According to a report by The Information, Misha Laskin and Ioannis Antonoglou, former employees of Google's DeepMind, left the company to start their own startup earlier this year. In a recent podcast hosted by Sequoia, Laskin emphasized the importance of a universal agent. He explained that such an agent should possess a broad skill set and be capable of handling various inputs while also being adept at complex tasks. Laskin discussed the different types of AI agents in the market during the podcast. For example, he mentioned AlphaGo, an AI program renowned for defeating professional Go players. While AlphaGo excels at this particular task, Laskin noted that it lacks versatility and cannot play other games like tic-tac-toe. Laskin also referred to large language models, such as Google's Gemini, Anthropic's Claude, and OpenAI's ChatGPT and GPT models. He mentioned that these models possess breadth but haven't been specifically trained for agency. Laskin, who has conducted AI research at the Berkeley Artificial Intelligence Research Lab and worked at Google DeepMind, joined forces with Ioannis Antonoglou, one of the creators of AlphaGo.

Antonoglou led the reinforcement learning from human feedback (RHLF) for Google's Gemini language model. In addition to Laskin and Antonoglou's startup, there are other companies focusing on building AI agents. Imbue, for example, has gained significant valuation and focuses on reasoning-focused agents. Decagon specializes in customer support, while Sybill targets sales representatives. Infrastructure providers, such as Emergence, AgentOps, Crew AI, and Phidata, enable enterprises to build their own AI agents. Furthermore, multi-agent systems have become a popular topic among venture capitalists. Acquisitions of agent startups have also taken place, such as Amazon hiring the cofounders of Adept, an AI agent startup that received substantial funding and licensed its technology. Laskin mentioned that he and Antonoglou could have remained at DeepMind to work on agents. However, they chose to pursue their own path, believing they can make faster progress towards their goals. Laskin added that their sense of urgency stems from the belief that a digital AGI, reminiscent of a universal agent, is approximately three years away.