Extraterrestrial evaluation of AI

The field of Artificial Intelligence is haunted by a problem it can’t define. The only example of advanced intelligence we have are ourselves, which historically, has made the field of AI rather human-centric.

Many of the most remarkable AI achievements comes in domains which are about people:

• Natural language processing and generation
• Image processing and generation (images as percieved and used by humans)
• Game (TV, board and video) playing systems

One may argue these achievements are in fact of limited relevance with respect of the AI field in general, given their limited applicability to other tasks. A useful mental exercise to measure such relevance is to consider the extent to which other life forms from earth may benefit from AI breakthroughs. Or similarly, how an extraterrestrial being (ET) would evaluate our AI technology.

Squirrels are unaffected by language models, and an ET may be equally unable to value advances in that field. There are however AI contributions that can be considered to be of relevance for all earth life forms, and even, universally praised. For example:

• Self-driving cars represent a technology able to interact directly with the world.
• Protein folding models in particular, and AI applied to life sciences in general, is of relevance beyond humanity.

It’s a short list. If you think other contributions deserve to be in it, let me know.