Language Imitation Games and the Arrival of Broad and Shallow AI
Large Language Models, or LLMs, are what artificial intelligence systems use to imitate, analyze, contextualize, interpret and decipher human language. The models are proving to be accurate and valuable in ways that inform or fulfill the well-meaning purposes for which they are most often employed, but still problematic in some significant ways, says Subbarao Kambhampati (pictured), a Fulton Schools professor of computer science and a former president of the Association for the Advancement of Artificial Intelligence. Some uses of LLMs have raised concerns about their tendency to reflect societal biases and stereotypes, Kambhampati writes in a commentary on the blog of the Association for Computing Machinery. There’s a need for LLMs to evolve to prevent the negative ramifications of their current limitations, he says, noting that researchers are beginning to explore remedies to such drawbacks.