Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

High-res image reconstruction with latent diffusion models from human brain

github.com

Post reply on HN
Print view View original

161–170 of 170 posts

Re: High-res image reconstruction with latent diffusion models from human brain

#161

Post by armatav »

00F_ wrote: post #159
armatav wrote: post #156

Earlier quoted context omitted.

You can’t stop it - certainly not with regulation; so why so much concern for the “geopolitical equation”? Fear won’t change that - and we are at least 20 years away from a neuromorphic revolution. You’ve got enough time to come to grips with it.

lets say there was a global coalition of countries that considered the creation or advancement of AGI a material threat to the safety of all humans. this would qualify countries that do advanced AI research for retaliation from NATO or other international bodies that might be created. it is clear that AI is only able to move forward in an environment of cheap compute and international academic collaboration. progress…

How exactly are you planning on regulating that in China, smartass?

Or did you forget a factor in that geopolitical equation of yours?

Re: High-res image reconstruction with latent diffusion models from human brain

#162

Post by 00F_ »

armatav wrote: post #161
00F_ wrote: post #159

Earlier quoted context omitted.

lets say there was a global coalition of countries that considered the creation or advancement of AGI a material threat to the safety of all humans. this would qualify countries that do advanced AI research for retaliation from NATO or other international bodies that might be created. it is clear that AI is only able to move forward in an environment of cheap compute and international academic collaboration. progress…

How exactly are you planning on regulating that in China, smartass? Or did you forget a factor in that geopolitical equation of yours?

i have no idea why people go straight to china every time regulating AI is brought up. its something to do with a rudimentary understanding of geopolitics i think. china. the answer to your question is the same answer to the same question regarding any country: a healthy majority of world powers forming a military-backed coalition would definitely stop outliers from carrying out the type of research that is at the leading edge of the current AI explosion. the chinese government is already worried about AGI so its ironic that everyone imagines them to be the outlier because in fact they would probably one of the first and most enthusiastic members of such a coalition. any country that wanted to resist, and that is highly unlikely given the fact that the need for regulation will become blindingly clear with every surge forward, would much rather cooperate than pursue far-fetched geopolitical strategies that involve AGI. most countries dont even output enough research to qualify for sanctions.

nobody has ever offered a lucid and axiomatic argument that shows regulation cannot work. there are two options, TRY to regulate or face a living nightmare where neither the best nor the worst outcome is even close to acceptable. it is so blindingly obvious that it boggles my mind: the only reasonable, rational response is to try to regulate, slow or stop AGI.

edit: i have a guilty confession. i was looking through your comments. i saw that you said that humans must eat meat to be healthy. i was surprised to see that and i want to tell you that i completely agree with you and i have often tried to explain to people why this is true and its like talking to a brick wall. there arent very many people who seem to get this even though its blindingly obvious. just wanted to give you a little encouragement to keep the fight going on the meat thing.

Re: High-res image reconstruction with latent diffusion models from human brain

#163

Post by armatav »

00F_ wrote: post #162
armatav wrote: post #161

Earlier quoted context omitted.

How exactly are you planning on regulating that in China, smartass? Or did you forget a factor in that geopolitical equation of yours?

i have no idea why people go straight to china every time regulating AI is brought up. its something to do with a rudimentary understanding of geopolitics i think. china. the answer to your question is the same answer to the same question regarding any country: a healthy majority of world powers forming a military-backed coalition would definitely stop outliers from carrying out the type of research that is at the le…

Haha, yeah that guy had a thing against “carnists”. Deranged.

I don’t think as the tech curve goes up there will be a long enough time period, even with a globally enforced military pact, to stop the rise of the machine. My reasoning is that there are more than enough clandestine organizations and families with a vested interest in pursuing the “power” it brings, a lot of them with a whole lot of control over these countries in our pact.

To each their own. Keep eating meat.

Re: High-res image reconstruction with latent diffusion models from human brain

#164

Post by nfgrep »

00F_ wrote: post #153
TechBro8615 wrote: post #119

Earlier quoted context omitted.

When in human history have we ever been able to stop technological advancement? > we need regulation No, we don't. Regulation doesn't stop technological progress - it puts it in the hands of an elite few. And besides, there are 130+ regulatory jurisdictions. For example, the US government doesn't fund human cloning research, but that doesn't mean China won't fund it. Or perhaps you'd also like a one world government…

from your comment i am pretty confident that you are around 19 years old. it doesnt seem like you actually read my comment. i guess i will respond to you but there is no comment in the world that could set you straight. you need more experience and personal development before you could begin to understand this topic. someone once said "do not see things as they are, see them as they might be." this quote is really ab…

> you have no idea what AGI will lead to

Neither do you? None of us do, in fact I’d imagine the people trying for AGI right now would have a better guess than you or I.

> there would be an immediate and total power vacuum caused by the advancements. these advancements would be so huge that it would change the geopolitical equation beyond recognition.

This sounds like you’re assuming someone will flip a switch one day and the most powerful mind in history will be let loose. I’m not sure AGI will advance that fast. We might have alot of incredibly “stupid” iterations of AGIs first, for many years before a clever one rolls around.

> this is intrinsic and unavoidable. it cannot be disproved or denied.

Were all just making assumptions here, I don’t think yours get to be called “intrinsic and unavoidable”.

I understand the concerns here, but if you’re willing to claim the end of the world, I would suggest basing your claims on something, or atleast making your assumptions explicit. E.g. “assuming we achieve AGI, and its equipped to rapidly become more powerful/intelligent than the whole of the human population…)

Re: High-res image reconstruction with latent diffusion models from human brain

#165

Post by 00F_ »

nfgrep wrote: post #164
00F_ wrote: post #153

Earlier quoted context omitted.

from your comment i am pretty confident that you are around 19 years old. it doesnt seem like you actually read my comment. i guess i will respond to you but there is no comment in the world that could set you straight. you need more experience and personal development before you could begin to understand this topic. someone once said "do not see things as they are, see them as they might be." this quote is really ab…

> you have no idea what AGI will lead to Neither do you? None of us do, in fact I’d imagine the people trying for AGI right now would have a better guess than you or I. > there would be an immediate and total power vacuum caused by the advancements. these advancements would be so huge that it would change the geopolitical equation beyond recognition. This sounds like you’re assuming someone will flip a switch one day…

you can predict the behavior of complex systems axiomatically. my predictions are very general because they are axiomatic. the most important axiom in play is that natural selection will guide the development and behaviour of the creatures of the singularity. there may be points of friction that cause small deviations from this path, such as the total effort of all humans post-pandoras-box, but the ultimate shape of things is inevitable. these are assumptions in name only.

there are many possibilities so the idea that we get an outcome that is good for us is unlikely. its just basic probability. i think people get hung up on this because there isnt an example of it to reference in history.

of course AGI will immediately rocket upward. the only way it wouldnt is if it were created in total secrecy and held in perfect captivity forever. laughable. all that is needed is for word to get out that AGI has been created and it would be re-created the next day somewhere else. and one iteration of it would rocket upward. AGI, once created, is intrinsically unstable.

the burden of evidence and proof is on you, not me. we know what things will be like without AGI. it is only right for the people who advocate for the creation of sentient machines to produce evidence that they will not open the doors to a living nightmare. the same thing should have been done with nuclear weapons. it really makes me scratch my head when people demand evidence from me as if i were the one encroaching. you are right, people are only making assumptions when they talk about the singularity. and the idea that we will not bitterly regret the singularity is the most tenuous assumption of all. until they show up with something more substantive i will be firmly against the creation of AGI.

Re: High-res image reconstruction with latent diffusion models from human brain

#166

Post by maxerickson »

notfed wrote: post #49
donohoe wrote: post #29

As people and groups increasingly move this direction do we think about vectors for abuse in 10, 20 or 50+ years? The human mind is considered the only place where we have true privacy. All these efforts are taking that away. At this rate all notions of privacy will soon be dead.

If this technology becomes accessible to courtrooms or police, they will use it. There will never be a way to encrypt thoughts.

When they tell you to think about what you were doing at the bank on Tuesday, think about what you were doing at the bank on Friday.

Re: High-res image reconstruction with latent diffusion models from human brain

#168

Post by sampo »

Aransentin wrote: post #15

I immediately found the results suspect, and think I have found what is actually going on. The dataset it was trained on was 2770 images, minus 982 of those used for validation. I posit that the system did not actually read any pictures from the brains, but simply overfitted all the training images into the network itself. For example, if one looks at a picture of a teddy bear, you'd get an overfitted picture of anot…

> The dataset it was trained on was 2770 images, minus 982 of those used for validation.

I don't think you got that 2770 correct. Might be 9250 images, minus 982 (that one you got right). Then again, the paper is so badly written, I find it difficult to decipher what they did. From section 3.1:

Briefly, NSD provides data acquired from a 7-Tesla fMRI scanner over 30–40 sessions during which each subject viewed three repetitions of 10,000 images. We analyzed data for four of the eight subjects who completed all imaging sessions (subj01, subj02, subj05, and subj07).

We used 27,750 trials from NSD for each subject (2,250 trials out of the total 30,000 trials were not publicly released by NSD). For a subset of those trials (N=2,770 trials), 982 images were viewed by all four subjects. Those trials were used as the test dataset, while the remaining trials (N=24,980) were used as the training dataset.

https://www.biorxiv.org/content/10.1101/2022.11.18.517004v2....

Re: High-res image reconstruction with latent diffusion models from human brain

#170

Post by lproven »

csomar wrote: post #131
hiddencost wrote: post #67

Earlier quoted context omitted.

Nope. If you train a model where the input is an integer between 1 and 10, and the output is a specific image from a set of ten, the model will be able to get zero loss on the task. That is what's happening here.

But unless they tested this on a single human being; doesn't this mean that we can read brains (it's just this one particular reader is bad).

From the paper, it's four people from the NSD:

« We analyzed data for four of the eight subjects who completed all imaging sessions (subj01, subj02, subj05, and subj07 »

P3 here: https://www.biorxiv.org/content/biorxiv/early/2022/11/21/202...

Post reply on HN