Google’s new artificial intelligence is capable of understanding jokes

Pathways is one of the first Artificial Intelligences to understand the jokes of humanity.

Advertisements

Understanding jokes is often even difficult for a human, but it is no longer impossible for a machine: Google has presented the first artificial intelligence  to achieve it.

This is Pathways, an AI developed by the Google Research team capable of understanding more than 540 billion patterns and concepts.

How do you do it?

Pathways maintains a 6144-chip brain and has been trained by thousands of source code conversations on Github, as well as information drawn from virtual libraries such as Wikipedia itself.

This is “the largest TPU configuration” used in history, according to the company. PaLM , its language model,  also achieves a training efficiency of 57.8% in the use of ‘hardware’ flops, “the highest achieved so far for language models at this scale”.

Among the improvements it offers with respect to other models is that of being able to distinguish, for example, causes and effects in different situations, in addition to correlating efforts and results.

It is also capable of guessing movies from an emoji or solving simple mathematical reasoning problems. Not the calculations, but the logic behind them.

But through  PaLM, Pathways is also able to understand and explain jokes he’s hearing for the first time in his existence.

Lastly, Google also points out that PaLM is capable of programming by translating code “from one language to another”, writing code based on “a natural description of the language” and “fixing compilation errors”.