Could we ever decipher an alien language?

Abstract

In this article, I discuss how research in emergent communication can help us understand the development of language. By giving AI agents simple communication tasks and minimal restrictions, we can study how they evolve their own ways of sharing information. Using tools from information theory, we analyze patterns in AI conversations to decode their linguistic structure, similar to how archaeologists piece together ancient languages. Our recent work, to be presented at NeurIPS, demonstrates methods for reverse-engineering AI language and syntax by matching patterns between visual contexts and communication logs. I also theorise that the methods we’re developing could one day prove valuable in the field of xenolinguistics - the study of alien languages. Read it here!

Olaf Lipinski

PhD Student in Artificial Intelligence

I study how autonomous agents develop langauge from scratch.