4 ms·
It's unzoomable on mt phone, and I don't have a portative microscope, could someone give 2 sentences whats "berserk" about responses?
by dgan 3y ago
It's unzoomable on mt phone, and I don't have a portative microscope, could someone give 2 sentences whats "berserk" about responses?
- jsemrau 3y agoIt is a collection of screenshots and embeds of tweets with replies and the statement that something has broken. Seemingly a confirmation by OpenAI that something has broken. A complaint that the system prompt is now 1700 tokens. ----- Feels like there is nothing to see here.
- notso411 3y ago[dead]
- seanhunter 3y agoNothing. It's Gary Marcus though and he's carved a niche for himself with doing this sort of thing. It's strange to me that it's given airtime on hn but there you go.
- bambataa 3y agoI kinda feel sorry for Gary Marcus. He’s carved this niche as an LLM critic and must have been delighted to have this bug to post about. I stopped reading his Substack because he was always trying to find a negative. Meanwhile I use LLMs most days and find them very useful.
- xanderlewis 3y ago> I stopped reading his Substack because he was always trying to find a negative. It’s a bit much, isn’t it? I think he’s just trying to counter the fairly dominant AI is the future of everything and in less than a year’s time it’ll be omniscient and we’ll all be living under it as our new God view though. It can be nice to see some scepticism for once.
- kombookcha 3y agoIt's bugging out in some way where it outputs reams and reams of hallucinated gobbledygook. Like not in the normal way where it makes up plausible sounding lies by free associating - this is complete word salad.