- Thomas Wolf said AI excels at following instructions but struggles to create new knowledge.
- AI needs to question its training data and take counterintuitive approaches, the Hugging Face exec wrote on X.
- Wolf’s comments come as tech focuses on agentic AI.
AI excels at following instructions — but it’s not pushing the boundaries of knowledge, said Thomas Wolf.
The chief science officer and cofounder of Hugging Face, an open-source AI company backed by Amazon and Nvidia, analyzed the limits of large language models in a Thursday post on X.
He wrote that the field produces “overly compliant helpers” rather than revolutionaries.
Right now, AI isn’t creating new knowledge, Wolf wrote. Instead, it’s just filling in the blanks between existing facts — what he called “manifold filling.”
Wolf argues that for AI to drive real scientific breakthroughs, it needs to do more than retrieve and synthesize information. AI should question its own training data, take counterintuitive approaches, generate new ideas from minimal input, and ask unexpected questions that open new research paths.
Wolf also weighed in on the idea of a “compressed 21st century”— a concept from an October essay by Anthropic’s CEO Dario Amodei, “Machine of Loving Grace.” Amodei wrote that AI could accelerate scientific progress so much that discoveries expected over the next 100 years could happen in just five to 10.
“I read this essay twice. The first time I was totally amazed: AI will change everything in science in five years, I thought!” Wolf wrote on X. “Re-reading it, I realized that much of it seemed like wishful thinking at best.”
Unless AI research shifts gears, Wolf warned, we won’t get a new Albert Einstein in a data center — just a future filled with “yes-men on servers.”
Wolf did not respond to a request for comment, sent outside standard business hours.
The rise of agentic AI
Wolf’s comments come as the AI world focuses on agentic AI.
Sam Altman, the CEO of OpenAI, has predicted that this may be the year the first “agents” — a set of artificial intelligence tools that can perform tasks independently — “join the workforce.”
“If 2024 was the year of LLMs, we believe 2025 will be the year of agentic AI,” Praveen Akkiraju told Business Insider in January. He’s a managing director at Insight Partners, a VC firm whose agentic plays include Writer, Jasper, and Torq.
Investors are betting big on this idea. According to PitchBook data, startups exploring the application of agents raised $8.2 billion last year.
Unlike AI assistants, which mainly retrieve and summarize information, agents can break down complex tasks, make decisions, and refine their approach based on outcomes.
Researchers have also used AI to achieve scientific breakthroughs.
Oxford professor Matthew Higgins used AlphaFold2, an AI tool from Alphabet-owned DeepMind, to crack the shape of a key malaria protein — something his lab had struggled with for years. That breakthrough led to an experimental malaria vaccine being tested in people.
Without AlphaFold, “we’d probably still be trying, to be honest,” Higgins told BI in 2023.
Read the full article here
