Most times I've asked it for/about code, its answers have either been plain wrong or contained hard to spot problems that could well trip many people up. AFAICS it's a fun toy, but not to be trusted at all. I put more faith in my Google-Fu.
The thing is that many are hailing it as 'the future of search' or some kind of all knowing Oracle - which are dangerous things to be treating it as. It certainly has the potential to improve many things, but it is currently far from reaching that potential.
The good thing about existing search engines for me is that they AVOID a lot the ambiguities of natural language and allow you (if you know what you are doing) to quickly find exactly the kind of results you're looking for. Making a bot that you talk to 'naturally' to get your answers brings back all the linguistic fuzziness, with all the potential for miscommunication and misunderstanding that that engenders - not to mention the dubious 'summarised knowledge' being spouted by the model as the 'correct' response.
Accessibility First DevRel. I focus on ensuring content created, events held and company assets are as accessible as possible, for as many people as possible.
Oh this is a whole other can of worms. I am currently exploring this as you raise a lot of concerns I have. I also worry if SEO will become AIRO (AI Relevance Optimisation) and AI will end up even less reliable as industry works out how to manipulate it, as has happened with SEO and SERPs.
Most times I've asked it for/about code, its answers have either been plain wrong or contained hard to spot problems that could well trip many people up. AFAICS it's a fun toy, but not to be trusted at all. I put more faith in my Google-Fu.
The thing is that many are hailing it as 'the future of search' or some kind of all knowing Oracle - which are dangerous things to be treating it as. It certainly has the potential to improve many things, but it is currently far from reaching that potential.
The good thing about existing search engines for me is that they AVOID a lot the ambiguities of natural language and allow you (if you know what you are doing) to quickly find exactly the kind of results you're looking for. Making a bot that you talk to 'naturally' to get your answers brings back all the linguistic fuzziness, with all the potential for miscommunication and misunderstanding that that engenders - not to mention the dubious 'summarised knowledge' being spouted by the model as the 'correct' response.
Oh this is a whole other can of worms. I am currently exploring this as you raise a lot of concerns I have. I also worry if SEO will become AIRO (AI Relevance Optimisation) and AI will end up even less reliable as industry works out how to manipulate it, as has happened with SEO and SERPs.
IDK what "it" you are referring to but if Karpathy says it writes 80% of his code with 80% accuracy, maybe try another model.