It’s not about fear, but about utility. Right now, it mostly generates rubbish for HA configs. Just last night I was doing a search on bing.com and the Bing bot was generating an answer on the side. It was wrong: it was using options that don’t exist on the card I was looking for, neither am I aware that there are plugins that add these options. I then actually tried to follow the trail to understand why it generated that answer, but couldn’t get to the source of it. If I lazily posted that as an answer to someone’s question, I’d be misleading them.
Don’t make this bigger than it is: Simply put, post answers that are your own. That’s the matter in question here. It’s not about whether it can be useful in the (near) future for your own uses, or whether it can be a useful assistant. There are tangents on those fronts here, but that is not the topic.
It’s fine to have different views. Some people will adopt it and others won’t. So be it, but using fear as an argument isn’t helpful, especially not when you’re addressing people’s personal convictions.
Because it cannot reason about what it does — not the way a human does.
I’m reasoning you need to be even more of an expert to use AI generated content in many cases, since you need to be able to spot the very subtle ways it often is wrong. Novice users can’t do that, yet they will be the ones using it.
I did the test, I asked for a Tetris bot code and it generates full garbage code, there is conflicts with some parts of the code, it’s knowledge of libraries is outdated.
And it’s hard to fix the generated code, the only thing you might do is to go on the libraries documentation and learn how it works.
GPT will not try to run the code, but code on GitHub worked at least once for the ones who wrote it.
So even if human wrote code is full of issues it’s still easier to fix it
Well I’ve seen both results, I’ve gotten valuable help from chatgpt (regarding nginx, and sticky ip loadbalancing), and I’ve gotten useless replies (about nginx IP blocking with two nginx’s in cascade). It all depends on the refinements of the questions asked. But I think it’s lovely that the replies shouldn’t be posted in the forum at least, one can use chatgpt on one’s own.
And how do you know that you have to ask in a different way? Only if you verified that the answer is correct/incorrect.
But that’s the thing about using AI here for answers ( and that’s what this thread is about) I don’t get…
I’m reading a question about topic xyz, I have no idea about the topic and don’t know how to answer and … I raise this question with GPT, it comes back with an answer and I post this as response … WHY would I do that?
If I have no clue about the topic I just don’t answer it. Because I wouldn’t know how to validate what the AI was returning.
When I answer here I either verify it before posting or clearly write that my answer is not fully verified.
One would be to fake expertise, someone with no knowledge but want to be seen as an expert,
Another is feeling the need to be helpful for others and thinking they got a nice tool
The last one might be to promote the AI
For the ones who thinks they are helpful we just have to explain them that the tool inst reliable and that we want to interact with real people and for others with less valuable motivation we need to tell them that they aren’t welcome.
Hey, it’s working out great for me! I’m feeling pretty darn satisfied with it. I think I’ll keep my tips and tricks under wraps though - gotta maintain that competitive edge, you know? And who knows, this thing might just outlast us all. It’ll be ruling the world long after I’m gone!
A bit tangential and facetious, but hopefully enjoyable (tongue in the cheek):
[A] man cannot enquire either about that which he knows, or about that which he does not know; for if he knows, he has no need to enquire; and if not, he cannot; for he does not know the very subject about which he is to enquire. — translated by Benjamin Jowett, 1871
I suggest not bothering to revisit that rule ever. If we wanted to discuss things with an AIBot, we wouldn’t be here. We could do that directly. No one needs someone to cut and paste GPT for them. We are not that incompetent that we can’t query an AI ourselves.
Well, I agree with you 100% @somates . I myself haven’t seen any posts where ChatGPT helps others with questions about fixing the home assistant. I would think they were deleted by now, as I cannot see them or have not seen them in the past. Whoever copied and pasted this for the helping hand should be ashamed. You could say that they made this blog post, which makes this AI technology look bad. It’s not the post, it’s people thinking that everything with the words AI, OpenAI, GPT, and ChatGPT is disliked here. After all, this technology can do more than just answer simple questions.
I do not dislike ChatGPT as it has saved me a lot. I love all my forums, but I have a reading and writing disability. Now, I write something and then pass it on to ChatGPT to fix any errors. Then, I ask Edge to use the read-out-loud feature to see if it sounds good . It makes me feel so much better that I can do this after 38 years of not being able to write without feeling embarrassed about my big grammar and spelling mistakes.
What a great use for AI! I remember back in the day when managers in an office environment who had poor written language skills had secretaries write everything for them. Then along came e-mail and we all got to find out how bad they really were. It’s always smart to have someone (or in this case, some thing) proof-read whatever we write before sending it out into the world.
YES! Thank you.
As a newcomer, I can’t tell how much time I’ve wasted, trying someone’s “answer” that didn’t work. I’m trying to research the problem, and go about it in a way to learn, but I keep running across things that supposedly “work,” but they do not. The YAML is often not correct, in “oh, so subtle” ways.
Next, it would be great if someone could figure out how to remove the posts that people have put in here that were generated that way…
Since we are talking about Ai and all does anyone know if there is a Telegram/Discord bot that incorporates Chatgpt and Home Assistant? If not it would be awesome to have one.
It would be awesome to have this as an add-on, then you can define what sensors / switches / services you want to expose to it. Then you can use it as a normal bot or just ask it to do house stuff. Or maybe it even has a special mode where it helps you with HA related questions / answers
This has been a fun discussion to read. There are some serious ChatGPT fanboys here who appear to think ChatGPT hung the moon. Any AI that has been reported as telling many people to commit suicide as a solution to their problem is still not quite ready for prime-time. Sure ChatGPT has some interesting things it has done, but it’s not quite up to snuff yet.
As for banning it, I agree. Not because I don’t embrace new tech or because I somehow hate ChatGPT, but no other reason than it takes volunteers precious time who mod this site to have to now audit post after post and delete and warn and so on. This is analogous to spammers on this site. If the answer is more often than not incorrect then why would anyone advocate for ChatGPT to be allowed?
Someone suggested having an “Ask ChatGPT” as an option, I think that is a pretty good idea. Link that to a pop-up that explains why it’s been problematic before they get to ask, then let them ask away and then return to the forums to get help on why their answer 100% doesn’t work.
Nobody is also considering the ramifications of the fact that ChatGPT can be easily biased. It wouldn’t take much to beef up a competing solution (say Hubitat for example) and keep HA incorrect for the sake of getting users to jump to a new platform. I’m not saying this will happen, but it can happen and it’s one of the biggest warnings experts issue regarding the future of AI in general.
It will just result in people making posts with vague statements like “this doesn’t work” which leaves us in the same situation (at least for now) when such a bot can’t actually solve the problem. It just encourages lazy behaviour if not used responsibly. I maintain that the use of this tech is not for the novice.
While I understand the concerns raised by the author regarding ChatGPT, I respectfully disagree with the suggestion that ChatGPT should be banned from this site.
Firstly, it is important to note that ChatGPT is not the only AI model available. There are many other AI models that can be used to assist users on this site, and it is possible to carefully select and configure an AI model that meets the site’s specific needs and requirements.
Furthermore, the reported instances of ChatGPT suggesting suicide as a solution were the result of biased training data, rather than a flaw in the model itself. It is important to recognize that AI is only as good as the data it is trained on, and biases in the data can lead to biased or inaccurate responses from the model. However, this is a problem that can be addressed by carefully selecting and curating the training data, as well as regularly monitoring and auditing the model’s performance to identify and correct any biases.
Additionally, while it is true that implementing an AI system can require additional time and resources for moderators to audit and manage, this is a trade-off that can be worth it in terms of the benefits that an AI system can provide. AI can help improve the accuracy and speed of responses to user queries, and can potentially reduce the workload on moderators by automating certain tasks.
In conclusion, while it is important to be aware of the potential risks and limitations of AI, I believe that banning ChatGPT from this site would be a premature and overly restrictive response. By carefully selecting and configuring an AI model, and regularly monitoring and auditing its performance, it is possible to harness the power of AI to improve the user experience on this site.
None of us want to do that. Use AI to your hearts content. Don’t post it’s results as answers here.
That’s it.
Making more arguments for AI here isn’t going to change our opinions. The only thing that will change our opinions is when AI answers questions correctly more often than incorrectly without us intervening.
Is it just me or did @nick2525’s post, above, read like it was written by AI?
Either way, it totally misses the point. Nobody here is “banning AI.” What’s being “banned” is posting answers generated by AI here in this forum.
I come to this forum seeking the knowledge of experienced (human) users. If I want search engine results, I’ll use a search engine. If I want AI results, I’ll use ChatGPT. And I have. We’re not saying there is anything wrong with AI in general, just that this isn’t the place.
Yes, it sounds like he did. He also posted something on my post about AI.
It’s disappointing because now everyone will be upset about the posts that I had rewritten with ChatGPT.
As I mentioned before, I use it a lot due to my disability.
Moving forward, I want to show that I’m not doing anything wrong. So, I’m posting like this now…
If you would like to see my original text before ChatGPT helped me with my disability, click on the edit history, and you’ll see my original writing before ChatGPT.
. It makes me feel so much better that I can do this after 38 years of not being able to write without feeling embarrassed about my big grammar and spelling mistakes.