Google’s “AI Scientist” Faces Heavy Criticism From Human Scientists!

Google recently unveiled its Gemini 2.0-based tool, designed to aid scientists by using “advanced reasoning” to generate hypotheses and research plans. According to Google, the tool mimics the scientific method, powered by multiple Gemini “agents” that debate and refine ideas over time. The company claims that this tool would give scientists “superpowers,” enabling them to come up with groundbreaking research ideas with ease. However, not everyone is convinced by the hype surrounding this AI tool.

The Buzz and the Criticism

While Google’s announcement made waves in the scientific community, many experts are skeptical about its real-world usefulness. Alan Karthikesalingam, an AI researcher at Google, described the tool as a way to “supercharge science.” Some early users, including biomedical researchers at Imperial College London, expressed excitement about its potential. They claimed it could dramatically speed up scientific progress. However, others aren’t as enthusiastic.

Sarah Beery, a computer vision researcher at MIT, told TechCrunch that while the tool is interesting, it’s unlikely to be a game-changer for the scientific community. “I’m not sure there is demand for this type of hypothesis-generation system,” she said.

The Claims and the Reality

Google proudly boasted that the AI tool had come up with novel ways to repurpose drugs for treating acute myeloid leukemia. However, critics quickly shot down these claims. Pathologist Favia Dubyk, from Northwest Medical Center-Tucson, questioned the legitimacy of the results, stating that the ideas generated were too vague to be taken seriously.

Furthermore, Google claimed that the AI discovered new treatments for liver fibrosis, but this too was debunked. Steven O’Reilly, from UK biotech company Alcyomics, pointed out that the drugs suggested by the tool were already well-known and established in the field.

Potential Advantages, but Not for Hypothesis Generation

Despite these criticisms, the tool does have some potential benefits. It can sift through vast amounts of scientific literature and summarize the findings, saving researchers valuable time. However, this is where the tool’s limitations become apparent. Like many large language models, it is prone to “hallucinations” or generating false or misleading information, which makes its usefulness questionable.

Google markets this tool as a hypothesis-generating machine, a lofty claim that many scientists find unnecessary. Lana Sinapayen, an AI researcher at Sony, pointed out that generating hypotheses is one of the most enjoyable aspects of research. “Why would I want to outsource my fun to a computer?” she asked, criticizing the tool for trying to automate a process that scientists cherish.

Conclusion

While Google’s Gemini 2.0-based tool may have some useful applications, it is far from the revolutionary breakthrough the company claims. Many scientists question the tool’s ability to generate meaningful hypotheses, and skepticism remains high. For now, it seems that hypothesis generation will remain a task best suited for human creativity and expertise.