A wave of new AI research labs, or neolabs, has emerged in 2024 as founders depart from major companies like OpenAI, Google DeepMind, and Meta. These startups are securing significant capital despite lacking public products, betting that current AI development methods—relying on massive language models and internet data—are insufficient for achieving AGI.
Investors are backing these ventures, anticipating that small, focused teams can identify breakthroughs that large corporations miss due to the innovator's dilemma. Notable funding rounds include Flapping Airplanes raising $1 billion at a $6 billion valuation in October, and Ineffable Intelligence securing a $1.1 billion seed round at a $5.1 billion valuation in April.
The research focus of these labs varies, with many targeting the limitations of current models in continual learning and data efficiency. For instance, Core Automation aims to develop models that update weights during deployment using significantly less training data, while Adaption is exploring efficient AI that avoids the scaling of model parameters.
Other labs are pursuing alternative learning paradigms. Flapping Airplanes is developing models that learn from limited data, drawing on biological principles. Ineffable Intelligence, led by David Silver, focuses on systems that learn through self-experience rather than human examples, while Ndea utilizes program synthesis to derive rules from few-shot examples.
The industry is seeing high-value exits and acquisition interest, as established tech giants seek to integrate the expertise of these teams. Examples include Microsoft hiring the core team from Inflection, Google acquiring founders from Character.ai, and AMD pursuing an acquisition of World Labs for approximately $8.2 billion.
