The core of an artificial intelligence program like ChatGPT is something called a large language model: an algorithm that mimics the form of written language.
While the inner workings of these algorithms are notoriously difficult to decipher, the basic idea behind them is surprisingly simple. They are trained on mountains of internet text, by going through them a few sentences or paragraphs at a time, repeatedly guessing the next word (or word fragment) and then grading themselves against the real thing.
To show you what this process looks like, we trained six tiny language models starting from scratch.
The Age of Artificial Intelligence
As A.I. systems continue to evolve and expand, so does their impact on our lives.
-
The novelist Stephen Marche is trying to teach artificial intelligence to write with him, not for him. Here is how it is working out.
-
Until now, Google, OpenAI and other companies have been able to use Reddit’s various chats in the development of their A.I. systems for free. That might no longer be the case soon.
-
A.I. competitors like the new Bing are quickly becoming the most serious threat to Google’s search business in 25 years. In response, Google is racing to build an all-new search engine powered by the technology.
-
Nick Bostrom, a philosopher at Oxford, has spent decades preparing for the prospect of A.I. sentience. Here is what he thinks that might look like.
-
Some researchers believe that A.I. won’t reach true intelligence, or true understanding of the world, until it is paired with a body that can perceive, react to and feel around its environment.
-
More than a decade ago, lawyers were singled out as the profession most likely to suffer job losses due to A.I. advances. That didn’t happen.
« TikTok and the Retreat of the Open Web“Simpler and Much Cheaper” – A New Way To Test Antibiotic Resistance »
