Stanford HAI released its latest findings on September 7, 2026, stating that AI is significantly accelerating scientific discoveries by generating hypotheses, designing experiments, and analyzing data. In this field, the Biomni tool can analyze massive amounts of medical data to accelerate disease treatment; PsychAdapter can generate text simulating real individuals for training and personalized content; Evo 2, launched by Stanford’s team in February 2025, is the first ultra-large-scale DNA language model trained in the biological domain. Funded by the Hofmann-I research fund, Evo 2 was trained with 9 trillion base pairs and 40 billion parameters, enabling it to predict gene mutations and design new sequences. Despite risks such as bias, data quality issues, and unfair access, AI continues to break through human limitations with its powerful computing capabilities, promoting interdisciplinary collaboration and helping scientists explore major questions from the origin of the universe to climate simulation.
New AI tools generate hypotheses, design experiments, and analyze data, accelerating scientists’ discoveries in various fields. Biomni can analyze massive amounts of medical data and design experiments to accelerate disease treatment; PsychAdapter generates text simulating real individuals for training and personalized content; Evo 2 is the first ultra-large-scale DNA language model trained in the biological field, launched by Stanford team in February 2025. This model, funded by the Hofman-Ei Research Fund and trained on 9 trillion base pairs and 40 billion parameters, can predict gene mutations and design new sequences. Despite risks such as AI bias, data quality issues, and unfair access, AI is breaking through human limitations by providing unprecedented computing power, promoting interdisciplinary collaboration, and helping scientists explore grand questions from the origins of the universe to climate simulation.