Behind AI

I’ve just been reading Stephen Wolfram’s “What is ChatGPT doing and why does it work?” It’s quite long, and though at times it gets a little complicated, it’s a really clear description of how the technology behind ChatGPT & other Large Language Models (LLMs) work.
It starts with simple examples using single words (e.g. if there’s a letter ‘q’ in a word, you know that the next letter has to be ‘u’) , then moves to examples of word combinations. It also explains why AI generated text can be a bit bland & boring – but also why it can generate sentences that are grammatically accurate (e.g. “The elephant travelled to the Moon”) but factually inaccurate.

You might also recognise the name – Stephen Wolfram is also behind Wolfram Alpha. Highly recommended, though I’d sit down with a cup of tea and a plate of biscuits while you ponder it 🙂