Take part in our daily and weekly newsletters to get the latest updates and exclusive content for reporting on industry -leading AI. Learn more
Since Mira Murati Openaai left last autumnMany were surprised about the next step of the former CTO. Well, now we know (or at least have a rough idea).
Murati has made itself X today to announce her new company Memory laboratoryA AI research and product company with the aim of “driving AI by making it largely useful and understandable through solid foundations, open science and practical applications”.
Muti posted: “We build three things: help people to adapt AI systems to work for their specific needs. Development of strong foundations to build more powerful AI systems; Promotion of a culture of open science that helps the entire field to understand and improve these systems. “
The team of around two dozen engineers and scientists from Machine is reinforced pioneer John Schulman and Chatgpt Co-creator Barret Zoph-, who could position the startup, with other open-ended coats. and to make development.
At the end of this article, the newly started company of the company X account @Thinkymachines had already accumulated about 14,000 followers.
Model intelligence, multimodality, strong infrastructure
Memorial not with the supercomputer and AI company of the 1980s that are no longer existing—there are still no specific examples of intended projects, but suggests a comprehensive focus on multimodal skills, the cooperation between human-AI (in contrast to pure actual systems) and strong infrastructure. The goal is to build “more flexible, more adaptable and personalized AI systems”, the company writes about its New website.
Multimodality is “critical” for the future of AI, says Thinking Machines, as it enables more natural and efficient communication that captures the intention and supports deeper integration. “Instead of concentrating exclusively on completely autonomous AI systems, we look forward to building multimodal systems that work with people,” writes the company.
In the future, the startup models will create “at the border of skills” in areas such as science and programming. Model Intelligence will be the key, as will the quality of the infrastructure. “In the long run, we want to build things up properly to maximize both productivity and security instead of taking abbreviations,” writes Thinking Machines.
The company will work with the AI community and publish the technical blog posts, papers and code. It will also require an “empirical and iterative” approach to AI security and commit to maintain a “high security bar” to prevent abuse, carry out the monitoring of rotting teams and after the introduction and best practice, data records, code and model specifications to share.
Expansion of a decorated team
Machines have an impressive team of scientists, engineers and builders behind Top-KI.
The startup wants to expand this basis. It is currently looking for a research program manager as well as product farmers and machine learning (ML). The goal is to put together a “small, highly cold team” from both doctoral students and the autodidact, the company writes. Those interested should Apply here.
Another Openai competitor?
Murati stepped back abruptly from Openai in September 2024 after the unexpected departure of other managers, including Schulman and co-founder of the former chief scientist Ilya Sutskever- after he came to the company in 2018 and rose to CTO in 2022 (the year we the groundbreaking chatt brought to CTO). .
At the time, She teased on x: “I go away because I want to create time and space to carry out my own exploration.”
Her next step was a topic in view of her reputation as a constant company force during the turbulent period of Openas at the end of 2023 when trying to displace CEO Sam Altman shortly before the board. Internally, she was regarded as a pragmatic guide, who kept openly opened by uncertainty. Your decision to go on alone is followed by a broader change in the AI research landscape in which the breakneck race is always dodging of an era of the ACI, the agent systems and the implementation of the real world.
Your remedial engine announcement takes place in the middle of a flood of new model functions and the benchmark break. Openaai continues to lead innovative breakthroughs-and the new O3-deep research mode, which scrolls the web and extensive reports curated-also confronted with strong competition in all directions. Only today, for example, Xai published Grok 3which competes with the performance of GPT-4O from the performance of GPT-4O.
Started with Openai co-founder Ilya Sutskever Safe superintelligence And other industry veterans who set up their own way now asks the question of where memorial machines put their bets in this developing landscape.
Source link
