OpenAI launches tool to understand how AI language models work

OpenAI released a tool to find out and explain the inner workings of Large Language Models (LLMs) in Artificial Intelligence (AI), which uses GPT-4 to analyze the models, explain them, and anticipate what will be potential problems with learning systems. AI.

Language models in the field of AI, such as OpenAI’s GPT, reach conclusions and produce surprising results independently, which is why they are increasingly implemented in more fields and day-to-day situations, learning more and becoming smarter.

However, human understanding of how these language models work internally is limited. It is difficult to know for sure why a model responds the way it does for each situation. It is also difficult to find out how they increase their knowledge, if they do it in a biased way or if they use deception.

In order to better understand how LLMs work, OpenAI released a tool with which it will be possible to analyze which parts of the model are responsible for each function and their behaviors.

This tool, as OpenAI explains in a statement on its website, is based on an automated process that uses the GPT-4 language model to “produce and score natural language explanations of the behavior of neurons” of a specific LLM, in this GPT-2 case and, later, apply it to neurons in another language model.

LLM models are made up of neurons that identify specific patterns in the text that the user enters in order to generate a response accordingly. With this in mind, the OpenAI tool divides the GPT-2 model into different pieces and performs a three-step process.

Source: dpa

(Reference image source: Eduardo Parra, Europapress)

Visit our news channel on Google News and follow us to get accurate, interesting information and stay up to date with everything. You can also see our daily content on Twitter and Instagram