Hacker News Reader

Comment by sanderjd | Hacker News Reader

I really think the ability to ask questions entirely free from all judgment is an under-emphasized aspect of the power of these tools. Yes, some people are intellectually secure enough to ask the "dumb" questions of other humans, but most people are not, especially to an audience of strangers. I don't think I ever once asked a question on Stack Overflow, because it was easy to see how the question I worried might be dumb might be treated by the community there. But I ask all sorts of dumb questions of these models, with nary a concern about being judged. I love that aspect of it.

View on HN

The other thing is that it will make an earnest attempt to answer the question. On the other hand with places like SO, many questions will be incorrectly marked as duplicate with the “answer” link pointing to a post that might seem similar at first glance but is different enough to not actually be the same, which is supremely unhelpful.

You can also ask it to explain the subject like you’re 5, which might not feel appropriate when interacting with a human because that can feel burdensome.

All of this is heavily caveated by how dramatically wrong LLMs can be, though, and can be rendered moot if the individual in question is too trusting and/or isn’t aware of the tendency of LLMs to hallucinate, pull from bad training data, or match the wrong patterns.

Yep, this is exactly what I mean!

Personally, I find that even when it's wrong, it's often useful, in that I come away with hints toward how to follow up.

I do have concerns that people who haven't lived a couple decades of adult life prior to the existence of these tools will be a lot more credulous, to their detriment.

That's a subtle, yet important point. Putting themselves out there is not easy for some. LLMs can take that pressure away.

The 'but' in that lies with how much freedom is given to the LLM. If constrained, its refusal to answer may become a somewhat triggering possibility.

I think this "refusal to answer" thing is so overblown. I have been using this technology every day for over two years now and have not one single time run into this.

You might be right. From my experience, it depends. The edgier your exploration, the more that can happen.

It's important to note that not everyone abide by the same morals. And a narrowly constrained model may end up refusing genuine inquiries just because.

In any case, if anything, this is a small 'but'. OP's point is the gold nugget here. That is, LLMs allowing exploring subjects without the fear of being judged for one's natural curiosity.

Yeah, I'm in no way claiming that this isn't a thing, or even that it isn't a problem.

But it isn't a problem for most people. The kind of edgelords that run into this are overrepresented on internet forums, including HN, but it's actually a pretty small group of people.

The tool is absolutely biased, what makes you think it wouldn’t be?

This guy is obviously unfamiliar with Tay lol.

My comment doesn't say anything about bias...

Reply on news.ycombinator.com