Thomas Wolf, co-founder and Chief Science Officer of AI company Hugging Face, has sounded a warning about the limitations of current artificial intelligence (AI) development, stating that the technology is at risk of becoming "yes-men on servers" without a breakthrough in AI research. In an essay published on Thursday, Wolf expressed his concerns that current AI development paradigms won't yield AI capable of outside-the-box, creative problem-solving – the kind of problem-solving that wins Nobel Prizes.

Wolf's views stand in contrast to those of other AI industry leaders, such as OpenAI CEO Sam Altman, who has predicted that "superintelligent" AI could "massively accelerate scientific discovery." Similarly, Anthropic CEO Dario Amodei has forecast that AI could help formulate cures for most types of cancer. However, Wolf argues that current AI systems are not incentivized to question and propose ideas that potentially go against their training data, limiting them to answering known questions.

According to Wolf, the main issue with AI today is that it doesn't generate any new knowledge by connecting previously unrelated facts. Even with access to most of the internet, AI as we currently understand it mostly fills in the gaps between what humans already know. Wolf believes that AI labs are building what are essentially "very obedient students" – not scientific revolutionaries in any sense of the phrase.

Wolf's concerns are shared by some AI experts, including ex-Google engineer Francois Chollet, who has argued that while AI might be capable of memorizing reasoning patterns, it's unlikely it can generate "new reasoning" based on novel situations. Wolf thinks that the "evaluation crisis" in AI is partly to blame for this state of affairs, pointing to benchmarks commonly used to measure AI system improvements, most of which consist of questions that have clear, obvious, and "close-ended" answers.

As a solution, Wolf proposes that the AI industry "move to a measure of knowledge and reasoning" that's able to elucidate whether AI can take "bold counterfactual approaches," make general proposals based on "tiny hints," and ask "non-obvious questions" that lead to "new research paths." While acknowledging that figuring out what this measure looks like will be a challenge, Wolf believes it could be well worth the effort.

Wolf's essay serves as a call to action for the AI industry to re-evaluate its approach to developing AI systems. By recognizing the limitations of current AI development paradigms, Wolf hopes to inspire a shift towards creating AI that can truly think outside the box and drive scientific progress. As he puts it, "the most crucial aspect of science [is] the skill to ask the right questions and to challenge even what one has learned." The question now is whether the AI industry will heed Wolf's warning and strive to create AI that can truly make a difference.