▲ 684 ▼ OpenAI confirms that AI writing detectors don’t work (arstechnica.com) submitted 2 years ago by fne8w2ah@lemmy.world to c/technology@lemmy.world 113 comments fedilink hide all child comments
[–] stevedidWHAT@lemmy.world 3 points 2 years ago (2 children) Isn’t that article about GANs? Isn’t GPT not a GAN? permalink fedilink source parent hideshow 4 child comments replies: [–] PetDinosaurs@lemmy.world 5 points 2 years ago (1 child) It almost certainly has some gan-like pieces. Gans are part of the NN toolbox, like cnns and rnns and such. Basically all commercial algorithms (not just nns, everything) are what I like to call "hybrid" methods, which means keep throwing different tools at it until things work well enough. permalink fedilink source parent hideshow 2 child comments replies: [–] stevedidWHAT@lemmy.world 3 points 2 years ago* (1 child) The findings were for GAN models, not GAN like components though. permalink fedilink source parent hideshow 2 child comments replies: [–] PetDinosaurs@lemmy.world 1 point 2 years ago (1 child) It doesn't matter. Even the training process makes it pretty much impossible to tell these things apart. And if we do find a way to distinguish, we'll immediately incorporate that into the model design in a GAN like manner, and we'll soon be unable to distinguish again. permalink fedilink source parent hideshow 2 child comments replies: [–] stevedidWHAT@lemmy.world 0 points 2 years ago Which is why hardcoded fingerprints/identifications are required to identify the individual as a speaker rather than as an AI vs Human. Which is what we’re ultimately agreeing on here outside of the pedantics of the article and scientific findings: Trying to find the model who is supposed to be human as an AI is counter intuitive. They’re direct opposites if one works, both can’t be exist in this implementation. The hard part will obviously be making sure that such a “fingerprint” wouldn’t be removable which will take some wild math and out of the box thinking I’m sure. Tough problem! permalink fedilink source parent [–] bioemerl@kbin.social 2 points 2 years ago It's not even about diffusion models. Adversarial networks are basically obsolete permalink fedilink source parent
[–] PetDinosaurs@lemmy.world 5 points 2 years ago (1 child) It almost certainly has some gan-like pieces. Gans are part of the NN toolbox, like cnns and rnns and such. Basically all commercial algorithms (not just nns, everything) are what I like to call "hybrid" methods, which means keep throwing different tools at it until things work well enough. permalink fedilink source parent hideshow 2 child comments replies: [–] stevedidWHAT@lemmy.world 3 points 2 years ago* (1 child) The findings were for GAN models, not GAN like components though. permalink fedilink source parent hideshow 2 child comments replies: [–] PetDinosaurs@lemmy.world 1 point 2 years ago (1 child) It doesn't matter. Even the training process makes it pretty much impossible to tell these things apart. And if we do find a way to distinguish, we'll immediately incorporate that into the model design in a GAN like manner, and we'll soon be unable to distinguish again. permalink fedilink source parent hideshow 2 child comments replies: [–] stevedidWHAT@lemmy.world 0 points 2 years ago Which is why hardcoded fingerprints/identifications are required to identify the individual as a speaker rather than as an AI vs Human. Which is what we’re ultimately agreeing on here outside of the pedantics of the article and scientific findings: Trying to find the model who is supposed to be human as an AI is counter intuitive. They’re direct opposites if one works, both can’t be exist in this implementation. The hard part will obviously be making sure that such a “fingerprint” wouldn’t be removable which will take some wild math and out of the box thinking I’m sure. Tough problem! permalink fedilink source parent
[–] stevedidWHAT@lemmy.world 3 points 2 years ago* (1 child) The findings were for GAN models, not GAN like components though. permalink fedilink source parent hideshow 2 child comments replies: [–] PetDinosaurs@lemmy.world 1 point 2 years ago (1 child) It doesn't matter. Even the training process makes it pretty much impossible to tell these things apart. And if we do find a way to distinguish, we'll immediately incorporate that into the model design in a GAN like manner, and we'll soon be unable to distinguish again. permalink fedilink source parent hideshow 2 child comments replies: [–] stevedidWHAT@lemmy.world 0 points 2 years ago Which is why hardcoded fingerprints/identifications are required to identify the individual as a speaker rather than as an AI vs Human. Which is what we’re ultimately agreeing on here outside of the pedantics of the article and scientific findings: Trying to find the model who is supposed to be human as an AI is counter intuitive. They’re direct opposites if one works, both can’t be exist in this implementation. The hard part will obviously be making sure that such a “fingerprint” wouldn’t be removable which will take some wild math and out of the box thinking I’m sure. Tough problem! permalink fedilink source parent
[–] PetDinosaurs@lemmy.world 1 point 2 years ago (1 child) It doesn't matter. Even the training process makes it pretty much impossible to tell these things apart. And if we do find a way to distinguish, we'll immediately incorporate that into the model design in a GAN like manner, and we'll soon be unable to distinguish again. permalink fedilink source parent hideshow 2 child comments replies: [–] stevedidWHAT@lemmy.world 0 points 2 years ago Which is why hardcoded fingerprints/identifications are required to identify the individual as a speaker rather than as an AI vs Human. Which is what we’re ultimately agreeing on here outside of the pedantics of the article and scientific findings: Trying to find the model who is supposed to be human as an AI is counter intuitive. They’re direct opposites if one works, both can’t be exist in this implementation. The hard part will obviously be making sure that such a “fingerprint” wouldn’t be removable which will take some wild math and out of the box thinking I’m sure. Tough problem! permalink fedilink source parent
[–] stevedidWHAT@lemmy.world 0 points 2 years ago Which is why hardcoded fingerprints/identifications are required to identify the individual as a speaker rather than as an AI vs Human. Which is what we’re ultimately agreeing on here outside of the pedantics of the article and scientific findings: Trying to find the model who is supposed to be human as an AI is counter intuitive. They’re direct opposites if one works, both can’t be exist in this implementation. The hard part will obviously be making sure that such a “fingerprint” wouldn’t be removable which will take some wild math and out of the box thinking I’m sure. Tough problem! permalink fedilink source parent
[–] bioemerl@kbin.social 2 points 2 years ago It's not even about diffusion models. Adversarial networks are basically obsolete permalink fedilink source parent