|
|
|
|
|
|
by shh_labs
794 days ago
|
|
|
But the trains with differently sized tracks didn’t fall off the tracks because (and I imagine there is some obscure exception to this, but for the very most part) someone had made sure that the systems were safe. The distinction here is that Google is rolling out half-baked rubbish that is palpably unreliable. No sane person is going to eat rocks or put glue on their pizza, but the race is on to find some genuinely dangerous answer from this system. It would be regrettable if the result of rolling out this type of product too soon has the effect of eroding trust in AI-based search, and delaying people from engaging with it when it actually is ready. |
|
|
nowhere in that sentence did you mention AI, though. If you take AI out of the picture, isn’t it equally problematic if they had some other mechanism, say some kind of analytical/stochastic mechanism, that selected data retrieved from actual results, and cited those?
Because, that’s what they’re doing. Google isn’t using an AI. They’re using a stochastic textual generator driven by results of the search. Sometimes it spits out weird and incorrect answers, especially when high-ranking sites suddenly spoof malicious results - that is a general problem with any attempt to pull data out of the results automatically, including the lesser variants of this approach they’ve deployed for decades.
The problem is (a) google putting out a poor product, similar to the pre-google era of search engines being really bad, and (b) high-ranking sites engaging in malicious and disruptive behavior, for which they probably should be de-ranked or blacklisted.
But you cannot gave this one both ways, it can’t be an “AI” when you want to be an maximalist and “just a textual stochastic algorithm” when you want to minimize and downplay. By the latter terms Google isn’t even using AI here, just a stochastic algorithm, driven by the search results. It’s just a bad one. So you’re arguing against an approach that doesn’t exist anyway.
Again: why specifically is google answer problematic, given that it’s been rolled out for like 15+ years now? The new algorithm behind it just kinda sucks, but the feature isn’t new. Citing and summarizing results has been a thing for a long time, and naturally it’s not always correct/accurate.
This isn’t the first time people have been upset over it either - newspapers got mad about this summarization so they got a law passed in Canada which outlawed it.
https://www.reuters.com/technology/canada-news-industry-body...