Artificial Intelligence Accelerates Scientific Research But Does Not Impact Laboratory Experiments at the Same Rate
A new report by Google, Google DeepMind, and MIT reveals that artificial intelligence accelerates theoretical work and idea generation, but fails to provide the same speed increase in physical laboratory experiments and drug discovery processes.
While artificial intelligence technologies contribute to solutions in areas such as mathematics, they fail to reach the same speed in disciplines like biology and chemistry, creating piles of untested hypotheses in front of scientists.
Scope and Data of the Research
Published by Google, Google DeepMind, and MIT FutureTech within the Massachusetts Institute of Technology, the report revealed the gaps in this field by analyzing a survey of 637 scientists, 15 million Gemini conversations, and more than 2,600 specialized models.
Experiment and Data Collection Bottleneck
While 44 percent of the scientists participating in the research stated that the main research bottleneck has shifted toward physical experiments and data collection in the last two years, 41 percent stated that there is an increasing accumulation of untested hypotheses.
Verification of Artificial Intelligence Outputs
Researchers spend significant time checking artificial intelligence outputs; 89 percent of those who state that they save time spend more than one-tenth of this gained time verifying artificial intelligence predictions.
Causes of Interdisciplinary Differences
Experts emphasize that fields like mathematics have clear rules for checking proofs, whereas fields like medicine lack absolute certainty, and physical laboratory automation faces obstacles such as high costs and safety regulations.