That’s a little harsh. I don’t think any offence was meant by “you guy”. It’s just casual speak.
He was being challenged and I thought he didn’t rise to it and deal with it politely!
If you follow the chronology, you’ll see that my challenge came after “you guys” not the other way around.
@dublinpilot don’t most hotels have a card in the rooms with breakfast, lunch and dinner times, plus all sorts of other services offered? Do we need AI for that.
I have an accent. If I was in a Dublin hotel and I ring down and tell them my shower needs fixing, do you think AI would know what I am on about? It’s bad enough with some human interfaces.
Yes internet translation is pretty good now but can AI really take it further in dealing with the voice and not the written word?
AI is IMO a double edged sword with the future benefits side being pretty blunt at the moment.
The calculator is great but it means that most under 20’s can no longer do even the simplest of calculation in their heads.
Something can be extremely limited and yet be very useful !
For translation it’s hard to beat. Worlds better than Google / DeepL. For finding ways to phrase a sentence too. Just ask suggestions and see what works for you. I encourage skeptics to use Mistral chat (it’s free and uses Mistral Large model, mostly as good as ChatGPT 4).
It still lacks « thinking » because it has no world model and no planning, but in practice later models are so large that they mostly find the correct context and stay on track (they find correlations with what you’re talking about in their vast « knowledge »). We are just lending another capacity to computers (the ability to use words), just like we’ve done many times before. It’s a productivity tool that’s good for some tasks. With the carried risk, like gallois said, to lose the ability to do it ourselves. I would find it concerning if people aren’t able to phrase sentences anymore, but youngsters today already struggle with syntax, phrasing and vocabulary, and that was before the popularization of LLM’s.
I don’t think people will « fall behind » noticeably more than they already have with regards to technology. I have many anecdotes where the use of and familiarity with technology makes my life a little more efficient. LLM’s are no different.
An interesting article which might explain why forums everywhere are filling up with real people posting bland questions intended to generate answers for incorporation in the phrase databases:
EuroGA has had a few. Normally I delete those people but some get through. I think somebody managed it in the insurance thread.
Jumping back on this thread after a while.
ChatGPT content posted in the main forum will be deleted
What do you think of using ChatGPT for editing a post? For those of us not great at English? Not that I really have, but just curious, seems a valid use to me.
A few more random thoughts:
- copyright: this is a very interesting and complicated issue. Beside being just inspired (isn’t a lot of what human do tweaks on things they have learn from other), these model make it much easier to generate content “in the form” of someone else, but on original content.
- I believe the model show creativity. They come up with original content, inspired by human content, but this also I just what human do.
- I am personally in awe at what simple matrix communication at scale can do. That said no amount of questioning ChatGPT got me to understand what “attention” really is while one or two video from smart YouTuber got me much closer.
- I belong to the optimist group. I think this is really just the beginning, I still believe the models show interesting ability to reason and this coupled with specialised “narrow” ai tool should take us a long way, the “agent” approach.
What do you think of using ChatGPT for editing a post?
I think you meant “translating” a post?
Some people do post google translated text, which is fair enough. I get it at work a lot from customers and it tends to result in a lot of going around in circles. The problem with ChatGPT output is not that the English is bad (it is usually very good) but that the output is bland. Hard to define “bland”… most output generated by a human is coloured by their experience, and this is readily detectable by anyone intelligent.
I’ve not tried using it for translations, and would need a native of the target language to check the output. Is it supposed to be just a good translation?
Peter wrote:
I think you meant “translating” a post?
I mean for example to “edit”. To be clear, I haven’t convinced myself, as it has the draw back that you don’t naturally improve/learn by using it, but here for example, my above post “edited” by chatgpt, would look like this:
Here are a few more random thoughts:
- Copyright: This is an intriguing and multifaceted issue. Beyond mere inspiration (after all, much of human creativity involves building on what we’ve learned from others), these models streamline the generation of content that resembles someone else’s work but remains original.
- I find these models to be quite creative. They produce original content inspired by human creations, which mirrors our own creative processes.
- Personally, I’m fascinated by the power of simple matrix communication at scale. However, despite my best efforts, ChatGPT couldn’t quite clarify what “attention” truly means for me, whereas a couple of insightful YouTube videos from knowledgeable creators got me much closer to grasping the concept.
- Count me among the optimists. I believe we’re just scratching the surface. These models demonstrate intriguing reasoning abilities, and when combined with specialized “narrow” AI tools, they could propel us forward significantly—the “agent” approach holds immense promise.
Definitely more “polished” but less natural…
Yes; more “PC” and more bland…
Someone told me there is another ChatGPT interface which doesn’t do the “PC treatment” i.e. it is a post-processing thing.
Somehow, I would be suspicious of your post if you were a brand new signup.
I don’t think it’s post processing, but rather pre training with lots of question and answer examples to become an assistant.
You can run raw models on your computer or hugging face. They’re far from being politically correct or even polite.
Roznets non chat gpt post definitely sounds more natural and therefore more interesting to me.
The bland label is a good description of what chapgpt did to his post.
