My friend created a pretty cool Youtube video explaining some of the mechanisms behind ChatGPT; and you’d realise that behind the whole engine of this ‘artificial intelligence’ is really lots of brute calculations and updates. Never at any point of the calculations would ChatGPT ‘fact-check’ or validate anything that it ‘says’. It relies upon statistical accuracy in matching and finding the next most relevant word to string together.
So the entire system wasn’t quite ever designed to give you accurate or right information – just something you might believe rather than something real. In some sense, technology as an aid to humans have reach the new next level of helping us to ideate more by bringing up relevant connections to what you’ve first provided as a prompt. But it doesn’t replace thinking; it simply supercharge one’s memory prowess more.
What is interesting in this whole march of progress with technology is that we can say rote memory is no longer that useful and should not be tested in school. But associative memory remains important and the ability to actually triangulate and fact-check things becomes even more important. Search engines and all of the internet are now no longer ‘objective’ but creating bubbles to house its users, trying to deliver the most ‘relevance’ to them. Validation of information now becomes a super-skill that depend on random and highly dispersed connections.