<div id="readability-page-1" class="page">A star Chinese researcher laid off by Facebook owner Meta Platforms has co-founded a start-up focused on self-improving artificial intelligence systems, joining a wave of US and Chinese firms developing models capable of autonomously refining their own code and reasoning. Tian Yuandong, former research scientist director at Meta’s FAIR (Fundamental AI Research) team, launched Recursive Superintelligence alongside seven other co-founders. The company said on Wednesday it had raised more than US$650 million at a US$4.65 billion valuation. The round was led by venture capital firms GV and Greycroft, with “major participation” from US chipmaking giants Nvidia and Advanced Micro Devices, the start-up said. Often referred to as “recursive self-improvement”, AI systems that can automate the AI research and development process have long been seen as the holy grail of AI research. A company that can first achieve this milestone would, in theory, be able to rapidly extend its lead over competitors as its AI development efforts accelerated exponentially. “[T] he fastest path to superintelligence will be realised by AI that recursively improves itself, and does so via open-ended algorithms that drive endless innovation,” the company said in a statement.  With offices in San Francisco and London, the company has only around 25 employees, although it said the number was “growing”. Tian is one of several Recursive co-founders to have previously led AI research at a US AI giant, with others previously working at OpenAI and Google. The Shanghai-native made headlines in October when he accused his former employer Meta of making him and hundreds of others scapegoats for the company’s subpar AI development efforts at the time. Recursive self-improvement has emerged as a top priority for the AI industry this year. In January, another US start-up founded by two former Google researchers called Ricursive Intelligence raised US$335 million at a US$4 billion valuation to work on developing AI systems that can automate AI chip hardware development. In China, meanwhile, efforts are also under way to automate the AI development process. On Tuesday, Zhipu AI’s co-founder Tang Jie wrote on X that “self-evolution” was the “endgame” for the AI industry: “I suspect that models like [Anthropic’s] Claude may have already achieved a baseline for self-training: writing their own code, cleaning their own data, generating synthetic data, and then training on it.” However, experts have also warned that recursively self-improving AI could result in humans losing control over AI systems, leading to unprecedented economic and social impacts. In a blog post on Wednesday, another Recursive Superintelligence co-founder, American computer scientist Jeff Clune, wrote that there were “tremendous risks associated with this extremely powerful technology”, but concluded that “the best thing I can do is to work hard to make it as safe, aligned and net beneficial for humanity as possible”. </div>

META

<p>Tian Yuandong, a former Meta researcher, co-founded Recursive Superintelligence, a start-up focused on self-improving AI systems, raising over US$650 million at a US$4.65 billion valuation. The company aims to automate AI development, a key goal in the industry. With offices in San Francisco and London, it has around 25 employees and is backed by major investors like Nvidia and AMD. Experts warn of the risks associated with such powerful technology, emphasizing the need for safety and alignment with human values.</p>

- A former Meta researcher co-founded Recursive Superintelligence, focusing on self-improving AI systems and raising over US$650 million at a US$4.65 billion valuation.  
- The startup aims to automate AI research and development, a process viewed as the holy grail of AI innovation, potentially allowing rapid advancements compared to competitors.  
- Concerns exist regarding the risks of such AI systems, with experts warning about potential loss of human control and significant economic and social impacts.

SCMP

Meta Platforms

iShares Expanded Tech Software Sector ETF

Invesco AI and Next Gen Software ETF

GraniteShares 2x Long META Daily ETF

iShares Semiconductor ETF

VanEck Semiconductor ETF

Invesco PHLX Semiconductor ETF

Direxion Daily Meta Bull 2x Shares

VG Info Tech

Global X Cloud Computing ETF

Spdr Select Tech

Invesco S&P 500 Equal Weight Tech ETF

First Trust DJ Internet Fd

Roundhill META WeeklyPay ETF

SPDR S&P Software

Direxion Semicon Bull 3X

Invesco Semiconductors ETF

SPDR S&P Semicon

ISHRS S&P Glb It

Alphabet

Alphabet - C

NVIDIA

OpenAI

KNOWLEDGE ATLAS

NVIDIA Corporation

Ex-Meta Chinese star joins race for self-improving AI with US$4.6b start-up