AI has so completely disrupting Search that it’s destroyed leading platforms effectiveness in a matter of months.
But because of its current lack of optimization for accuracy, we shouldn’t consider it disruptive because it’s not yet proven technology?
You can call it dangerous but you can’t call it useless. It’s also only going towards improvement from here, including drastic reductions in hallucinations.
You have to remember too that AI models are generally attempting to interpret the intent behind the prompt, so many of these crazy articles are happening because people aren’t yet good at writing clear instructions for AI and AI isn’t yet mature enough to disambiguate poor instructions in its output and is trying to deliver on unclear instructional intents.
>> You can call it dangerous but you can’t call it useless. It’s also only going towards improvement from here, including drastic reductions in hallucinations.
> Why?
A pseudo-religious belief in progress, especially the technological kind.
It's bullshit. If it were true, 2022 pre-LLM Google would have been better than 2010-era Google, but it most definitely wasn't. Consumer printer technology, for the most part, has been getting worse for decades at this point.
If it cannot be trusted to return accurate information without hallucinations, then it shouldn't be publicly available in a system meant to be used for finding accurate information like web search. I could see it's value in generating office documents, but even for summarizing them you still get the same problem of hallucinations.
Misinformation and disinformation is already a problem on the web and thrusting unproven technology like generative AI that has a tendency towards misinformation is opening a Pandora's box. But as long as Microsoft and Google and Meta can make their money...
But because of its current lack of optimization for accuracy, we shouldn’t consider it disruptive because it’s not yet proven technology?
You can call it dangerous but you can’t call it useless. It’s also only going towards improvement from here, including drastic reductions in hallucinations.
You have to remember too that AI models are generally attempting to interpret the intent behind the prompt, so many of these crazy articles are happening because people aren’t yet good at writing clear instructions for AI and AI isn’t yet mature enough to disambiguate poor instructions in its output and is trying to deliver on unclear instructional intents.