Chaos Lever Podcast

 Welcome to another episode of Tech News of the Week, where we break down the hottest stories in technology with a dose of snark and insight. This week, we’re diving into Google’s AI extravaganza, GM’s retreat from robo-taxis, quantum computing milestones, and Microsoft’s push for eco-friendly data centers. Grab your coffee and let’s get into it!

Google’s AI Blitzkrieg 🚀
Google is doubling down on artificial intelligence, unveiling a slew of announcements including Gemini 2.0, a next-gen model with audio and video capabilities, and ASTRA, which can summarize videos and answer questions. They also revealed Trilium, an AI training chip, and Jules, an AI coding assistant capable of fixing bugs. But don’t get too excited—most of these features are still in development, with limited demos and no hands-on access yet. For now, it’s a lot of promises, but we’ll see how it shakes out in January. Read more: https://techcrunch.com/2024/12/11/gemini-2-0-googles-newest-flagship-ai-can-generate-text-images-and-speech/

GM Pulls the Plug on Robo-Taxis 🚗
GM is merging its Cruise division into its Super Cruise engineering team and stepping back from the robo-taxi race. This comes after a string of incidents, including a high-profile crash that led to California revoking Cruise’s operating license. With plans to integrate the tech into personal vehicles instead, GM hopes to salvage their investment. But for now, fully autonomous taxis are hitting a red light. Read more: https://techcrunch.com/2024/12/11/gm-is-giving-up-on-cruise-robotaxis-pivots-to-personal-autonomous-vehicles

Google’s Quantum Leap 🧮
Google made headlines with Willow, a quantum chip boasting over 100 quality qubits. They claim it can complete specific computations that would take conventional supercomputers septillions of years. While this sounds groundbreaking, the practical use of this technology is still unclear, and some remain skeptical of Google’s claims. However, advancements in error rates and quantum stability are undeniably impressive. Read more: https://www.tomshardware.com/tech-industry/quantum-computing/google-claims-its-new-willow-quantum-chip-can-swiftly-solve-a-problem-that-would-take-a-standard-supercomputer-10-septillion-years

Microsoft’s Water-Cooled AI Data Centers 🌊
Microsoft announced a shift to water-saving, closed-loop cooling systems for their AI-heavy data centers. These new systems reduce water usage in drought-prone areas like Phoenix, Arizona, and are part of a larger push to retrofit existing facilities. It’s a smart move to address the environmental toll of AI, though some wonder why this wasn’t done sooner. Read more: https://siliconangle.com/2024/12/10/microsoft-previews-new-water-efficient-data-center-design/ 

Thanks for tuning in! Be sure to subscribe, share your thoughts in the comments, and take our listener survey at https://chaoslever.com/survey. See you next week! 

What is Chaos Lever Podcast?

Chaos Lever examines emerging trends and new technology for the enterprise and beyond. Hosts Ned Bellavance and Chris Hayner examine the tech landscape through a skeptical lens based on over 40 combined years in the industry. Are we all doomed? Yes. Will the apocalypse be streamed on TikTok? Probably. Does Joni still love Chachi? Decidedly not.

[01:00:00.00]
Announcer: Welcome to Tech News of the Week with your host, a recursive joke you'll never hear the end of.

[01:00:09.22]
Ned: Welcome to Tech News of the Week. This is our weekly Tech News podcast, where Chris and I talk about four interesting articles that we found. Before we jump into it, I would like to once again remind you that we are running our listener survey over at chaoslever/survey, so please take the survey, or don't, whatever. I don't care. Chris, go ahead. Just go. Let's go.

[01:00:32.12]
Chris: Google releases a lot of AI stuff. Like a lot. Fast on the heels of OpenAI's release of Sora, which is video-generating AI, Canvas for finetuned document editing in a GUI environment, and their Apple intelligence integrations going live. Yay. Advertising company Google thought it was time to do a huge dump of AI features of their own. First up was Gemini 2.0 flash. This is a new model that is an experimental release, but it is expected to add audio and video capabilities. Google claims that it will be twice as fast as the current 1.5 models and be generally available around the end of January. Other things they announced: Trilium, an AI training chip, Jules, an AI coding assistant that's to be able to fix bugs on its own, terrifying concept, ASTRA, which can summarize videos and answer questions about them, DeepResearch, which is supposed to make the Gemini chatbot into a research assistant. And AI agents that can do things like, quote, understand the rules of video games like Clash of Clans. That last one, I expect, will be superbly popular as if there's one thing that gamers like more than complaining about games online, it's cheating while they play those games online.

[01:02:04.10]
Chris: I'm sorry, I meant getting assistance. It should also be noted that, one, this was not all the announcements, but two, they were announcements only. What I mean by that is there were no functioning testable examples, and only a few had demonstration videos available, at least at the time of this writing. Expect me to copy this This whole thing basically verbatim with 30% more snark in January when this stuff is all actually released.

[01:02:38.12]
Ned: I assume you also saw that the Sorra website is not accepting new logins. That happened almost immediately after they released it, almost like it wasn't ready yet. What? Gm is cruising away from robo-taxis. Gms, crews, not division, they're a crew's subsidia. January, maybe we could call it, is being merged with the autonomous driving engineering team at GM to work on their Super Cruise driving feature. As part of the transition, GM is abandoning their investment in robo-taxi service in all current markets. Cruise was one of a handful of fully autonomous robotaxis that had been approved to work in markets like San Francisco, Phoenix, and Austin. Why does Arizona hate pedestrians so much? That's like a different channel, but it is definitely true. Strode's are the worst thing ever. Why is GM backing away from billions of dollars of investment in the robotaxi service? They mentioned the time and resources necessary to scale the business in the increasingly competitive Robotaxi market. There's a bit more to the story. In the last year, well, actually in the last three months, Cruise has been in multiple traffic incidents, including one where a woman was dragged under one of their cars for 20-ish yards.

[01:04:08.28]
Ned: Yeah, not great. Cruise lost their license based on that to operate in California. In the wake of the investigation, GM paused testing in other states, laid off 900 people, and replaced the CEO of Cruise. You could say that the writing was on the wall. Gm currently owns 90% of the company and has plans to increase that to 97% shortly. They're still planning to use the technology as Super Cruise, which is already available on 20 models. Models that I will absolutely not be buying and probably run away from screaming if I encounter them in the wild.

[01:04:48.23]
Chris: Or you could just put a parking cone on it.

[01:04:51.28]
Ned: Touche.

[01:04:53.22]
Chris: Google releases a quantum chip that Google claims is the best quantum chip. Pivoting from theoretical AI releases to potentially parallel universe probing releases now, as Google this week also announced Willow. Willow is their latest quantum processor, which they state has over 100 quality qubits. Quality qubits, as long-time listeners will remember, refers to the ability to do error correction, which is extremely difficult at present in quantum environments. Google claims that Willow has set a new record on a specific computation that has been used for years to benchmark quantum computers. On that particular computation, a conventional A real supercomputer would take, 10 septillion years. Now, it should be noted that this particular calculation is completely and entirely useless in the real world. And it is only used because it tests quantum computers in ways that are known to fail dramatically in conventional processing. Now, I say currently because it's entirely possible that this problem, which is known as random circuit sampling, it might only be hard to do on a conventional computer because it's so goddamn useless that nobody's figured out how to do it properly. Also, there have been many instances in the past where Google's announcement of their computational dominance have been accused of overstating.

[01:06:37.16]
Chris: Still, if their claims about error rates are accurate, this is still big news. Error rates cut in half and then half again while retaining quantum state for up to five times longer than previous efforts is a big deal. Google did release a research paper to go along with this announcement that was published in Nature. I I don't have access to nature, though, so I'll just go ahead and assume that the paper is great.

[01:07:06.28]
Ned: Yeah, that little weird thing in the middle of the announcement where he just says, This proves that parallel universes that exist? That was a strange take.

[01:07:17.15]
Chris: Yeah. Yeah.

[01:07:20.26]
Ned: Did that not go through a copy editor first? Anyway. Microsoft is building water-cooled data centers. When they already have water-cooled data. But I digress. Ai workloads run hot. Very hot. Hot, hot, hot, some might say. So they need to be cooled. Data centers that use evaporative cooling run through a lot of water, which might be fine for some communities, but droughts are becoming more common by the minute. Microsoft announced in a blog post this week that they are implementing chip-level cooling systems with zero water evaporation to cool their AI-focused data centers. The water will be cycled through a closed loop between the servers and chillers without needing to add fresh water to the system. Why this wasn't already the case is beyond me. We have radiators. We've had closed-loop radiators for over 100 years. This isn't exactly new territory, but whatever, sure. Good job on Microsoft for burning up the planet slightly less faster than before. Okay, so the reason is because it uses less electricity to use evaporative cooling, and it's less complex because you don't have to run it straight to the tip and blah, blah, blah, whatever. Still stupid, but at least there is a reason.

[01:08:43.02]
Ned: I should also point out that this is for new projects only, including ones at Phoenix, Arizona, and Mount Pleasant, Wisconsin. That one in Arizona seems particularly important due to its lack of water. The longer term goal is to retrofit existing data centers with this technology, although no timeline was shared for that goal. That's it. We're done. Go away now. Bye.