David Silver has long been one of the most influential names in international AI research. Now he has taken a bold step out of DeepMind and into a new era — with one of the fastest billion-dollar fundraisings in British AI history.
Record Funding for Newly Founded Company
According to TechCrunch, Ineffable Intelligence, which Silver founded only a few months ago, has secured $1.1 billion in funding. The valuation is set at $5.1 billion — a remarkable figure for a company that has barely had time to find office space.
The company is based in the UK and builds on an ambition to develop artificial intelligence that does not rely on large amounts of human-produced data to learn. This is a direction that fundamentally differs from how most large language models are trained today.
Why Learn Without Human Data?
The rationale behind Ineffable Intelligence's approach is not just about technological originality — it's about resource scarcity. The research community has long warned that the availability of high-quality text data for AI training may be reaching a limit.
According to research overviews in the field, some analysts predict that high-quality text data could be virtually exhausted as early as 2026 if current training trends continue. Low-quality text data is expected to last longer — perhaps until between 2030 and 2050 — but the development necessitates new solutions.
A new model called Absolute Zero Reasoner, which trains itself with minimal human data, is said to have outperformed traditional AI on key benchmark tests — with over 22 percent better performance on one complex benchmark.
In practice, three main approaches are being experimented with: synthetic data generation, self-supervised learning, and training in simulated environments. All methods seek to reduce the need for costly and time-consuming human labeling and collection of real datasets.
Synthetic Data on the Rise
According to available research, the analysis company Gartner estimates that most AI models will be trained on synthetic data by 2030, and that synthetic data will constitute over 95 percent of training datasets for image and video models by the same year. Already in 2024, over 60 percent of data used in AI applications is said to have been synthetically generated.
However, it is important to emphasize that synthetic data is not without challenges. The research community points to the risk of bias from original real datasets being perpetuated, and to the phenomenon of “model collapse” — where models trained on their own outputs gradually degrade. These issues are active research fields.
Norwegian and Nordic Relevance
For Norwegian AI researchers and investors, Ineffable Intelligence's approach is interesting for several reasons. Norwegian business and public sectors are beginning to invest heavily in AI solutions but quickly encounter problems related to data access — especially in healthcare, where privacy concerns limit the use of real patient data. Technology that allows models to learn without such data could potentially open new opportunities.
Nevertheless, there is reason to be sober: Ineffable Intelligence is a very young company, and there is currently no publicly available technical documentation or independent verification of the company's progress. The valuation and capital raised primarily reflect confidence in Silver's professional reputation rather than demonstrated results.
What Happens Next?
David Silver is particularly known for his central role in the development of AlphaGo and AlphaZero at DeepMind — systems that learned to play board games at an extraordinarily high level through self-play, without human example data. It is plausible to believe that similar principles will shape the direction of Ineffable Intelligence.
International tech media, including TechCrunch, report on the funding round, but details about which investors are participating and what the money will specifically be used for are currently limited public information. 24AI will continue to follow the development.