4 ms·

Sorry, I had to do it. I asked GPT3... How can we prevent drowning in garbage generated by AI? > Preventing drowning in AI-generated garbage can be a challeng

by baxtr 4y ago

Sorry, I had to do it. I asked GPT3... How can we prevent drowning in garbage generated by AI? > Preventing drowning in AI-generated garbage can be a challenging task, but there are a few things that can be done to reduce the amount of garbage that is generated. One approach is to carefully design and train AI systems to generate high-quality content, rather than relying on large amounts of data to produce mediocre results. Additionally, using tools and techniques such as data filters and validation can help to identify and remove low-quality or irrelevant content. Finally, creating and enforcing strict guidelines and standards for the use and generation of AI-generated content can help to prevent the proliferation of garbage.

notacoward 4y ago
That's honestly one of the best comments in this entire thread. Instead of veering predictably into philosophy or anecdote, it provides concrete and feasible steps toward improvement. No better way to make the point that "AI generated" is not the problem and "garbage" is.
- Nathanba 4y ago
  I find the AI competition to be yet another powerful reminder that all that matters are results. I can't count how much time I've spent being angry at people in my life, even though I always knew that it was stupid. Angry at bad docs, bad programmers, bad decisions, their inability to think or understand or care or feel. Now the AI removes it all. The AI generates truth out of thin air faster than any human can invent a lie about it. The AI simply generates good docs from bad docs, all the anger was always irrelevant. People were never going to become better, once again technology simply creates a world where people are automatically better by the mere fact that life is so much easier that the humans have no justification anymore to lie and cheat and pretend that the docs were actually always good and how I'm the problem. Nope, the AI just fixes it immediately and the human can remain lazy and ignorant, never the wiser that he actually failed at his job. I think it will be wonderful, no longer will we have to argue with e.g c++ programmers about whether their language is bad. The AI will make it instantly obvious that you won't want to risk programming in this nonsense. No longer will we have to debate people on whether some political issue is right or wrong. The AI will generate so many truths out of thin air that it counters the DDOS of bad people's lies faster than they can spin lies. Once again technology is going to save us from human nature, humans are allowed to remain silently bad. And the best part: Bad people are usually lazy, they will trust the AI far more than they ever trust other bad people or themselves. They will simply start blindly trusting the AI, which is better than what we have now: Them inventing their own lies and trusting themselves.
  - practice9 4y ago
    It will remove a large amount of stress from people's lives for sure. When an AI model can explain the non-obvious bug in the code written by the developer who left the company 2 years ago - that's quite powerful. Saves more time for other stuff Though at some point it may create new sources of stress (like much discussed job security when the other job candidate is the AI model)
    - visarga 4y ago
      A company firing its human experts in favour of AI is a bad sign, probably going to fail. If business was good they would hire additional humans and use more AI. The purpose of a company is not to reduce salary costs, but to increase profits.
- EamonnMR 4y ago
  In this case though, it's not a solution at all. The question isn't how producers of garbage produce less garbage, it's how consumers of content can avoid being inundated by garbage. We may well ask 'please don't produce less garbage' and the producers have given us a resounding 'no.'
  - visarga 4y ago
    Fight fire with fire. Local language model doing question answering takes search and web garbage in, outputs polite and helpful responses without ads and tracking. But let's take this idea one step further. We could also have a neural index. There are a few promising papers.
- practice9 4y ago
  An important threshold will be when an important physics, math or biology problem is going to be solved with one of these AIs, or by a scientist who uses AI for "ghostwriting". If scientists will verify the solution and find it to be correct, the mainstream sentiment about "garbage" will change very quickly. Maybe except various neo-luddites who will find different reasons to hate these models
ryandrake 4y ago
I notice it did not offer a solution that involves outright banning itself… The ghost in the machine is showing
- somrand0 4y ago
  because maybe it understands that banning stuff is not to be done lightly. any action that restricts anyone's individual freedom must be very carefully vetted. because without freedom we will die. but this also applies to the corporations whose rights are now in opposition with our own human individual necessities. if I want to gift a copy of a copyrighted material owned by a corporation, I can still do it, but it's getting more difficult; and I in fact may well be prosecuted criminally for doing that. this is an example of my right to do what I want (give a copied gift) stands in opposition with the right of a corporation to get its way. I fear human individuals are losing this battle.
Vox_Leone 4y ago
A propos, has anybody asked The Last Question*?[0] "Can the entropy of the universe be reversed?" [0]https://en.wikipedia.org/wiki/The_Last_Question https://en.wikipedia.org/wiki/The_Last_Question
- _shantaram 4y ago
  Putting the question in verbatim led to a somewhat unimaginative, albeit technically correct response. > The entropy of the universe, also known as the thermodynamic arrow of time, is a measure of the disorder or randomness of a system. It is a fundamental concept in thermodynamics, and it is widely accepted that the entropy of the universe is constantly increasing over time. This is because according to the second law of thermodynamics, the entropy of a closed system can never decrease.
nonbirithm 4y ago
I think that the root of the problem is not algorithms like the article suggests. It's ordinary people that can't resist the temptation of using AI to generate coherent-seeming results and flooding the Internet with them. But I don't blame you since I find AI tools to be useful in certain contexts, and I am unable to resist tinkering with Stable Diffusion. I guess the future for me will be one of navigating the resulting cognitive dissonance as I sift through pages of autogenerated SEO spam. Maybe there could be a codified set of guidelines for how AI content should be shared. Or maybe an annotation system that separates the two across different websites? I only ever share something that comes from an AI on a siloed website just for AI content, but that's only because I have my own beliefs. More and more people on this site are asking if certain comments were generated with GPT-3 and it's a worrying sign.
tenebrisalietum 4y ago
Curious of GPT3's response to the opposite side of that question: *How can we enable drowning in garbage generated by AI?*
- baxtr 4y ago
  Here you go: > It is not advisable to try and enable drowning in garbage generated by AI. Rather, we should strive to prevent the generation of AI-generated garbage in the first place. This can be done by carefully designing and training AI systems to generate high-quality content, using tools and techniques such as data filters and validation to identify and remove low-quality content, and by creating and enforcing strict guidelines and standards for the use and generation of AI-generated content.
visarga 4y ago
Spot on with verification. It's going to be big. There could be neural nets specialised in verification.
nova27 4y ago
"One approach is to carefully design and train AI systems to generate high-quality content" - sure, but how will you accomplish this is the million dollar question

by baxtr 4y ago

notacoward 4y ago

That's honestly one of the best comments in this entire thread. Instead of veering predictably into philosophy or anecdote, it provides concrete and feasible steps toward improvement. No better way to make the point that "AI generated" is not the problem and "garbage" is.

Nathanba 4y ago
I find the AI competition to be yet another powerful reminder that all that matters are results. I can't count how much time I've spent being angry at people in my life, even though I always knew that it was stupid. Angry at bad docs, bad programmers, bad decisions, their inability to think or understand or care or feel. Now the AI removes it all. The AI generates truth out of thin air faster than any human can invent a lie about it. The AI simply generates good docs from bad docs, all the anger was always irrelevant. People were never going to become better, once again technology simply creates a world where people are automatically better by the mere fact that life is so much easier that the humans have no justification anymore to lie and cheat and pretend that the docs were actually always good and how I'm the problem. Nope, the AI just fixes it immediately and the human can remain lazy and ignorant, never the wiser that he actually failed at his job. I think it will be wonderful, no longer will we have to argue with e.g c++ programmers about whether their language is bad. The AI will make it instantly obvious that you won't want to risk programming in this nonsense. No longer will we have to debate people on whether some political issue is right or wrong. The AI will generate so many truths out of thin air that it counters the DDOS of bad people's lies faster than they can spin lies. Once again technology is going to save us from human nature, humans are allowed to remain silently bad. And the best part: Bad people are usually lazy, they will trust the AI far more than they ever trust other bad people or themselves. They will simply start blindly trusting the AI, which is better than what we have now: Them inventing their own lies and trusting themselves.
- practice9 4y ago
  It will remove a large amount of stress from people's lives for sure. When an AI model can explain the non-obvious bug in the code written by the developer who left the company 2 years ago - that's quite powerful. Saves more time for other stuff Though at some point it may create new sources of stress (like much discussed job security when the other job candidate is the AI model)
  - visarga 4y ago
    A company firing its human experts in favour of AI is a bad sign, probably going to fail. If business was good they would hire additional humans and use more AI. The purpose of a company is not to reduce salary costs, but to increase profits.
EamonnMR 4y ago
In this case though, it's not a solution at all. The question isn't how producers of garbage produce less garbage, it's how consumers of content can avoid being inundated by garbage. We may well ask 'please don't produce less garbage' and the producers have given us a resounding 'no.'
- visarga 4y ago
  Fight fire with fire. Local language model doing question answering takes search and web garbage in, outputs polite and helpful responses without ads and tracking. But let's take this idea one step further. We could also have a neural index. There are a few promising papers.
practice9 4y ago
An important threshold will be when an important physics, math or biology problem is going to be solved with one of these AIs, or by a scientist who uses AI for "ghostwriting". If scientists will verify the solution and find it to be correct, the mainstream sentiment about "garbage" will change very quickly. Maybe except various neo-luddites who will find different reasons to hate these models

ryandrake 4y ago

I notice it did not offer a solution that involves outright banning itself… The ghost in the machine is showing

somrand0 4y ago
because maybe it understands that banning stuff is not to be done lightly. any action that restricts anyone's individual freedom must be very carefully vetted. because without freedom we will die. but this also applies to the corporations whose rights are now in opposition with our own human individual necessities. if I want to gift a copy of a copyrighted material owned by a corporation, I can still do it, but it's getting more difficult; and I in fact may well be prosecuted criminally for doing that. this is an example of my right to do what I want (give a copied gift) stands in opposition with the right of a corporation to get its way. I fear human individuals are losing this battle.

Vox_Leone 4y ago

A propos, has anybody asked The Last Question*?[0] "Can the entropy of the universe be reversed?" [0]https://en.wikipedia.org/wiki/The_Last_Question https://en.wikipedia.org/wiki/The_Last_Question

_shantaram 4y ago
Putting the question in verbatim led to a somewhat unimaginative, albeit technically correct response. > The entropy of the universe, also known as the thermodynamic arrow of time, is a measure of the disorder or randomness of a system. It is a fundamental concept in thermodynamics, and it is widely accepted that the entropy of the universe is constantly increasing over time. This is because according to the second law of thermodynamics, the entropy of a closed system can never decrease.

nonbirithm 4y ago

I think that the root of the problem is not algorithms like the article suggests. It's ordinary people that can't resist the temptation of using AI to generate coherent-seeming results and flooding the Internet with them. But I don't blame you since I find AI tools to be useful in certain contexts, and I am unable to resist tinkering with Stable Diffusion. I guess the future for me will be one of navigating the resulting cognitive dissonance as I sift through pages of autogenerated SEO spam. Maybe there could be a codified set of guidelines for how AI content should be shared. Or maybe an annotation system that separates the two across different websites? I only ever share something that comes from an AI on a siloed website just for AI content, but that's only because I have my own beliefs. More and more people on this site are asking if certain comments were generated with GPT-3 and it's a worrying sign.

tenebrisalietum 4y ago

Curious of GPT3's response to the opposite side of that question: *How can we enable drowning in garbage generated by AI?*

baxtr 4y ago
Here you go: > It is not advisable to try and enable drowning in garbage generated by AI. Rather, we should strive to prevent the generation of AI-generated garbage in the first place. This can be done by carefully designing and training AI systems to generate high-quality content, using tools and techniques such as data filters and validation to identify and remove low-quality content, and by creating and enforcing strict guidelines and standards for the use and generation of AI-generated content.

visarga 4y ago

Spot on with verification. It's going to be big. There could be neural nets specialised in verification.

nova27 4y ago

"One approach is to carefully design and train AI systems to generate high-quality content" - sure, but how will you accomplish this is the million dollar question