AI at the Service of Science: A Transformation in Progress | Best generative ai courses | Generative ai use cases in healthcare | Generative ai use cases in banking | Turtles AI

AI is transforming science, with applications ranging from structural biology to computer science, accelerating scientific research. In the face of complex and growing challenges, its use could optimise progress towards global goals, but requires targeted policies to realise its full potential.

Key points:

• AI is radically changing the scientific landscape, accelerating research in fields such as genomics and biology.
• The opportunities for science arising from AI are rapidly growing, but require collective effort to be fully exploited.
• AI models are reducing research time and costs, as demonstrated by AlphaFold 2 in biology.
• Although the benefits are clear, effective public policies must be implemented to avoid risks related to scientific reliability and creativity.

A quiet revolution is taking place in laboratories around the world, where AI is rapidly becoming a key ally for scientists. Today, one in three postdoctoral researchers uses large language models to optimize everyday tasks such as reviewing scientific literature, programming and editing data. The recent awarding of the Nobel Prize in Chemistry to Demis Hassabis and John Jumper, pioneers of AlphaFold 2, is a testament to the impact that AI is having on science. This recognition recognizes their work, which has allowed them to predict the structure of proteins, paving the way for new drugs and materials designed with AI.

In the scientific world, the adoption of AI is seen as a necessity to address increasingly complex challenges, such as managing the vastness of academic literature and designing experiments that, without AI, would require very long times and enormous resources. The increasing difficulty of keeping up with the increase in knowledge and research demands has pushed the scientific community to resort to advanced technological solutions. AI, with its deep learning algorithms, is establishing itself as a privileged tool for reducing scientific development times and multiplying productivity, tackling problems ranging from designing more effective proteins to predicting climate change.

Despite progress, the path towards full integration of AI in science is still under development. Although a growing number of scientists are using tools based on language models, the adoption of methodologies entirely focused on AI remains limited. This phenomenon can be attributed to concerns about the reliability of the results generated by AI models, but also to the need for adequate training to make the most of its potential. The scientific community, therefore, is called upon to find a balance between innovation and verification, to avoid risks that could compromise the quality of research.

There is no shortage of concrete examples of how AI is already transforming research. A case in point is AlphaFold, which has drastically reduced the time and costs required to determine the three-dimensional structure of proteins. While traditionally, X-ray crystallography experiments to determine the structure of a protein could take years and cost close to $100,000, AlphaFold has made a database of over 200 million predicted protein structures freely available, having a significant impact on structural biology. Another example of AI application is in the field of genomics, where machine learning models are being used to decipher genetic sequences and predict diseases with a precision that would never have been possible without the help of these tools.

Despite promising results, the benefits of using AI in science are not yet guaranteed. Some scientists have had to deal with the reliability of models, which sometimes have not produced the expected results. The most common concerns are the potential lack of creativity and the reliability of AI-generated solutions, which are still being debated. The growing availability of these tools, however, suggests that, if well used, AI can become a fundamental resource to accelerate scientific progress in many areas, solving issues of scale and complexity that today seem intractable.

To ensure that the benefits of AI can be fully realized, it is crucial to adopt targeted public policies that direct the use of AI towards high-impact sectors. Institutions such as the US Department of Energy, the European Commission and the US National Academies have already recognized the importance of investing in AI for science. However, no country has yet implemented a comprehensive national strategy to promote and regulate the use of AI in science. It is essential that future policies are able to stimulate scientific innovation, responding effectively to global challenges.

AI has the potential to become a driver of unprecedented progress in many areas of science, but a joint effort between scientists, governments and institutions is needed to ensure that its use is responsible, effective and beneficial for society as a whole.

https://deepmind.google/public-policy/ai-for-science/