national security. I hear some CI respond, so what? We've always faced attacks from Russia.

At its core, that response shows a misunderstanding of what has and is happening. As the philosopher Ron White once said, it's not that the wind is blowing, it's what the wind is blowing. Past foreign adversary activity in critical infrastructure has been to establish long-term access and learn about the environment in preparation for future conflict. What CISA is trying to tell you is that future may be now. If Russia is sufficiently unsatisfied with the West's response to an invasion in Ukraine, they will look for opportunities to apply pressure on the administration. They will have learned from events like Colonial Pipeline, where reports were that the Biden administration was very concerned about the potential political impact of gas lines, etc. They would also have learned that you can have

real-world impact without destroying the technical underpinnings if you make it economic infeasible to pump gas, then the gas will not flow. You can also expect a continuation of disinformation and misleading activity. There will always be just enough about an attack that makes certain people question whether it really was Russia. Perhaps it will look like ransomware, or it will have Black Lives Matter rhetoric. In the end, the intent isn't to harm the long-term capability of the U.S. to supply the needs of its citizens, but to damage the immediate political ability of the administration to interfere in Russian affairs. They will look to do that in a targeted painful way, but in a way that can both be walked back easily, and that can be denied at least

at the surface level, even though anyone paying attention will know the truth. Finally, they aren't going to come straight at you. They have perfected the supply chain attack and the abuse of your trust in your partners. They will turn your complexity against you in ways you will have not anticipated. And that is what CISA is trying to tell you, end quote. Nothing really to add to that sobering note. Talk to you tomorrow.