The AI Thread
- Thread starter Truthy
- Start date
It is beyond awesome, I am already in love with itI am playing with ChatGTP a lot. I hope that he will not use it against me. He he
I have stressed for several weeks about a letter I needed to write to a builder I used in the past. It turns out some of his work was shoddy and needs to be redone.
I am asking him to pay for the repairs.
I know this has almost zero percent chance of a positive outcome, but feel if I don't at least ask I am being a pushover in life.
I asked ChatGPT.
I GASPED at how perfect the letter was. I sent it to my old builder with some minor edits.
What a relief, what an amazing tool!
I am asking him to pay for the repairs.
I know this has almost zero percent chance of a positive outcome, but feel if I don't at least ask I am being a pushover in life.
I asked ChatGPT.
I GASPED at how perfect the letter was. I sent it to my old builder with some minor edits.
What a relief, what an amazing tool!
Thorgalaeg
Deity
Yep. It is really hard to believe the ai really doesn't understand what it's writing. In fact it works similarly to stable diffusion, dalle-e and other AI image generators.
I have experimented a lot with stable diffusion and working with it you slowly get an intuitive grasp of how it works and where the limitations are. While the generated images are really breathtaking it is evident the ai doesn't really understand what it's painting and errors are usual. For instance a leg can become a arm, long hair can become a scarf, a person can have two heads and siamese twins are usual, specially in high resolution images. For the ai images are only a bunch of noise with irregularities which become things when refined again and again being guided by data from a billion of pre-trained images selected according to our prompt.
With text models pixels are replaced by words and images by texts but it is basically the same at the end, however these inner mechanisms become less evident and the magic is even stronger.
Still we don't know what "intelligence" really is or how is it formed, so I wouldn't be so sure about these models not being intelligent in some way different, or not so different to ours. So I wouldn't discard a Terminator scenery in a couple of years, in the meantime we can enjoy and use this amazing tools to our benefit.
I have experimented a lot with stable diffusion and working with it you slowly get an intuitive grasp of how it works and where the limitations are. While the generated images are really breathtaking it is evident the ai doesn't really understand what it's painting and errors are usual. For instance a leg can become a arm, long hair can become a scarf, a person can have two heads and siamese twins are usual, specially in high resolution images. For the ai images are only a bunch of noise with irregularities which become things when refined again and again being guided by data from a billion of pre-trained images selected according to our prompt.
With text models pixels are replaced by words and images by texts but it is basically the same at the end, however these inner mechanisms become less evident and the magic is even stronger.
Still we don't know what "intelligence" really is or how is it formed, so I wouldn't be so sure about these models not being intelligent in some way different, or not so different to ours. So I wouldn't discard a Terminator scenery in a couple of years, in the meantime we can enjoy and use this amazing tools to our benefit.
Imo it's not intelligence, but yes, it can do a lot of complicated things. The fact that there is no sense behind it, would easily lead to tragedy if it was given a body, like ED209 ^^Yep. It is really hard to believe the ai really doesn't understand what it's writing. In fact it works similarly to stable diffusion, dalle-e and other AI image generators.
I have experimented a lot with stable diffusion and working with it you slowly get an intuitive grasp of how it works and where the limitations are. While the generated images are really breathtaking it is evident the ai doesn't really understand what it's painting and errors are usual. For instance a leg can become a arm, long hair can become a scarf, a person can have two heads and siamese twins are usual, specially in high resolution images. For the ai images are only a bunch of noise with irregularities which become things when refined again and again being guided by data from a billion of pre-trained images selected according to our prompt.
With text models pixels are replaced by words and images by texts but it is basically the same at the end, however these inner mechanisms become less evident and the magic is even stronger.
Still we don't know what "intelligence" really is or how is it formed, so I wouldn't be so sure about these models not being intelligent in some way different, or not so different to ours. So I wouldn't discard a Terminator scenery in a couple of years, in the meantime we can enjoy and use this amazing tools to our benefit.
Even now, it can massively help with limited (constructed from a set base, up) environments, such as writing code or (one level above) providing real time instructions on how to do things in 3d modellers such as Blender.
Thorgalaeg
Deity
AI will be inserted in most human activities. It has been hidden behind internet search engines for some time, now text and images generators are finally available for anyone, 3D is almost here too, it is already being used in science, in design, production, business etc. In the near future it will drive our cars, will control or houses, our bank accounts, our education, our health, our security, even our wars...
What I find somewhat worrysone is an huge unified AI which has access to every aspect of human life. I think there are incentives to create such thing, since it would probably work much better than smaller specific AIs, because it would have access to a much bigger set of data and the results would be more useful globally, but the power of such being would be absolutely terrifying.
Btw, has been playing a bit more with the new Bing which incorporates chatgpt, and yes it is chatgpt, but in some kind of prudish infant mode. In contrast with chatgpt which will answer almost anything, if Bing finds something a bit difficult it will inmediately say: "sorry I can't answer that, I am too young and I am still learning"
The interface is also much worse, it is a mess and conversations are easily lost. They are not registered anywhere either as with chatgpt. So there is some way for Microsoft to walk before this new Bing becomes really useful IMO.
What I find somewhat worrysone is an huge unified AI which has access to every aspect of human life. I think there are incentives to create such thing, since it would probably work much better than smaller specific AIs, because it would have access to a much bigger set of data and the results would be more useful globally, but the power of such being would be absolutely terrifying.
Btw, has been playing a bit more with the new Bing which incorporates chatgpt, and yes it is chatgpt, but in some kind of prudish infant mode. In contrast with chatgpt which will answer almost anything, if Bing finds something a bit difficult it will inmediately say: "sorry I can't answer that, I am too young and I am still learning"
The interface is also much worse, it is a mess and conversations are easily lost. They are not registered anywhere either as with chatgpt. So there is some way for Microsoft to walk before this new Bing becomes really useful IMO.
Last edited:
I thought Bing was for porn?
Maybe the AM (from I have No Mouth and I Must Scream) or the univac or what it was ultimately called in that Asimov storyAI will be inserted in most human activities. It has been hidden behind internet search engines for some time, now text and images generators are finally available for anyone, 3D is almost here too, it is already being used in science, in design, production, business etc. In the near future it will drive our cars, will control or houses, our bank accounts, our education, our health...
What I find somewhat worrysone is an huge unified AI which has access to every aspect of human life. I think there are incentives to create such thing, since it would probably work much better than smaller specific AIs, because it would have access to a much bigger set of data and the results would be more useful globally, but the power of such being would be absolutely terrifying.
Btw, has been playing a bit more with the new Bing which incorporates chatgpt, and yes it is chatgpt, but in some kind of prudish infant mode. In contrast with chatgpt which will answer almost anything, if Bing finds something a bit difficult it will inmediately say: "sorry I can't answer that, I am too young and I am still learning"
The interface is also much worse, it is a mess and conversations are easily lost. They are not registered anywhere either as with chatgpt. So there is some way for Microsoft to walk before this new Bing becomes really useful IMO.
But chatGpt is only good with responses given as text; I did ask it (trying to override that) to provide some ascii drawings, but it failed even in drawing triangles with right angles - it would just draw a triangle and then try to convince you why it has a right angle despite it being the same collection of acute angles. It crashed on me multiple times, while trying to do something relatively simple, such as typing the arabic numerical value of the angle inside the triangle drawn with ascii.
Last edited:
Thorgalaeg
Deity
It is still not ready to kyriakos level of weirdness but it will be at some point.Maybe the AM (from I have No Mouth and I Must Scream) of the univac or what it was ultimately called in that Asimov story
But chatGpt is only good with responses given as text; I did ask it (trying to override that) to provide some ascii drawings, but it failed even in drawing triangles with right angles - it would just draw a triangle and then try to convince you why it has a right angle despite it being the same collection of acute angles. It crashed on me multiple times, while trying to do something relatively simple, such as typing the arabic numerical value of the angle inside the triangle drawn with ascii.
Seriously, It all depends on the things the models have been trained on I suppose. Obviously ASCII graphics is not among its strongest topics.
But It is curious as it can write the most complex part of a program correctly to get it wrong in the more simple parts. Shows it doesn't really knows what is doing.
Yes, and since it doesn't know what it is typing, it is excellent for closed and strictly hierarchical systems - since no one needs to know stuff there to just produce a correct answer (illustrated in a famous "Chinese Room" metaphor) ^^
Gori the Grey
The Poster
- Joined
- Jan 5, 2009
- Messages
- 13,482
we'll call it an amazing tool when you report that the builder has in fact paid you!
He replied 8am this morning asking to come view the problem. He's never been so polite and prompt!
Gori the Grey
The Poster
- Joined
- Jan 5, 2009
- Messages
- 13,482
Now that is something. Would you be willing to post the letter (with specifics redacted, of course)? And say which parts of it not that you couldn't have written on your own, but that you didn't want to during the time that you were disinclined to write the letter? I'd be very interested.
Last edited:
Hygro
soundcloud.com/hygro/
He was like oh shiß a lawyer wrote this!
Gori the Grey
The Poster
- Joined
- Jan 5, 2009
- Messages
- 13,482
Or actually I suspect something else, Hygro: an appeal to the builder's reputation. But yeah, quasi-legal language might do it, too. That's why I would love to see Aiken's letter. Or rather, the letter Chat GPT generated on Aiken's behalf.
Hygro
soundcloud.com/hygro/
I am assuming it was just worded so friendly and warmly in that chatgpt way that it reached the guy. I don't think it was fear based, just a joke.
Lexicus
Deity
Chat GPT does not comprehend anything. It is as intelligent as a rock.
I'll post it tomorrow. I would guess the difference was that I am normally very casual and this had a whiff of lawyer.
Or that I did this at 3am so it was the top of his inbox in the morning...
Or that I did this at 3am so it was the top of his inbox in the morning...
The AI letter:
Dear [] ,
I am writing to bring to your attention some repairs that need to be carried out in [] . As you are aware, you were contracted to undertake a building project in my home, and although the work was completed initially to a satisfactory standard, I have since noticed some issues that require your attention.
Specifically, the following repairs are needed:
[]
I am disappointed that these issues have arisen, and I am requesting that you take responsibility for these repairs. I believe it is reasonable to expect the work to be carried out at no additional cost to me as these are issues that have arisen as a result of the work that you undertook. [.]
I would appreciate it if you could respond to this email as soon as possible so that I can arrange for these repairs to be completed. I would also like to arrange a time for you to visit the property to discuss the matter further.
Thank you for your attention to this matter.
Sincerely,
Dear [] ,
I am writing to bring to your attention some repairs that need to be carried out in [] . As you are aware, you were contracted to undertake a building project in my home, and although the work was completed initially to a satisfactory standard, I have since noticed some issues that require your attention.
Specifically, the following repairs are needed:
[]
I am disappointed that these issues have arisen, and I am requesting that you take responsibility for these repairs. I believe it is reasonable to expect the work to be carried out at no additional cost to me as these are issues that have arisen as a result of the work that you undertook. [.]
I would appreciate it if you could respond to this email as soon as possible so that I can arrange for these repairs to be completed. I would also like to arrange a time for you to visit the property to discuss the matter further.
Thank you for your attention to this matter.
Sincerely,
