Tech Brew Ride Home - Thu. 12/05 – Bitcoin $100k!

Episode Date: December 5, 2024

Bitcoin crosses the big $100k mark for the first time. We now know who Trump wants to fill the roles Silicon Valley cares about the most. Is Amazon about to become a top tier AI model player? Two new models from Google DeepMind, one of them promises to revolutionize weather prediction. And Waymo says bienvenido a Miami. Sponsors: WashingtonPost.com/ride Links: Crypto Trading Volume Surged to $10 Trillion for the First Time in November (Bloomberg) Trump Picks Paul Atkins to Run the S.E.C. (NYTimes) Trump Taps Vance Aide Gail Slater as Top DOJ Antitrust Cop (Bloomberg) Amazon announces Nova, a new family of multimodal AI models (TechCrunch) First impressions of the new Amazon Nova LLMs (via a new llm-bedrock plugin) (Simon Willison's Blog) DeepMind’s Genie 2 can generate interactive worlds that look like video games (TechCrunch) Google Introduces A.I. Agent That Aces 15-Day Weather Forecasts (NYTimes) Key leaders behind Google’s viral NotebookLM are leaving to create their own startup (TechCrunch) Waymo to expand to Miami, aims to launch robotaxi service there in 2026 (CNBC) 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 TechMeme right home for Thursday, December 5th, 2024. I'm Brian McCullough today. Bitcoin crosses the big 100K mark for the first time. We now know who Trump wants to fill the roles Silicon Valley cares about the most. Is Amazon about to become a top-tier AI model player, two new models from Google DeepMind, one of them promising to revolutionize weather prediction. And Waymo says, bienvenito amoe ami. Here's what you miss today in the world of tech. Well, it's finally happened. Bitcoin has passed 100,000,000. $1,000 per coin for the first time ever. But get these crazy stats. Bitcoin has rallied more than 126% just since January 2024. And according to CC data, crypto trading volume hit an all-time high in

November with more than $10 trillion of assets traded for the first time ever on centralized spot and derivatives exchanges. Quoting Bloomberg. The combined volume doubled last month, according to data compiled by CC data, crypto markets surged amid optimism for a friend's regulatory environment for the industry under the newly elected Trump administration, with market bellwether Bitcoin jumping 38 percent and setting a record high of almost $100,000. This sentiment is evident in the increased appetite for assets like Ripple, which has historically faced heightened regulatory scrutiny, said Jacob Joseph, senior research analysts at CCData. Optimism is also evident on the institutional side, with CME volumes seeing a significant

uptick and substantial inflows into spot Bitcoin ETFs over the past month, end quote. Monthly spot trading volume on centralized exchanges increased $128% to $3.43 trillion, the second highest number ever since May 2021. Derivatives volume climbed 89% to $6.99 trillion, surpassing the previous all-time high from March, according to CC data. The data doesn't include volume from so-called decentralized finance platforms, end quote. We now know who President Trump wants for two of the roles the tech industry cares about the most. Trump has selected former SEC Commissioner Paul Atkins, who is seen as a strong advocate for looser regulation of crypto assets to succeed Gary Gensler as SEC chair. And Gail Slater, an economic policy advisor to J.D. Vance and tech policy advisor during Trump's first term, has been proposed as head of the DOJ's antitrust unit. Quoting the times, Mr. Atkins, 66, who was an SEC commissioner under President George W. Bush, is a well-known and generally admired figure in Washington legal circles and the securities.

regulatory community. In the early 1990s, he worked at the SEC during both Republican and Democratic administrations. He is also seen as a strong advocate for looser regulation of crypto assets in issue that Mr. Trump adopted as part of his campaign. Shortly after his term as SEC Commissioner ended in 2008, Mr. Atkins founded Petomac Global Partners, a financial services consulting firm. Potomac provides advisory services to banks and investment firms on regulatory and compliance matters. More recently, the firm has advised clients on issues related to crypto and digital assets. Mr. Atkins, a lawyer, has been playing an active role in helping to draft best practices for crypto trading platforms as co-chair of the Token Alliance, which is part of the Digital Chamber of Commerce.

He is also on the advisory board for scrutinize a digital asset firm that promotes the use of digital tokens. His nomination will need Senate confirmation, end quote. And quoting Bloomberg. Slater is seen as a pro-enforcement populist Republican, particularly when it comes to the tech sector and her nomination signals a new Trump administration is unlikely to fully back down from the aggressive stance of Joe Biden's administration. Her boss, J.D. Vance, who has also represented Ohio in the Senate since 2023, has expressed his support for much of the agenda of outgoing FTC chair, Lena Kahn. Trump hasn't yet named Kahn's replacement at the FTC, which enforces both antitrust and consumer protection. laws, Slater would replace Biden's Assistant Attorney General for antitrust Jonathan Cantor, who has held the post since 2021 and brought a record number of monopolization cases against companies including Google, Apple, Visa, and others. At the Justice Department, Slater will need to tackle a handful

of pending mergers, including Capital One Financial's proposed acquisition of Discover Financial Services, as well as a half-dozen monopolization lawsuits against some of the largest U.S. companies. The agency has two ongoing cases against Google and is heading back to trial in April on what changes the tech giant must make to its business after being found to have illegally monopolized the online search market. Slater will also be responsible for deciding whether to move forward or settle recent cases filed against Live Nation Entertainment's Ticketmaster, Visa, and Apple. Slater has focused on issues facing big tech companies before. Following her FTC stint, Slater joined the Internet Association, a now-defunct trade group for Internet companies,

whose members included Amazon and Google. She moved to the White House during the First Trump administration as a special assistant to the president for economic policy with a broad portfolio that included technology, telecom, and cyber issues, end quote. Somehow this slipped through the cracks at Reinvent. AWS this week unveiled Nova AI models for Bedrock, including Micro, Light, Pro, and Premier versions for text generation. Also, canvas for image generation and reel for video generation.

Quoting TechCrunch. The text-generating Nova models, which are optimized for 15 languages, but primarily English, have widely varying sizes and capabilities. Micro can only take in text and output text, but delivers the lowest latency of the bunch, processing text, and generating responses the fastest. Light can process image, video, and text inputs reasonably quickly. Pro offers a balanced combination of accuracy speed and cost for a range of tasks, and Premiere is the most capable designed for complex workloads. Pro and Premier, like Light, can analyze text, images, and video. All three are well suited for tasks like digesting documents and summarizing charts, meetings, and diagrams.

AWS is positioning Premier, however, as more of a teacher model for creating tuned custom models rather than a model to be used on its own. Micro has a 128,000 token context window, meaning it can process up to around 100,000 words. Light and Pro have 300,000 token context windows, which works out to around 225,000 words, 15,000 lines of computer code, or 30 minutes of footage. In early 2025, certain Nova models context windows will expand to support over 2 million tokens, AWS says. Canvas and Real are AWS's strongest play yet for generative media. Canvas lets users generate and edit images using prompts to remove backgrounds, for example,

and provides controls for the generated images, color schemes, and layouts. Real, the more ambitious of the two models creates videos up to six seconds in length from prompts or optionally reference images. Using Real users can adjust the camera motion to generate videos with pans, 360-degree rotations, and zoom. Real is currently limited to six-second videos, which take about three minutes to generate, but a version that can create two-minute-long videos is, quote, coming soon, according to AWS, end quote. Now, here's what's interesting. You know that I tend to turn to Simon Willison to tell me who's up and who's down in the whole model race. And interestingly, Simon says his first impressions

of Nova are that the models are competitive with Google Gemini and extremely inexpensive, and thus may position Amazon as a top-tier model provider for the first time. As a general initial vibe check, these Nova models pass with flying colors. Google Gemini now has competition in terms of pricing. This is a relief to me. Gemini is so cheap, I've been nervous that they'll jack the prices up through lack of competition. These appear to be good multimodal models. Their image handling feels solid, and it's neat to have a new video input model, even if it's quite limited compared to Gemini. Anthropics' decision to raise the price for Claude 3.5 Haiku isn't looking great, given the continued competition at the bottom end of the market.

Just a few months ago, Claude 3 Haiku was the cheapest model available. Now it's hardly worth considering undercut by even GPT40 Mini. The cost of multimodal models, First Gemini, now Nova, continues to leave me almost speechless. at 0.0136 per image, I could generate a description of every photo in my 67,771 personal photo library using Nova Mini for $9.21. With this release, I think Amazon may have earned a spot among the top tier of model providers. Maybe we need a new Fang acronym that covers OpenAI, Anthropic Google Meta, and Amazon. I like Gamoa, end quote. Google DeepMind has also released two new models, but each are interesting enough in their own way that they deserve their own segments. First up, say hello to

Genie 2, a model that can generate 3D worlds from a single prompt image playable by humans or AI agents using keyboards and mouse inputs, according TechCrunch. Called Genie 2, the model, the successor to Deep Minds Genie, which was released earlier this year, can generate an interactive real-time scene from a single image and text description, e.g. acute humanoid robot in the woods. In this way, it's similar to models underdevelopment by Phi Phi Li's company World Labs and Israeli startup Descartes. DeepMind claims that Jeannie 2 can generate a vast diversity of rich 3D worlds, including worlds in which users can take actions like jumping and swimming by using a mouse or keyboard. Trained on videos, the models able to simulate object interactions, animations, lighting, physics, reflections, and the behavior of NPCs. Many of Genie 2 simulations look like AAA video games.

and the reason could well be that the model's training data contains play-throughs of popular titles. But DeepMind, like many AI labs, wouldn't reveal many details about its data-sourcing methods for competitive reasons or otherwise. One wonders about the IP implications. DeepMind, being a Google subsidiary, has unfettered access to YouTube, and Google has previously implied that its terms of service give it permission to use YouTube videos for model training. But is Jeannie 2 basically creating unauthorized copies of the video games it watched? That's for the courts to decide. DeepMind says that Genie 2 can generate consistent worlds with

different perspectives like first person and isometric views for up to a minute with the majority lasting 10 to 20 seconds. Most models like Genie 2, world models, if you will, can simulate games and 3D environments, but with artifacting, consistency, and hallucination-related issues. For example, Descartes's Minecraft simulator oasis has a low resolution and quickly forgets the layout of levels. Genie 2, however, can remember parts of a simulation. scene that aren't in view and render them accurately when they become visible again. World Labs models can do this too. Now, games created with Jeannie 2 wouldn't be all that fun, really, given that they'd erase your progress every minute or so. That's why DeepMind is positioning

the model as more of a research and creative tool, a tool for prototyping interactive experiences, and evaluating AI agents, end quote. But maybe more interesting and applicable for everyday folk like you and me, DeepMind also unveiled Gencast, an AI weather model. that the company claims outperforms traditional methods on up to 15-day weather and deadly storm forecasts. Quoting the New York Times. It's a big deal, said Kerry Emanuel, a professor emeritus of atmospheric science at the Massachusetts Institute of Technology, who was not involved in the deep mind research. It's an important step forward, they said.

In 2019, Dr. Emanuel and six other experts writing in the Journal of the Atmospheric Sciences argued that advancing the development of reliable forecasts to a length of 15 days from 10 days, would have, quote, enormous socioeconomic benefits by helping the public avoid the worst effects of extreme weather. Ilyn Price, the new paper's lead author, and a senior research scientist at DeepMind describes the new AI agent which the team calls GenCast as much faster than traditional methods, and it's more accurate, he added. He and his colleagues found that GenCast ran circles around DeepMind's previous AI weather program,

which debuted in late 2023 with reliable 10-day forecasts. Remy Lam, the lead scientist on the project and one of a dozen co-authorized, co-authors on the new paper described the company's weather team as having made surprisingly fast progress. I'm a little bit reluctant to say it, but it's like we've made decades' worth of improvements in one year, he said in an interview. We're seeing really, really rapid progress. The world leader in atmospheric prediction is the European Center for Medium Range weather forecasts. Comparative tests regularly show that its projections exceed all others in accuracy. Deep Mine tested its new AI program against the center's ensemble prediction system,

a service that 35 nations rely on to produce their own weather forecasts, the team compared how the 15-day forecasts of both systems performed in predicting a designated set of 1,30 global wind speeds, temperatures, and other atmospheric features. The Nature Report said the new agent outdid the center's forecasts 97.2% of the time. The AI achievement, the authors wrote, helps open the next chapter in operational weather forecasting. Matthew Chantry, an AI specialist at the European Center for Medium Range weather forecasts, said his agency was already adopting some of its features. That's how highly we think of it, he said. Machine learning in general, Dr. Chantry added, was accelerating human bids to outmaneuver

some of nature's deadliest threats, end quote. And this was probably inevitable. Fresh off the tool being part of Spotify Rapp, as I told you about yesterday, three of Google's Notebook LM team members, including Razum Martin, a designer and an engineer, are leaving to launch a startup focused on a user-first AI product. Quoting TechCrunch. On LinkedIn, ex-team lead, Reza Martin, said she and her two other co-founders, designer Jason Spielman and engineer Stephen Hughes, couldn't shake the feeling that there's a massive opportunity to build something transformative in the space.

The startup is in full stealth mode with only a bare-bones website that gives no details about its purpose or even its name. It's not clear if the startup will focus on things that Notebook LM went viral for, such as AI-generated podcasts and AI-assisted note-taking, or if it will do something totally different. In comments to TechCrunch, Martin gave few details, but hinted that the startup would be consumer-facing, emphasizing the team wanted to create something that leverages the latest AI models

to build something useful to regular people, end quote. And finally, Waymo is continuing its region-by-region expansion. Somebody alert Will Smith because Waymo is coming to Miami. Quoting CNBC. Waymo said it will first begin cruising through the Florida City with human safety drivers in 2025, before opening doors to riders for its Robo Taxi Service through its Waymo One app in 26. The expansion into Miami is indicative of Waymo's growing confidence in operating its self-driving

vehicles in harsher weather conditions in large metropolitan areas in the U.S. Waymo first tested in Miami in 2019, which the company said helped improve the ability of its self-driving vehicles to navigate in wet and rainy conditions. We deepened our learning and understanding of the Waymo driver's performance in adverse weather conditions, a company spokesperson said. Waymo will use what it learned when it returns. the city with its all-electric jaguar eye pace next year. The company said its initial territory in Florida will include some parts of Miami's larger metropolitan area, which has a population of

more than 6 million people. Waymo has been rapidly expanding its operations over the last year, thanks to additional funding. In November, the company announced it was removing its waitlist of about 300,000 people in Los Angeles, so anyone would be able to use Waymo One to hail a self-driving robotaxie throughout the nearly 80 square miles of Los Angeles County. The company's writing hailing service also operates citywide in Phoenix and San Francisco. And in September, Waymo announced a partnership with Uber in Austin and Atlanta. Through that deal, Uber riders will be able to access Waymo's robo-taxies through the Uber app, starting in early 2025, and Uber will be responsible for fleet management and operations of the Waymo vehicles, including maintenance and infrastructure,

such as vehicle charging, cleaning, and repairs, end quote. I never learn, do I? Never pre-announce an interview, because that only jinxes it. Jay Graber had to reschedule right before we were scheduled to talk yesterday. So I do hope to have that interview for you soon, just not sure how soon. Never pre-announce, never pre-announce, never pre-announce, never pre-announce, never pre-announce.