Open a I, Google and Others Won't Publish the Code, Model or Training Data.

3min Snip

00:00

Play full episode

Summary

Transcript

Episode notes

Researchers have reported that they can extract sensitive data used to train large language models by posing careful questions. They retrieved personal contact information that gpt two had memorized verbatum. The best defence, they write, is simply to limit the sensitive information in the training data. All of these concerns suggest that, at a minimum, researchers should publicly document the training data that goes into their models. Some university teams andfir s, including gogle and facebook, have done this, but others, including invidia, microsoft and open a i, have not.

In 2020, the artificial intelligence (AI) GPT-3 wowed the world with its ability to write fluent streams of text. Trained on billions of words from books, articles and websites, GPT-3 was the latest in a series of ‘large language model’ AIs that are used by companies around the world to improve search results, answer questions, or propose computer code.

However, these large language model are not without their issues. Their training is based on the statistical relationships between the words and phrases, which can lead to them generating toxic or dangerous outputs.

Preventing responses like these is a huge challenge for researchers, who are attempting to do so by addressing biases in training data, or by instilling these AIs with common-sense and moral judgement.

This is an audio version of our feature: Robo-writers: the rise and risks of language-generating AI

Hosted on Acast. See acast.com/privacy for more information.