Ex-DeepMind Scientist David Silver Building AI Startup Ineffable
Former AlphaGo architect David Silver discusses his startup, Ineffable, which aims to build self-teaching AI systems using reinforcement learning.
Key points
- David Silver is building a startup named Ineffable.
- The company is developing an AI "super learner" based on reinforcement learning.
- Silver views this approach as an alternative to large language models.
- He cites a moral obligation to continue advancing artificial intelligence technology.
Former Google DeepMind scientist David Silver is building an artificial intelligence startup called Ineffable, according to a report by the Financial Times. In an interview with journalist Madhumita Murgia, Silver discussed the technical direction of the venture, focusing on creating machines that can teach themselves using reinforcement learning.
Silver, widely recognized for his role in creating AlphaGo, explained that Ineffable is focused on building what he describes as an AI "super learner." The core technology relies on reinforcement learning rather than the generative methods currently dominating the industry. This strategic focus suggests an attempt to explore different pathways for machine intelligence beyond the current standard.
Technical Approach and Alternatives to LLMs
The Financial Times report highlights that Silver views reinforcement learning as a viable alternative to large language models. While most recent AI advancements have centered on scaling language models, Silver’s new work emphasizes systems that can teach themselves through interaction and reward signals. This method mirrors the techniques used in AlphaGo, which mastered the board game Go through extensive self-play and iterative improvement.
Silver’s involvement signals a significant shift in how some leading researchers view the trajectory of AI development. By prioritizing reinforcement learning, Ineffable aims to create machines that learn from experience rather than relying solely on vast datasets of text. This distinction could influence how the industry approaches general intelligence and autonomous decision-making in the future.
Who Is Affected
The development primarily affects the research community and investors tracking post-DeepMind innovations. Silver’s work at Ineffable draws attention to the mobility of top-tier AI talent. Companies and academic institutions monitoring alternative AI architectures may find this work relevant to their own strategic planning and talent acquisition efforts.
Furthermore, the broader technology sector may observe changes in funding and research priorities as high-profile scientists pursue independent ventures. The focus on self-teaching systems could inspire other researchers to revisit reinforcement learning techniques that were previously overshadowed by the rise of generative AI tools.
What should you do?
Security and technology professionals should monitor developments from Ineffable to understand potential shifts in AI capabilities. Tracking the technical publications and public statements from Silver’s team can provide early insights into emerging threats or defensive opportunities. Organizations should also evaluate their current AI strategies to ensure they are prepared for diverse technological approaches.
Background: Large language models
Large language models are a specific type of neural network trained on text. Generative AI is the broader umbrella term for any system that creates new content, including images and code. Not all generative AI uses large language models, but most modern text generators do.
What to do and how to stay safe: David Silver
- Monitor academic and industry publications from Ineffable for technical details on their reinforcement learning models.
- Evaluate internal AI security policies to ensure they account for self-teaching autonomous systems.
- Track talent movements from major AI labs to identify emerging research trends and potential partnerships.
- Review current vendor assessments to determine if competitors are adopting similar non-LLM architectures.
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Who is building the startup Ineffable?
David Silver, a former scientist at Google DeepMind, is building the startup Ineffable.
What technology is Ineffable using?
The company is using reinforcement learning to build an AI system that can teach itself.
How does this differ from current AI trends?
Silver views this reinforcement learning approach as an alternative to large language models.
