AI, Explained Plainly
GPT-3, the language model OpenAI described in a 2020 paper, was trained on 300 billion tokens of text, and a token averaged about 0.7 of a word. This site explains what numbers like those mean for someone running a business. Each article takes one part of how an AI tool works, like how a chatbot picks its next word or why it can state something false with total confidence, and walks through it in plain language with the research linked. The goal is that you can look at an AI tool for your business and understand what it's doing well enough to decide for yourself.
Articles
- How a chatbot writes an answer, one piece at a timeA chatbot writes by predicting the next piece of text, adding it, and predicting again. Here's how that works and what it means for your business.
- If your chatbot gets a policy wrong, it's still your policyA tribunal made Air Canada pay for a wrong answer from its website chatbot. Here's why chatbots state wrong facts with confidence.