Big players like Microsoft with Bing AI (and Copilot), Google Bard and OpenAI ChatGPT-4 are making AI chatbot technology, previously limited to test labs, more accessible to the public.
How do these LLM programs work? OpenAI GPT-3 told us that the AI uses "autocomplete similar programs to learn languages" and that these programs analyze "statistical characteristics of language" to "make educated guesses based on advanced words". " Or in the words of James Vincent, the man: "These AI tools are giant autocomplete systems trained to predict what word will come next in a sentence. Therefore, they have no hard database of 'facts' to rely on - just the ability to write believable -sounding statements. This means that they tend to pass off false information as truth because the fact that a certain sentence sounds plausible does not guarantee its truth. But there are so many other parts of the AI landscape that are coming in – and there will be problems
COMMENTS