Keep pulling the thread on Jonathan Siddharth.
Jonathan Siddharth believes the era of data labeling companies is over and has been replaced by the era of research accelerators.
Jonathan Siddharth predicts that all knowledge work will eventually be automated.
Jonathan Siddharth believes the traditional SaaS business model is "completely over."
Turing works with seven out of the eight major frontier AI labs.
The dominant paradigm for training AI agents has shifted to reinforcement learning, which requires building complex "RL environments."
Jonathan Siddharth believes in a "slow, steady takeoff" for AGI and superintelligence, rather than a rapid takeoff scenario.
Turing provides custom, fine-tuned AI model services to enterprise clients including Disney, Pepsi, BlackRock, Fiserv, and Johnson & Johnson.
Research indicates that smaller language models (0.5B to 10B parameters) can be faster and more accurate than large, trillion-parameter models for specific enterprise tasks like insurance underwriting.
According to OpenAI's GDPVal study, top AI models produced work indistinguishable from human experts approximately 50% of the time.
In OpenAI's GDPVal study, Claude 4 Opus was identified as the top-performing model, with GPT-5 also performing well.
Jonathan Siddharth believes that data-driven feedback loops will be a key competitive moat for companies in the age of AI.
NVIDIA derives 39% of its revenue from just two clients and approximately 50% from four clients.