AI in protein design and disease diagnosis

Advances in artificial intelligence allow scientists to design entirely new molecules in seconds instead of months of research. In July this year, DeepMind Technologies Limited presented the latest version of AlphaFold, a protein structure prediction software that has predicted the structure of every protein known to science. Scientists and companies using AI to design proteins would like to create proteins that can, for example, clean up toxic waste and help treat diseases.

Most current AI diagnostic models learn to recognise diseases and patient conditions from chest x-rays previously marked by specialists. However, this is a time-consuming process. The new model, CheXzero, can “learn on its own” from existing medical reports that specialists have written in normal language. As a result, after reviewing thousands of chest X-rays and accompanying clinical reports, CheXzero, developed at Harvard Medical School, has learned to diagnose disease as accurately as radiologists.

With access to AI art generators becoming more widespread, artists are raising questions about the ability of artificial intelligence to mimic their work. They also point out that these programmes primarily make money for large corporations. AI massively creates on-demand graphics using machine learning algorithms in response to text prompts, and from images from the web or public data collections it is able to replicate the style of individual artists with remarkable accuracy.