Google unveils new models, Karpathy joins
- Google on May 19 unveiled new Gemini models, AI agents in Search and a premium AI Ultra subscription at its I/O conference. - OpenAI committed more than S$300 million, or $234 million, to Singapore and said it would open its first overseas Applied AI Lab. - Anthropic said Andrej Karpathy joined its pre-training team; OpenAI is lobbying U.S. states as federal AI legislation remains stalled.
Alphabet's Google used its I/O conference on May 19 to widen its AI push across consumer products, software development and corporate sales, adding new Gemini models, agentic tools and a higher-priced subscription tier. The Mountain View, California, company also introduced a lower-cost enterprise model as it tries to answer pressure from OpenAI and Anthropic, according to Reuters. One day later, Anthropic said former OpenAI founding member Andrej Karpathy had joined its pre-training team, while OpenAI announced a more than S$300 million ($234 million) commitment to Singapore's AI ecosystem and plans for its first overseas Applied AI Lab. Together, the moves showed competition spreading across products, hiring and national partnerships. ### What did Google actually roll out at I/O? Google on May 19 put AI agents directly into Search and introduced a faster, cheaper Gemini model aimed at enterprise customers, Reuters reported. The company also unveiled new AI tools for coders and consumers, including product updates tied to Search and YouTube, as it tried to blunt gains by OpenAI and Anthropic with business users. (newsbreak.com) CNBC reported that Google also launched a higher-end AI Ultra subscription and additional model updates at the conference. Reuters said the cheaper enterprise offering was part of a broader effort to court developers and businesses at a time when pricing and access to capable models have become a competitive lever. (newsbreak.com) ### Why does Karpathy's move matter to Anthropic? Andrej Karpathy said on May 19 that he had joined Anthropic, adding a high-profile researcher to one of OpenAI's closest rivals. Reuters identified Karpathy as a former Tesla AI executive and one of OpenAI's founding members. (money.usnews.com) Anthropic assigned Karpathy to its pre-training team, according to reports surfaced in search results. That group handles the large-scale training runs that shape Claude's core capabilities, making the hire notable in a market where companies are competing not only on chips and products but also on researchers with frontier-model experience. That characterization of the competitive stakes is an inference drawn from the role description and Reuters' framing of Anthropic's effort to strengthen itself in the AI race. (msn.com) ### Why is Singapore showing up in this fight? Singapore on May 20 announced separate AI partnerships with Google and OpenAI, while OpenAI committed more than S$300 million to the country's local ecosystem, CNBC reported. The ChatGPT maker said it would establish the OpenAI Singapore Applied AI Lab, its first overseas lab, and focus on areas including education, public services, finance, healthcare and digital infrastructure. (msn.com) Google's agreement with Singapore did not include a disclosed investment amount, CNBC said, but it focused on workforce readiness, enterprise innovation, healthcare, scientific research and secure AI deployment. CNBC also said Singapore has already committed more than 1 billion Singapore dollars over 2025-2030 to strengthen public AI research capabilities. (cnbc.com) ### Where does policy fit into the latest corporate moves? Politico reported on May 20 that OpenAI is pressing blue-state lawmakers to pass legislation that would support what its lobbyists describe as a national AI framework. The report said Chris Lehane, OpenAI's top lobbyist and political strategist, has called the approach "reverse federalism" as Congress remains deadlocked on AI safety rules. (cnbc.com) Those state-level efforts are unfolding as companies expand products, recruit talent and strike government partnerships. The sequence of announcements this week — Google's I/O launches, Karpathy's move to Anthropic, and OpenAI's Singapore commitment — shows that the contest is being fought simultaneously in model performance, hiring and public-sector alignment. That is an inference based on the timing and substance of the announcements. (politico.com) ### What comes next? OpenAI said its Singapore lab will be its first overseas Applied AI Lab and that the investment will support hiring and local ecosystem development over the next few years. Singapore's government said the partnerships are tied to deployment across public services, healthcare, education and enterprise. (newsbreak.com) Google's next test will be whether developers, businesses and consumers adopt the new Gemini tools and the AI Ultra tier after I/O. Anthropic's next benchmark will be future Claude model work from its pre-training team, now including Karpathy. OpenAI, meanwhile, is continuing its state-level lobbying campaign in the United States while it expands abroad. (newsbreak.com) (cnbc.com)