Google revealed a major step forward in the development of an AI system capable of handling a wide range of activities, including complicated learning and reasoning. The Pathways Language Model, or PaLM, is the name of the new system.
PaLM is capable of outperforming existing AI state-of-the-art as well as outperforming humans in language and reasoning tests. This includes Few-Shot Learning which imitates how humans learn new things and combine diverse bits of knowledge to tackle new challenges that have never been seen before and its benefit is a machine that can use all of its knowledge to solve new challenges, one example of this skill in PaLM is its ability to explain a joke that it has never heard before.
Another skill is Inference Chaining in order to construct an answer, several concepts and entities must be correctly connected and considered.
However, the researchers warn that they can’t get over the restrictions that come with large-scale language models, which can unintentionally lead to unethical outcomes such as discrimination, exclusion and toxicity, information hazards, misinformation harms, malicious uses, human-computer interaction harms, automation, access, and environmental harms.