Subscribe now

Letter: AI's advances really aren't quite as great as we think (1)

Published 9 August 2023

From Eric Kvaalen, Les Essarts-le-Roi, France

The article on using artificial intelligence to tackle scientific challenges paints too rosy a picture of its accomplishments up to now. AlphaFold sometimes inaccurately predicts protein folding. The matrix multiplication methods found by AI only achieve a 10 to 20 per cent improvement for certain small matrices and only when using certain hardware. The sorting algorithms found by AI were 70 per cent faster only for a list of five items. The concrete formulation found by AI entails 40 per cent less carbon dioxide emissions only when compared with a form of “standard” concrete that hasn’t been used much for the past 25 years (29 July, p 32).

Issue no. 3451 published 12 August 2023

Sign up to our weekly newsletter

Receive a weekly dose of discovery in your inbox. We'll also keep you up to date with New Scientist events and special offers.

Sign up
Piano Exit Overlay Banner Mobile Piano Exit Overlay Banner Desktop