Different researchers have been suggesting that Google is promoting the teaching of language to Artificial Intelligence (AI). Many have expressed concern that this practice can be risky and harmful to marginalized groups. These language models, such as GPT-3, have enabled chatbots to communicate very accurately with humans.
Particularly, because of the fact that it is often indistinguishable when a contact is with a machine or a human. Recently, a paper published by a Google research team has argued that the PaLM language model is very realistic. In addition to generating texts to perfection, he also interprets and explains jokes that are told by people.
Google and AI deepening
Google’s Artificial Intelligence team has presented a paper detailing the model’s capability to logical reasoning. Emphasis is also placed on the possibility of achieving other complex language tasks. This is in addition to the discernment of the context to carry forward the use of an appropriate technique in the chain of thought.
The system targets an improvement in AI analysis capabilities to analyze logical problems. In addition to this, steps are used that aim to simulate the thinking of a human being. Among the highlights, and worrying for some, is how this model can recognize and interpret jokes.
In some cases, the AI has managed to decipher those who arm themselves to generate bewilderment in the listener. The capacity of PaLM, as one of the largest language models ever built, is about 540 billion parameters. These model elements are trained in the learning process when the system receives new example data.