Google said that when a comedian says something sarcastic or controversial, usually the audience can discern the tone and know it's more of an exaggeration, something that's learned from years of human interaction.
Apparently, its PaLM, or Pathways Language Model, learned a person was joking without being explicitly trained on humour and the logic of jokes. After being fed two jokes, it was able to interpret them and spit out an explanation.
In a blog post, Google shows how PaLM understands a novel joke not found on the internet. The reason it was not found on the internet was probably because it was not funny. Basically, it was Google has just hired a sperm whale for its TPU team because it wanted someone who could communicate between pods.”
The AI worked out that it was a pun between whales who live in pods and TPU which also uses pods. Yep, about as funny as a road accident. But the fact that the AI worked out that the right was supposed to be joking is more important than he was funny.
PaLM is Google's largest AI model to date and trained on 540 billion parameters. It can generate code from text, answer a math word problem and explain a joke. It does this through chain-of-thought prompting, which can describe multi-step problems as a series of intermediate steps.
Of course, it does not mean that we are at the point where anyone in America will ever get European humour or sarcasm. They still think fake news is satire, Donald Trump was joking, and “Friends” was ever funny.