20VC: Scale, Surge, Turing, Mercor: Who Wins & Who Loses in Data Labelling | Is Revenue in Data Labelling Real or GMV? | Why 99% of Knowledge Work Will Go and What Happens Then? | Why SaaS is Dead in a World of AI with Jonathan Siddharth @ Turing
20VC: Scale, Surge, Turing, Mercor: Who Wins & Who Loses in Data Labelling | Is Revenue in Data Labelling Real or GMV? | Why 99% of Knowledge Work Will Go and What Happens Then? | Why SaaS is Dead in a World of AI with Jonathan Siddharth @ Turing
20VC: Scale, Surge, Turing, Mercor: Who Wins & Who Loses in Data Labelling | Is Revenue in Data Labelling Real or GMV? | Why 99% of Knowledge Work Will Go and What Happens Then? | Why SaaS is Dead in a World of AI with Jonathan Siddharth @ Turing
In this forward-looking discussion, Jonathan Siddharth of Turing explores the transformative impact of AI on knowledge work, enterprise adoption, and the future of software. Moving beyond conventional narratives, he unpacks how reinforcement learning, data loops, and agentic systems are redefining what’s possible in AI development and deployment.
Turing is pioneering a shift from traditional AI training to reinforcement learning-based agents that operate in simulated real-world environments, enabling automation of complex workflows. The company emphasizes data-driven feedback loops as the new competitive moat, surpassing raw technology. Enterprises struggle with outdated infrastructure and unstructured data, slowing AI adoption, while startups and front-office functions lead in integration. With AI models now matching or exceeding human performance in many cognitive tasks, up to 99% of knowledge work could be automated within a decade, drastically increasing individual productivity. This evolution threatens traditional SaaS models, as custom AI built on foundation models reduces reliance on packaged software. Meanwhile, smartphones are poised to become primary AI interfaces, acting as cognitive extensions. Despite challenges, long-term trends favor research-driven companies leveraging proprietary data, with breakthroughs expected in areas like drug discovery and self-improving AI systems.
06:17
06:17
Agents require teaching the model tool use
15:03
15:03
A smaller, fine-tuned model can be faster and more accurate than a giant world model for specific enterprise workflows.
20:16
20:16
Convincing financial firms to adopt AI is easier when tied to profit rather than cost-saving.
24:01
24:01
Top AI models can produce work indistinguishable from human experts in simple tasks
35:30
35:30
Alex Wang from Scale AI is respected for his early insight into data's role in AI.
43:32
43:32
Achieving AGI could unlock unprecedented intellectual and economic value
45:20
45:20
If you believe in AGI, investing in SaaS apps is unwise because SaaS is over.
54:23
54:23
AI devices will become an extension of the brain
57:07
57:07
AI will improve incrementally, not through rapid takeoff
