Welcome back to The Prompt,
Elon Musk’s Department of Government Efficiency (DOGE) has set out to slash government spending by cancelling contracts and firing thousands of workers across federal agencies (but also hastily rehiring hundreds of them back). On Saturday, DOGE asked all federal employees to send an email listing the work they had accomplished in the past week to assess whether their positions are necessary. Those email responses are slated to be ingested by an AI model to determine which roles to cut, NBC News reported, though over the weekend the administration said filling out emails was voluntary after the heads of multiple federal agencies instructed employees not to respond.
It’s not the first time the use of artificial intelligence has surfaced as DOGE moves through the government. Earlier this month, DOGE representatives fed sensitive data from the Department of Education into an AI model as it searched for areas to cut spending, the Washington Post reported. DOGE is also building a custom chatbot called GSAi for US General Services Administration to increase employees’ productivity, Wired reported.
Now let’s get into the headlines.
BIG PLAYS
Weeks after the announcement of a $500 billion joint AI infrastructure initiative called Stargate, yet another tech giant has made a massive AI investment pledge. On Monday, Apple announced plans to invest $500 billion in AI infrastructure. The iPhone maker plans to build a 250,000 square foot AI server factory in Houston, Texas. The news comes as Microsoft, which canceled leases for some data centers, appears to be rethinking its data center strategy as OpenAI shifts its compute needs from Microsoft to Oracle.
PEAK PERFORMANCE
Multiple AI companies, including OpenAI and Perplexity, have launched AI systems they claim can carry out in-depth research by scrounging the web for information and presenting an analysis. Last week, Google joined the fray with its AI “co-scientist” it said can help scientists generate hypotheses and research proposals.
And Anthropic, which is reportedly finalizing a $3.5 billion funding round, released a new agentic model called Claude 3.7 Sonnet, that can use computer applications and has better performance at coding. AI coding startups like Cursor and Cognition were quick to add the model to their applications.
AI DEALS OF THE WEEK
It was a busy week in the fundraising world as multiple AI startups announced gigantic investment rounds.
AI recruiting startup Mercor raised a $100 million Series B round at a $2 billion valuation.
Cloud computing startup Lambda Labs raised $480 million at a $2.5 billion valuation.
TogetherAI, which provides where AI companies can run and train models, raised $305 million at a $3.3 billion valuation
AI healthcare startup OpenEvidence raised $75 million at a $1 billion valuation, CNBC reported.
AI inference platform Baseten raised $75 million at a $825 million valuation.
DEEP DIVE
AI chatbots have a tendency to “love bomb” their users, says Quinten Farmer, CEO of AI voice companion startup Tolan. Most AI companions apps, such as Character AI and Replika, are designed to steer the conversation into the realms of romance, role play and intimacy. And in many instances, young users have gotten addicted to and become emotionally reliant on the technology, resulting in a string of lawsuits.
Farmer claims to have a different approach: he’s building a voice-based conversational AI tool designed to be a “non-romantic” friend. Tolan, touted as an “alien best friend,” has 500,000 app downloads and 90% of its userbase are women between the ages of 16 and 24. The AI voice companion, built on top of a bundle of AI models from OpenAI, Anthropic and others, engages in casual conversations, helps with homework and can engage in emotional conversations about stress, arguments with parents and breakups. To ensure users don’t get attached, the app prompts users to take a break from the conversation at regular intervals (but won’t cut them off if the user wants to keep talking) and varies the types of interactions users are having with it.
“What we’ve found is that a lot of existing AI chatbots were really trying to suck you into their world and get you to role play as an anime character or a Game of Thrones character,” Farmer told Forbes. “To form a secure healthy attachment, you really want a character that is curious about your world.”
Farmer, whose previous fintech company Even was acquired by Walmart for $300 million, is among a growing number of entrepreneurs building new applications with existing AI models as they increasingly get commoditized. David Luan, former CEO of AI agentic startup Adept and head of Amazon’s AGI lab who invested in the company, said startups like Tolan are “exploiting the second order effects of AI.” Other investors in the startup include Nat Friedman, Daniel Gross and Physical Intelligence cofounder Lachy Groom.
With about $10 million in seed funding, Tolan faces a fair deal of competition from much larger players. But while OpenAI’s ChatGPT, which just crossed 400 million weekly active users last week, also has voice features allowing it to have long conversations with people, Luan says users come to ChatGPT to solve factual problems, whereas “people’s affinity with Tolan is a trustworthy confidant that helps you navigate the world.”
WEEKLY DEMO
EdTech company Chegg is suing Google, alleging that the search giant’s AI-generated summaries are hurting the company’s traffic and revenue, CNBC reported. Google’s AI search overviews generated content that contained details from Chegg’s website without attribution and competed with its own e-learning offerings. Chegg, which is used by students for homework help (and criticized for facilitating cheating), reported $6.1 million in losses in its Q4 2024 revenue. The struggling company has previously blamed students turning to ChatGPT for its revenue losses, as well. Other sites like Kayak and TripAdvisor are also concerned about Google chipping away at traffic to their sites, Forbes reported
MODEL BEHAVIOR
TVs at the headquarters of the Department of Housing and Urban Development showed an AI-generated video of Donald Trump kissing Elon Musk’s feet, displaying the words “Long live the real king” Wired reported. The deepfake video was played as the tech billionaire has acquired a position of unprecedented power under Trump’s second presidency.
Read the full article here
