Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
×

Comment Confusion (Score 1) 170

I'm not against all AI. I like the idea of using AI to find new protein folding configurations, or using it for automatic pedestrian detection, or helping with industrial automation. What I really dislike is when someone asks an LLM a question and assumes they're getting back a well reasoned and authoritative response. The whole point of an LLM is word/token prediction. All it's doing is taking the text of a conversation up to a certain point and statistically predicting what the next word or token is, with some randomness thrown in. This does a reasonably good job of mimicking chitchat and small talk, like how we just zone out to chat about the weather, but it's not "intelligent". It's literally a random text generator, heavily weighted towards text it's already seen. Don't assume it's correct. Don't rely on it's output. Never put it in charge of decision making. That's all I'm asking.

Comment Re:Conflating (Score 1) 74

And yet bugs will still be annoying. When someone changes their password and the password doesn't change, or someone uploads a video and the video doesn't "upload" (or at least, the key to the upload is lost). "Eventually consistent" should only be used in specific cases where a speedup can be demonstrated (ie, measured), it shouldn't be used as the default case.

Comment We're late to the party on this? (Score 1) 60

I remember hearing a few years ago that in China it was common for people to avoid driving new cars, and avoid inhabiting new construction, specifically due to these kinds of chemical risks. They'd let it air out for a month or two first. The article I read said they were also worried that the manufacturers probably aren't following all safety regulations and might be using materials that don't meet safety guidelines, so it was a good idea to err on the side of caution.

Comment Re:I'm hopeful that future AI (Score 1) 61

These systems, incapable as they are, are end-products of a long optimization process and there is little that can be done to make them better.

That might not be true. There are still a lot of ancient algorithms in AI that might be improved with large datasets. In particular I think combining a knowledge base (like Cyc) with an LLM might be particularly fruitful.

Comment Re:They're banning apostrophe's? (Score 1) 100

When it comes to punctuation its fine if its' missing or wrong so long as you spell correctly and people understand what youre getting at right?

Unrelatedly a coworker was telling me about his big plans for the weekend. He always loves when he gets to enjoy cooking his family and his dog. Another coworker was telling me about his concern for the hostages in Gaza, his ailing mother and President Biden.

(I hope that the above was as painful to read as it was to write)

Comment Re:Declare War on Climate Change (Score 2, Insightful) 268

I don't necessarily disagree with criticizing priorities, but just to be fair, those bombs for the war in Ukraine were almost entirely surplus and the US military was planning to get rid of them. It was actually cheaper to have the Ukrainians fire them at the Russians, and more informative to watch the results. Besides, do you think the bombs were manufactured based on market demand in the first place? The military, to some degree, is a government work program. I do agree that we could build renewable infrastructure as a government work program, but I also think China and India need to agree to do the same thing.

Slashdot Top Deals

Them as has, gets.

Working...