@dangillmor Wait, you're telling me a (admittedly extremely advanced) random number generator isn't trustworthy? ​​
Sarcasm aside, large language models have their uses, but being your one and only source of information isn't one of them and it's good that more people are starting to notice, although it's not even close to everyone yet

@dangillmor The image should have given the robot too many fingers.

@dangillmor neat, time for them all to jump ship to the next scam.

@dangillmor Just following the stages of the hype cycle.

@dangillmor why is it a good news?

@dangillmor @Gargron I think a big factor here is that Gen AI is superficially amazing but the deeper you dig, the more holes you find. Even people who are excited about the tech are noticing this.

@dangillmor If you're looking for information about an obscure topic, then AI definitely isn't something you should rely on. Of course, that's not to say all its features are no-good. Merely, AI is not ready to be a good, comprehensive knowledge base.

@dangillmor AI is in full gold rush, speculation mode. This started when one player in big tech achieved an advantage and the other big players rushed in to catch up. This was perceived as demand from the rest of the market and now we're in a speculation vortex. At a consumer level I'm detecting very little practical demand for, or uptake of AI products. At least the current offerings.

@dangillmor very funny (and exasperating) that they're calling it a "trust problem"

@dangillmor #Counterfactual : How much trust would we place in an employee who regularly and confidently hallucinates?

@dangillmor @briankrebs the only AI news I really want to read

@dangillmor on the other hand, the techbro scheme to blame it all on AI ™ seems to have worked like a charm.

@dangillmor I am not surprised to see a more positive view from people in third world countries. I am seeing a lot of skilled engineers and grad students for whom English is a second language eager to use LLM systems to improve their writing. And since the structure of text is the only thing the models really "understand", it might be good for that (though it is only safe if their English is good enough to catch mistakes). The problem is all the people and companies trying to use it for things that it just cannot do well, like customer support (fire support people and make customers talk to a chatbot).

@dangillmor How do you deal with companies just lying and not marketing it as AI though - just customer service?

@dangillmor I feel like public opinion won't count for much when it's use is required by employers. Right now we're collectively in the phase of try it and see if it can make things more efficient, which will be followed by "we're an openai shop", which will be followed by "xzy ai handles that, get back on the line and listen to what the machine tells you"