News for hackers. Stuff that matters.
Re:Finland's ChatGPT equivalent begins to think in Estonian ... by rjzzleep on Tuesday October 17, 2023 @05:28AM (#37912489)
I actually think you're wrong. I just had a discussion about this with a friend who was infinitely excited about it and felt I was being extremely dismissive when I said writing agent guidelines is very much akin to micromanaging an extremely cheap outsourcing company.

MOST people actually think ChatGPT is actual, sentient thinking. We're kind of the minority in this story.

Re:Finland's ChatGPT equivalent begins to think in Estonian ... by thih9 on Tuesday October 17, 2023 @05:37AM (#37912539)
According to one poll, we are not a minority: "10% of people say ChatGPT is sentient; 37% are not sure; and 53% say it is not."[1].

Also note that it's a very new technology. I'd say that over time and as the public gets more familiar with the tech, they will understand it better.

[1]: https://itbrief.co.nz/story/concern-over-ai-increasing-since...

Re:Finland's ChatGPT equivalent begins to think in Estonian ... by ljlolel on Tuesday October 17, 2023 @08:40AM (#37914029)
Do you think that 10% will become 1% or 50% in 5 years with GPT-7?
Re:Finland's ChatGPT equivalent begins to think in Estonian ... by thih9 on Tuesday October 17, 2023 @03:51PM (#37920629)
My guess is that it’s going to join other urban mysteries, from cellphone radiation causing DNA damage, to aliens.

As in: most people will be aware of the topic, most will dismiss it, but every now and then there will be a news story.

Surprisingly “Four in 10 Americans now think some UFOs that people have spotted have been alien spacecraft visiting Earth from other planets or galaxies.”[1], so we might see an increase to 40% actually.

Unless of course GPT-7 gains sentience ;)

[1]: https://news.gallup.com/poll/350096/americans-believe-ufos.a...

Re:Finland's ChatGPT equivalent begins to think in Estonian ... by rjzzleep on Tuesday October 17, 2023 @01:49PM (#37918819)
You're correct, I ignored the "sentient" part of the comment. If we were to reduce the question exclusively to the sentience part, my comment holds no ground.
Re:Finland's ChatGPT equivalent begins to think in Estonian ... by anileated on Thursday October 19, 2023 @02:21AM (#37939183)
From what I see on HN, many people (mostly, I guess, in AI industry) justify Microsoft/ClosedAI ignoring copyright when training LLMs because we don’t prevent humans from learning on copyrighted material and using it to create artworks.

Therefore, by their logic, LLMs are “like human” enough that they should have this basic right as well.

An understanding that “thinking” or “training” in case of an LLM is not at all like thinking in case of a human would go a long way to realizing this take doesn’t hold water.

Re:Finland's ChatGPT equivalent begins to think in Estonian ... by devjab on Tuesday October 17, 2023 @05:42AM (#37912578)
I’m not sure about that part about most people. We obviously come from different anecdotal stories, but in my business, which is non-tech enterprise, nobody views GPT as “thinking”. People generally see it as a “better Google”.

Our main issue is with people thinking it’s answers are actual factual. And while it’ll obviously get things right most of the time, we’re really struggling with how much our employees trust it. That and the data privacy, but that’s a different story that will eventually get solved once AzureGPT for Enterprise kicks off.

Another “interesting” issue is how GPT will sometimes “comment” texts if you don’t prompt it correctly. We had someone who asked GPT to correct the grammar in some contract, and it did. It also added some comments on how part of what was written was “impressive”. Which would have been extremely hilarious if it had been an internal thing, but our lovely employee didn’t proof read the proof read and actually sent the contract draft out to a customer. Luckily a good humoured customer who took it with a laugh, but hell, that’s going to be a nightmare going forward.

Re:Finland's ChatGPT equivalent begins to think in Estonian ... by cathalc on Tuesday October 17, 2023 @05:50AM (#37912634)
Anecdotally, I am yet to meet a single person, in or out of my professional life, who thinks ChatGPT is in any way sentient.
Re:Finland's ChatGPT equivalent begins to think in Estonian ... by raincole on Tuesday October 17, 2023 @05:57AM (#37912687)
My anecdotal experience is that we are the minority, but in a completely different sense.

My parents (whose tech-savvy level is they call iPad "PC") think ChatGPT is like Google, but "someone make it say the search result like a human".

Re:Finland's ChatGPT equivalent begins to think in Estonian ... by JeremyNT on Tuesday October 17, 2023 @10:15AM (#37915428)
This strikes me as a pretty reasonable way to conceive of these things.

Google uses the same ML techniques in their products after all - they just haven't exposed the chatbot UI until recently.