Welcome back to The Prompt.
President-elect Donald Trump is considering appointing an “AI czar” to lead efforts in artificial intelligence policy, as well as the government’s use of the technology, Axios reports. The official likely won’t be Trump advisor and Tesla CEO Elon Musk, who has become an influential figure on Trump’s transition team, but Musk will likely play a major role in shaping AI policy as well.
The potential hire comes as Trump has courted Silicon Valley, as some venture capitalists and founders have become more vocal supporters of his. He is reportedly also appointing a “crypto czar” according to Bloomberg, though the AI and crypto job could be rolled into one emerging tech position, Axios said.
Now let’s get into the headlines.
PEAK PERFORMANCE
Anthropic this week unveiled a new standard for connecting AI systems to data sources. Called the Model Context Protocol, the idea is to help frontier models gain access to more siloed information, like business tools and inside app development environments. The standard is open sourced, so it can be used with not only Anthropic’s models like Claude, but the models from other AI developers as well.
BIG PLAYS
Uber is entering a new market: training artificial intelligence, Bloomberg reports. The rideshare company has a new division, called Scaled Solutions, which provides contractors to clients for the data labeling and annotating work used to develop AI models. The company started signing up contractors this month in India, the US, Canada, Poland and Nicaragua. Uber is joining a growing market that includes rivals like Scale AI, valued at $14 billion.
AI DEALS OF THE WEEK
Amazon is pouring another $4 billion into Anthropic. The massive deal brings the e-commerce giant’s investment in Anthropic to $8 billion in total, after Amazon injected billions into the startup last year. As part of the deal, Amazon Web Services, the tech giant’s cloud provider, will become Anthropic’s primary cloud for training its AI. Anthropic will also use Amazon’s in-house AI chips, Trainium and Inferentia, to train and deploy its models.
/dev/agents, a company building an operating system for AI agents, launched on Tuesday with $56 million in seed funding, co-led by Index Ventures and CapitalG, Google parent Alphabet’s growth fund. The startup’s founding team is made up of prominent ex-Google and Stripe executives that help create Google’s Android platform, including David Singleton, former CTO of Stripe, and Hugo Barra, former VP of product management at Google. Angel investors also include Scale AI CEO b, OpenAI cofounder Andrej Karpathy and Sarah Guo, founder of the venture firm Conviction. The startup is valued at $500 million, according to Bloomberg.
DEEP DIVE
Why The DOJ Is Trying To Curtail Google’s AI Future
Late on Wednesday, the U.S. Department of Justice filed sweeping proposals for remedies in Google’s landmark antitrust case, where a federal judge ruled earlier this year that it has an illegal monopoly in the online search market. Among the marquee requests: Forcing Google to sell off its popular Chrome browser, banning multi-billion dollar distribution contracts like the one Google has with Apple, or potentially barring Google from requiring Android phone makers to include Google apps on their devices.
But beyond those headline-grabbing demands, the government also included provisions that could hobble Google’s future in the competitive race to control the future of AI. The DOJ proposed Google must sell any stakes in AI companies with technology that could compete in search, and divest within six months of a final judgment from the court. The agency also recommended barring any new acquisitions, joint ventures or partnerships with AI companies competing in search.
Notably, if the judge in the case agrees, that could mean forcing Google to sell off its investment in Anthropic, the firm founded by OpenAI defectors in 2021 and could reportedly be valued at up to $40 billion. Last year, Google said it would invest $2 billion in Anthropic, following a $4 billion deal Amazon announced with the company months before. On Wednesday, regulators in the U.K. cleared the investment from Google, saying it wouldn’t conduct a full-scale investigation to scrutinize the deal after an initial probe.
Anthropic is the maker of Claude, a language model that can generate answers to questions, similar to Google’s own Gemini model, which was integrated into Google’s search engine earlier this year. While the startup doesn’t pitch Claude as a search product, those kinds of chatbots are widely seen as a threat to Google search. Other startups, like Perplexity, backed by Nvidia and Jeff Bezos, are more forthright about competing with Google. “It’s a good morning to be Perplexity,” one prominent AI investor told Forbes. Also of note, the DOJ proposal would take Perplexity off the table as a potential Google acquisition target. (Disclosure: Forbes has threatened legal action against Perplexity for plagiarizing our content.)
The filing also proposes that Google gives all publishers and content creators — including those on Google-owned YouTube — a simple way to opt out of having their content be used to train or file-tune Google’s AI models or other AI products, and vow not to retaliate against those who choose to do so.
Read the full story on Forbes.
AI INDEX
700,000
The number of homes that could be powered annually by one data center campus with peak demand of one gigawatt, or one billion watts — the type of data centers that underpin the world’s AI usage. According to a report by CNBC, the energy consumed by AI systems could use more electricity than entire cities. In addition to the power drain, finding enough land to house the sprawling complexes is also becoming a challenge. For example, Tract, a developer that secures land for infrastructure projects, said it has assembled more than 23,000 acres of land for data center development across the U.S.
QUIZ
This country is launching a $240 million plan to invest in AI development and defense.
- Poland
- Singapore
- India
- Mexico
Check if you got it right.
MODEL BEHAVIOR
How do you fight against phone scammers preying on innocent victims? The answer might be Daisy, the AI granny. Daisy is a chatbot unveiled this month by the British phone company O2, the New York Times reports, which deploys the bot to waste the time of scammers by telling meandering stories that go nowhere, even providing fake personal and account information. The idea is to use the bias of scammers, who often target the elderly, against them, and keep them on the phone for hours–time that can’t be spent against real people.
