Former Top Google Researchers Have Made A New Kind of AI Agent

A new kind of one artificial intelligence The agent was taught to understand how a program was established in the information of a company and how it was built to the latest product, but a more skilled program assistant and a small step towards the AI.

The new agent called Asimov, on the contrary, the best of Google was developed by a perplexity, on the contrary, on the contrary, on the contrary. Asimov code, emails, slab messages, project updates and other documents, read other documents to learn how to prepare all this ended.

Reflection’s last goal is to build something that owns something that other leading AI laboratories are working. Meta recently created a new Super Administrative Laboratory, prospective of large amounts It is interesting to join the researchers in their new efforts.

Williamsburg, New York’s Brooklyn neighborhood, New York, along the road from a pakingball club to see, plan to pass through the super headline before competition.

This is the simplest way to interact with the world, because it is an ideal way to build Supersmart AI agents, supersmart AI agents of the company, because it’s really the simplest, most natural way to master. Other companies are building agents using human user interfaces and See the InternetLaskin, who previously working on Gemini and Agents in Google Deprmind, is unlikely to be natural as a natural language model. Laskin will also produce more useful coding assistants to teach the development of the EI software.

Laskin says Asimov is designed to read more read code. “Everyone really pays attention to the generation of code,” he said. “But how useful agents can be done in a team administration.

Asimov actually consists of several small agents inside the trench coat. Agents work together to understand the entire code and understand the inquiries of users. Small agents get information and a greater thinking agent synthesizes this information to a response to a query.

Reflection claims that Asimov is already accepted to pre-lead some of the leading AI tools by some measures. In a reflective study, 82 percent of ASMMOUD of ASMOV, which works on large open source projects, 82 percent compared to an anthropical claude code of the company, 82 percent compared to the anthropic code code.

A computer scientist Daniel Jackson at the Massachusetts Technology Institute, said that the approach of the reflection is promised to provide a wider range of information. Jackson is likely to be seen in the benefits of approach, and the company’s survey is not enough to convince him. He notes that the approach can increase calculation costs and can create potential new security issues. “All this would read this personal messages,” he says.

Reflection, multifaceted approach reduces computing costs, and it says it uses a reliable environment that provides more security than some ordinary Saas tools.