|
In order to ensure that this thread meets TL standards and follows the proper guidelines, we ask that everyone please adhere to this mod note. Posts containing only Tweets or articles adds nothing to the discussions. Therefore, when providing a source, explain why you feel it is relevant and what purpose it adds to the discussion. Also take note that unsubstantiated tweets/posts meant only to rekindle old arguments will be actioned upon. All in all, please continue to enjoy posting in TL General and partake in discussions as much as you want! But please be respectful when posting or replying to someone. There is a clear difference between constructive criticism/discussion and just plain being rude and insulting. https://www.registertovote.service.gov.uk |
On May 17 2025 23:39 Yurie wrote:Show nested quote +On May 17 2025 23:35 Billyboy wrote:On May 17 2025 22:26 BlackJack wrote:
Nobody mentioned the single biggest problem with this idea. If you need a hint for what that is google racial disparities in crime statistics and then imagine what happens when you try to program that into an AI hive mind that predicts your likelihood for committing crime. My experience on this forum tells me even if some of you are okay with this idea on paper none of you will be okay if the conclusions of the program end up being unsavory, regardless of how accurate they are. There are other people in this thread that are going to be really mad if it finds no reason to have racial bias, but a mathematical one for socioeconomic conditions or whatever. Isn't your postal code one of the classic indicators of future outcomes? Any big data crunching will likely find that relationship quickly. Living in a rich area you likely have successful parents and your peers are also from that background so you push the basic level up. Facilities, extracurriculars and so on are often better. While living in a very poor area the school is bad, there are no jobs. Parents are likely less successful and less likely to push you and your peers. With enough data access you could probably find strange and interesting data though. Random example, got drivers license as soon as allowed, played football while young and spend more than average on alcohol could be an extreme indicator, much stronger than any individual thing.
Agreed.
On May 17 2025 23:42 WombaT wrote:Show nested quote +On May 17 2025 23:35 Billyboy wrote:On May 17 2025 22:26 BlackJack wrote:
Nobody mentioned the single biggest problem with this idea. If you need a hint for what that is google racial disparities in crime statistics and then imagine what happens when you try to program that into an AI hive mind that predicts your likelihood for committing crime. My experience on this forum tells me even if some of you are okay with this idea on paper none of you will be okay if the conclusions of the program end up being unsavory, regardless of how accurate they are. There are other people in this thread that are going to be really mad if it finds no reason to have racial bias, but a mathematical one for socioeconomic conditions or whatever. I don’t think that would make many on here mad. We already know this, it’s precisely why many of us think trying to equalise socioeconomic conditions is so important to do.
Sensible people won't, people who think skin colour is what makes people good or bad will.
On May 17 2025 23:54 BlackJack wrote:Show nested quote +On May 17 2025 23:35 Billyboy wrote:On May 17 2025 22:26 BlackJack wrote:
Nobody mentioned the single biggest problem with this idea. If you need a hint for what that is google racial disparities in crime statistics and then imagine what happens when you try to program that into an AI hive mind that predicts your likelihood for committing crime. My experience on this forum tells me even if some of you are okay with this idea on paper none of you will be okay if the conclusions of the program end up being unsavory, regardless of how accurate they are. There are other people in this thread that are going to be really mad if it finds no reason to have racial bias, but a mathematical one for socioeconomic conditions or whatever. Spoiler alert - there’s also racial disparities in “socioeconomic conditions or whatever.” It would be foolish to think people would be satisfied if the disparities in race that the machine spits out is because of environmental reasons as opposed to “melanin content.” Any discrepancy is going to be extremely problematic regardless of the variables that are causing the discrepancies.
You have spent way to much reading facebook meme's and looking for reasons to be mad at the "left". People on the left don't think there are not a higher percentage of, for example Blacks in America, committing crime. They believe that it is no skin colour or genetically related.
If AI proves them right they are not going to be mad.
|
United States10501 Posts
Did you understand my post? If the AI hive mind uses the statistic that blacks commit more crime to predict they may be more dangerous and as a result maybe they should be policed differently, then no reasoning in the world for this conclusion is going to satisfy anyone on the left. Your delusion that the left would be okay with this so long as the conclusion is not based on black people being “genetically inferior” is pure fairytale.
|
Norway28667 Posts
Yeah I honestly don't know of people who reject that there is a racial difference in amount of crime committed. However, there are two caveats: many will argue that the real difference is likely to be smaller than what the difference is for punishment received (drug use is a well-documented example of this), and any real difference is more likely to be caused by environmental factors (not racist) than cultural factors (racist?) than biological factors (racist and this idea is basically entirely rejected by the left).
|
I think you're talking past each other.
BJ is saying: if the algorithm says that someone from Tottenham is more likely to commit a crime than someone from Notting Hill, then the police will intensify their presence in Tottenham following Tottenham locals around, frisking them, etc. Whereas Notting Hill residents go about their business unharrassed and carefree.
Instead of addressing the latter there's a lot of people saying the former is both obvious and not something to worry about. That isn't the main argument though.
That said, policing in Tottenham is already different than it is in Notting Hill. The cops don't need an algorithm for that. Will an algorithm make it worse? Maybe? Maybe not.
|
On May 18 2025 00:47 BlackJack wrote:
Did you understand my post? If the AI hive mind uses the statistic that blacks commit more crime to predict they may be more dangerous and as a result maybe they should be policed differently, then no reasoning in the world for this conclusion is going to satisfy anyone on the left. Your delusion that the left would be okay with this so long as the conclusion is not based on black people being “genetically inferior” is pure fairytale.
Yes I did, and your further explanation proved it. Your version on the "left" is not the real left, sure there are loonies just like everywhere, but the vast majority are like...
On May 18 2025 00:58 Liquid`Drone wrote:
Yeah I honestly don't know of people who reject that there is a racial difference in amount of crime committed. However, there are two caveats: many will argue that the real difference is likely to be smaller than what the difference is for punishment received (drug use is a well-documented example of this), and any real difference is more likely to be caused by environmental factors (not racist) than cultural factors (racist?) than biological factors (racist and this idea is basically entirely rejected by the left).
Agreed.
|
On May 17 2025 22:56 WombaT wrote:Show nested quote +On May 17 2025 19:07 Acrofales wrote:
UK is a bit scarier than most places due to the sheer amount of CCTV cameras you decided to install, but in general, I'd worry more about Google and Meta gathering your data than the government doing so directly. Best case they only use the data to sell you stuff and maybe pass it on to the government. Worst case, your profile is for sale to the highest bidder, e.g. Cambridge Analytics. I didn’t even realise we were notably CCTV heavy, I thought that was just generally the trend. It’s not something that massively bothers me as generally it’s a camera pointed at me, there’s not someone actually watching me. If CCTV footage is being pored through, it’s after the fact and investigatory, and my presence is going to be incidental. It may seem a somewhat arbitrary distinction to some, but I think it is quite a meaningful one. Myself and colleagues had zero issues with security cameras being a thing in our place of work. Many felt uneasy when we were told they’d become fully manned. Then the dynamic changes, is some busybody watching you and will they report non-serious infractions? Security cameras already covered the serious stuff like theft, or if an assault happened. As it turned out not much functionally changed. It’s pretty prohibitively difficult to actually survey a couple of hundred folks moving in and out. Even if you really wanted to be an arse about it. I totally understand some concerns, albeit I find it a little frustrating at the same time. Governments could do a lot of good with modern tech advances in various spheres, but it’s generally a no-go with the public out of privacy concerns. Reasonable concerns but then the public by and large just voluntarily cede privacy to corporate actors. Perhaps not this pre-crime example, I’m not sure on this one. But things like preventative medical screening. I think you could build more robust transport infrastructure if you hoovered up real-time data and crunched it. Just spitballing really but there are some pretty good use cases for big data in the public sector, folks just tend to be very, very reticent. Which I have no issue with, but it feels they don’t apply this reticence to other actors.
Regarding CCTV - first I will take workplace - I would say there is a difference between workplace and public space, also if I recall and nothing changed I believe company cannot review CCTV footage without police. There was some accident in place I worked at around 10 years ago and company had to ask police to come to check the footage.
"It’s pretty prohibitively difficult to actually survey a couple of hundred folks moving in and out. Even if you really wanted to be an arse about it. " - yes, for a human, add facial recognition software, AI and it becomes non effort.
"I totally understand some concerns, albeit I find it a little frustrating at the same time. Governments could do a lot of good with modern tech advances in various spheres, but it’s generally a no-go with the public out of privacy concerns."
Of course they can, it is the will which is the issue. I mean once they got nuclear power, power plants werent exactly first thing they used it for. I think you believe that government and citizens have the same goals, they however dont. Those two are separate entities.
"Reasonable concerns but then the public by and large just voluntarily cede privacy to corporate actors."
Bolded makes massive difference here. What corporate actors end of the day can do? Now lets exclude banks from those. And take FB for example - you can decide that you dont want to have an account and it is fine, you in no way, shape or form are obliged to behave in accord with FB policies. Even if you do decide to use FB, and break T&C, most they can do is ban you. More so if you are mistreated by them odds are government can actually protect you.
Now government - whether you like it, or not, you are bound by its policies, it is not exactly a choice. On top of that if you decide to break government policies, you just go to prison.
So you see comparison is rather unfair. One cant really do shit to you, other can do with you whatever it wants.
On May 17 2025 23:14 WombaT wrote:Show nested quote +On May 17 2025 22:26 BlackJack wrote:
Nobody mentioned the single biggest problem with this idea. If you need a hint for what that is google racial disparities in crime statistics and then imagine what happens when you try to program that into an AI hive mind that predicts your likelihood for committing crime. My experience on this forum tells me even if some of you are okay with this idea on paper none of you will be okay if the conclusions of the program end up being unsavory, regardless of how accurate they are. From the article, to me anyway it feels people who’ve already a decent gauge on predictive factors are pulling in data they think is relevant and building this thing. Rather than just unleashing an AI to crunch all the data. Prior convictions for domestic violence for example. It would stand to reason that someone who had such a conviction would be a person of interest here as it were. It’s less ‘sexy’ for the tabloids but murder via domestic violence is right up there in terms of overall murder rate. I’m not one who is massively concerned with such modelling, the thing here I’m somewhat confused by is how is this intended to be used? I’m still unsure what the end goal and application here is, by the sounds of it much of this thread is in the same boat.
bolded - it doesnt matter. It either wont get used (then why do it in the first place), or get used (which is effectively punishing someone for something he didnt do). So for example: police will check on you every now and again, thats not very nice, is it, considering you didnt do anything. I mean your neighbours and people from your work would probably have some talk about it. ( on the plus side though, your lunch would never go missing  )
On May 17 2025 22:26 BlackJack wrote:
Nobody mentioned the single biggest problem with this idea. If you need a hint for what that is google racial disparities in crime statistics and then imagine what happens when you try to program that into an AI hive mind that predicts your likelihood for committing crime. My experience on this forum tells me even if some of you are okay with this idea on paper none of you will be okay if the conclusions of the program end up being unsavory, regardless of how accurate they are.
You think thats the problem now??? Isnt terrorist prediction tool logical follow up? Bad times to be Irish or Muslim in UK.
Edit: some typo.
|
United States10501 Posts
On May 18 2025 08:33 Billyboy wrote:Show nested quote +On May 18 2025 00:47 BlackJack wrote:
Did you understand my post? If the AI hive mind uses the statistic that blacks commit more crime to predict they may be more dangerous and as a result maybe they should be policed differently, then no reasoning in the world for this conclusion is going to satisfy anyone on the left. Your delusion that the left would be okay with this so long as the conclusion is not based on black people being “genetically inferior” is pure fairytale. Yes I did, and your further explanation proved it. Your version on the "left" is not the real left, sure there are loonies just like everywhere, but the vast majority are like...
Riiiiiight… people on the left would totally be okay with racial disparities in policing so long as the AI mastermind tells them there is good reason for it not based on skin color… lmao. Believe what you want.
|
United States42689 Posts
On May 18 2025 10:15 BlackJack wrote:Show nested quote +On May 18 2025 08:33 Billyboy wrote:On May 18 2025 00:47 BlackJack wrote:
Did you understand my post? If the AI hive mind uses the statistic that blacks commit more crime to predict they may be more dangerous and as a result maybe they should be policed differently, then no reasoning in the world for this conclusion is going to satisfy anyone on the left. Your delusion that the left would be okay with this so long as the conclusion is not based on black people being “genetically inferior” is pure fairytale. Yes I did, and your further explanation proved it. Your version on the "left" is not the real left, sure there are loonies just like everywhere, but the vast majority are like... Riiiiiight… people on the left would totally be okay with racial disparities in policing so long as the AI mastermind tells them there is good reason for it not based on skin color… lmao. Believe what you want.
Why are we assuming the AI will say the answer is to discriminate against black people? Not a gotcha, but isn’t it more likely to say that we need more affordable housing.
|
On May 18 2025 10:15 BlackJack wrote:Show nested quote +On May 18 2025 08:33 Billyboy wrote:On May 18 2025 00:47 BlackJack wrote:
Did you understand my post? If the AI hive mind uses the statistic that blacks commit more crime to predict they may be more dangerous and as a result maybe they should be policed differently, then no reasoning in the world for this conclusion is going to satisfy anyone on the left. Your delusion that the left would be okay with this so long as the conclusion is not based on black people being “genetically inferior” is pure fairytale. Yes I did, and your further explanation proved it. Your version on the "left" is not the real left, sure there are loonies just like everywhere, but the vast majority are like... Riiiiiight… people on the left would totally be okay with racial disparities in policing so long as the AI mastermind tells them there is good reason for it not based on skin color… lmao. Believe what you want.
Obviously they would AI said so:
https://www.theguardian.com/technology/2024/feb/22/google-pauses-ai-generated-images-of-people-after-ethnicity-criticism
|
United States10501 Posts
On May 18 2025 10:28 KwarK wrote:Show nested quote +On May 18 2025 10:15 BlackJack wrote:On May 18 2025 08:33 Billyboy wrote:On May 18 2025 00:47 BlackJack wrote:
Did you understand my post? If the AI hive mind uses the statistic that blacks commit more crime to predict they may be more dangerous and as a result maybe they should be policed differently, then no reasoning in the world for this conclusion is going to satisfy anyone on the left. Your delusion that the left would be okay with this so long as the conclusion is not based on black people being “genetically inferior” is pure fairytale. Yes I did, and your further explanation proved it. Your version on the "left" is not the real left, sure there are loonies just like everywhere, but the vast majority are like... Riiiiiight… people on the left would totally be okay with racial disparities in policing so long as the AI mastermind tells them there is good reason for it not based on skin color… lmao. Believe what you want. Why are we assuming the AI will say the answer is to discriminate against black people? Not a gotcha, but isn’t it more likely to say that we need more affordable housing.
I mean you yourself brought up the idea of responding to calls differently based on this predictive technology
On May 16 2025 22:46 KwarK wrote:
You wouldn’t sentence people for precrime off of it but you would treat police calls from the victim a little more urgently.
.
Like do you honestly think the predictive technology would assess that an elderly white woman poses the same threat as a black male youth? Of course not. Of course you’re going to send more calvary for the latter than the former if that’s the larger threat. It’s going to make all kinds of judgements based on race, religion, sex, etc.
|
On May 18 2025 11:26 BlackJack wrote:
Like do you honestly think the predictive technology would assess that an elderly white woman poses the same threat as a black male youth? Of course not. Of course you’re going to send more calvary for the latter than the former if that’s the larger threat. It’s going to make all kinds of judgements based on race, religion, sex, etc.
Entirely dependant on who codes it, the AI doesn't actually come up with it on its own. If it's coded to be discriminatory then it will be discriminatory. If not, then it won't.
|
United States10501 Posts
That goes to Rayzda’s point. Sure you can program google gemini to give you ridiculous responses. You can also program an AI to predict the Asian grandmother to be equally likely to nick your bike as anyone else. But if you’re going to program the AI to exist in a clown world then it’s not very useful to the real world.
|
A reminder that there is nothing intelligent about current gen ai. It has no concept of what it is doing. Its just reproducing what it has been fed.
|
I have plenty of problems with AI, but am I misremembering that AI made significant progress with protein folding, or were those just propaganda headlines?
|
On May 18 2025 13:27 BlackJack wrote:
That goes to Rayzda’s point. Sure you can program google gemini to give you ridiculous responses. You can also program an AI to predict the Asian grandmother to be equally likely to nick your bike as anyone else. But if you’re going to program the AI to exist in a clown world then it’s not very useful to the real world.
Presumably when the AI is involved there would already be a context, and then you could be discriminatory within that context. If some random dude is robbing a store and the AI says send two units, and then some random black dude is robbing a store and the AI says send four units, probability of violence increased, then that would be an example of a discriminatory AI. There's no scenario that matters in which the AI is to be asked if an old lady is more dangerous than a young man. We can sort of see that you understand the problem with your own argument because you should have a comparison between an old asian lady and an old black lady in order to test discrimination by race, but of course the old black lady isn't particularly likely to steal your bike either.
Edit: and we could even extend that to have a reverse clown world example, for example an AI that says that an old black lady is more likely to steal a bike than a young white guy because black people steal more bikes.
|
On May 18 2025 17:23 Gorsameth wrote:
A reminder that there is nothing intelligent about current gen ai. It has no concept of what it is doing. Its just reproducing what it has been fed.
On May 18 2025 23:54 GreenHorizons wrote:
I have plenty of problems with AI, but am I misremembering that AI made significant progress with protein folding, or were those just propaganda headlines?
I'll reply to both of you at once. For starters, @GH: was not propaganda, it was a Nobel prize winning breakthrough in chemistry. It wasn't quite the same type of algorithm as what Gorsameth is talking about, though. But regardless, I don't think that's a counterpoint: nobody at all ever said that AI is not good at matching patterns, which is what protein folding is: it's very complex patterns of a kind us humans look at and think are just a jumble of unpredictability, but we (1) know there are deterministic rules underlying how proteins fold which are 100% determined by the order of their aminoacids, and (2) have lots of data about them. This is something ML algorithms are designed to deal with par excellence, and AlphaFold did. Note, it was still an extremely hard problem: they had to come up with the right way to represent the data so that the algorithm could make sense of it, and just like they did in AlphaGo, they let it learn by example a bit, and then released it to improve upon itself. It was very impressive, but it just doesn't work the same way or do the same stuff as LLMs do. It's like if someone says "all FPS games are is just pointing your mouse and clicking to shoot", and countering with the argument "but Starcraft is pretty cool".
Now, regarding the original point, that is, in extremis a tautology: yes, LLMs are by definition just reproducing what they have been fed. But are you sure that isn't true about you as well? Do you have any concept of what you are doing? Because neuroscience has increasingly shown that we really really do not have any idea of what is going on in our reasoning process. But in particular, we don't even have to go there: from the moment our umbilical chord is snipped, we are autonomous little creatures, with a whole bunch of senses that are our sole connection to the surrounding world. We learn about that world, other people in it, how things work, etc. But there is a debate raging within philosophy, and there kinda has been since Descartes formulated the dualist proposition as "cogito ergo sum", that applies here. Dualists believe that in addition to your material self, there is a something (call it a soul, a spirit, a conscience, or whatever) that is separate from your material self and not an emergent property of that material self. And that is the thing that is doing the thinking. The debate is kind of over, as far as science is concerned, because no evidence at all has been found of any such thing, and we are understanding the power of emergent properties more and more, and there seems to be no reason why conscience should not be an emergent property of a sufficiently complex process. That said, the debate still rages on. The main modern argument is that this thing arises from quantum effects (and there is indeed evidence of quantum effects playing a role in the functioning of the brain). I personally remain unconvinced: even if quantum effects are a necessary requirement of consciousness, that doesn't make them non-material, it just means you need material in which such quantum effects can occur. However, lets remember this.
The question of duality is important to AI: if consciousness, intelligence, etc. are due to some immaterial component, then no matter what machine we build, unless we are able to imbue it with that thing, it will never be conscious. If, however, consciousness is an emergent property, then if we build a machine that is sufficiently complex in all the necessary ways, then it stands to reason, consciousness can emerge in that too. Of particular interest here is John Searle's Chinese Room Argument. This is also not a new argument, and this debate has raged since the 60s (not quite as long as dualism in human consciousness, but long enough for there to be a lot written about it). The main argument that is sticking so far is that if quantum effects are a necessary component of knowing stuff, then the system argument is removed: the room system, with a human inside it then either:
- knows Chinese, because the human inside it provides the necessary quantum effects to the entire system. But this is trivial, because we aren't putting humans inside our AIs, and aren't providing the necessary structure for such quantum effects to occur. Thus the Chinese room argument holds.
- or the room doesn't know Chinese, because the human isn't being employed as a thinking autonomous creature, but rather as a rote rule follower, and we aren't letting their fancy quantum brain do anything useful. In that case, trivially, the Chinese room argument holds, because the system does not know Chinese.
In other words, in both these cases, current AIs are on the wrong track to ever gaining consciousness.
However, I don't personally think this is very convincing: even if the necessary complexity in human brains comes from quantum effects, that just shows the quantum effects are a sufficient condition, but not yet a necessary one. There may be other ways for consciousness to emerge from material processes that do not require quantum effects. And if that's the case, we may just be on the right track to building a truly conscious machine, even if all it does is just "reproduce what it has been fed".
And with that horribly off-topic bit of epistemology, I suggest we return to discussing UK politics 
|
On May 19 2025 00:40 Acrofales wrote:Show nested quote +On May 18 2025 17:23 Gorsameth wrote:
A reminder that there is nothing intelligent about current gen ai. It has no concept of what it is doing. Its just reproducing what it has been fed. Show nested quote +On May 18 2025 23:54 GreenHorizons wrote:
I have plenty of problems with AI, but am I misremembering that AI made significant progress with protein folding, or were those just propaganda headlines? I'll reply to both of you at once. For starters, @GH: was not propaganda, it was a Nobel prize winning breakthrough in chemistry. It wasn't quite the same type of algorithm as what Gorsameth is talking about, though. But regardless, I don't think that's a counterpoint: nobody at all ever said that AI is not good at matching patterns, which is what protein folding is: it's very complex patterns of a kind us humans look at and think are just a jumble of unpredictability, but we (1) know there are deterministic rules underlying how proteins fold which are 100% determined by the order of their aminoacids, and (2) have lots of data about them. This is something ML algorithms are designed to deal with par excellence, and AlphaFold did. + Show Spoiler +Note, it was still an extremely hard problem: they had to come up with the right way to represent the data so that the algorithm could make sense of it, and just like they did in AlphaGo, they let it learn by example a bit, and then released it to improve upon itself. It was very impressive, but it just doesn't work the same way or do the same stuff as LLMs do. It's like if someone says "all FPS games are is just pointing your mouse and clicking to shoot", and countering with the argument "but Starcraft is pretty cool".
Now, regarding the original point, that is, in extremis a tautology: yes, LLMs are by definition just reproducing what they have been fed. But are you sure that isn't true about you as well? Do you have any concept of what you are doing? Because neuroscience has increasingly shown that we really really do not have any idea of what is going on in our reasoning process. But in particular, we don't even have to go there: from the moment our umbilical chord is snipped, we are autonomous little creatures, with a whole bunch of senses that are our sole connection to the surrounding world. We learn about that world, other people in it, how things work, etc. But there is a debate raging within philosophy, and there kinda has been since Descartes formulated the dualist proposition as "cogito ergo sum", that applies here. Dualists believe that in addition to your material self, there is a something (call it a soul, a spirit, a conscience, or whatever) that is separate from your material self and not an emergent property of that material self. And that is the thing that is doing the thinking. The debate is kind of over, as far as science is concerned, because no evidence at all has been found of any such thing, and we are understanding the power of emergent properties more and more, and there seems to be no reason why conscience should not be an emergent property of a sufficiently complex process. That said, the debate still rages on. The main modern argument is that this thing arises from quantum effects (and there is indeed evidence of quantum effects playing a role in the functioning of the brain). I personally remain unconvinced: even if quantum effects are a necessary requirement of consciousness, that doesn't make them non-material, it just means you need material in which such quantum effects can occur. However, lets remember this.
The question of duality is important to AI: if consciousness, intelligence, etc. are due to some immaterial component, then no matter what machine we build, unless we are able to imbue it with that thing, it will never be conscious. If, however, consciousness is an emergent property, then if we build a machine that is sufficiently complex in all the necessary ways, then it stands to reason, consciousness can emerge in that too. Of particular interest here is John Searle's Chinese Room Argument. This is also not a new argument, and this debate has raged since the 60s (not quite as long as dualism in human consciousness, but long enough for there to be a lot written about it). The main argument that is sticking so far is that if quantum effects are a necessary component of knowing stuff, then the system argument is removed: the room system, with a human inside it then either:
- knows Chinese, because the human inside it provides the necessary quantum effects to the entire system. But this is trivial, because we aren't putting humans inside our AIs, and aren't providing the necessary structure for such quantum effects to occur. Thus the Chinese room argument holds.
- or the room doesn't know Chinese, because the human isn't being employed as a thinking autonomous creature, but rather as a rote rule follower, and we aren't letting their fancy quantum brain do anything useful. In that case, trivially, the Chinese room argument holds, because the system does not know Chinese.
In other words, in both these cases, current AIs are on the wrong track to ever gaining consciousness.
However, I don't personally think this is very convincing: even if the necessary complexity in human brains comes from quantum effects, that just shows the quantum effects are a sufficient condition, but not yet a necessary one. There may be other ways for consciousness to emerge from material processes that do not require quantum effects. And if that's the case, we may just be on the right track to building a truly conscious machine, even if all it does is just "reproduce what it has been fed".
And with that horribly off-topic bit of epistemology, I suggest we return to discussing UK politics It was more of a curiosity than a counterpoint in hopes I would get a response like yours, so thank you for that.
That said, one of my thoughts on this dystopian surveillance nightmare society being built alongside incompetent (even if occasionally impressive) AI, in the context of your response, is that some people still literally believe in racial determinism and they have political power.
|
United States10501 Posts
On May 19 2025 00:12 Nebuchad wrote:Show nested quote +On May 18 2025 13:27 BlackJack wrote:
That goes to Rayzda’s point. Sure you can program google gemini to give you ridiculous responses. You can also program an AI to predict the Asian grandmother to be equally likely to nick your bike as anyone else. But if you’re going to program the AI to exist in a clown world then it’s not very useful to the real world. Presumably when the AI is involved there would already be a context, and then you could be discriminatory within that context. If some random dude is robbing a store and the AI says send two units, and then some random black dude is robbing a store and the AI says send four units, probability of violence increased, then that would be an example of a discriminatory AI. There's no scenario that matters in which the AI is to be asked if an old lady is more dangerous than a young man. We can sort of see that you understand the problem with your own argument because you should have a comparison between an old asian lady and an old black lady in order to test discrimination by race, but of course the old black lady isn't particularly likely to steal your bike either. Edit: and we could even extend that to have a reverse clown world example, for example an AI that says that an old black lady is more likely to steal a bike than a young white guy because black people steal more bikes.
You can change my example to an old black lady if you want. You’re still discriminating on the basis of sex and on the basis of age. I’m not sure why discriminating on those immutable characteristics instead of race is a counter to my argument.
The only reason you think this changes the equation at all is because you’re more tolerant of discrimination on the basis of sex/age in this matter. As you should be. Men commit overwhelmingly more crime than women. An AI pre-crime bot that doesn’t judge men differently from women would also be pretty useless.
I’m quite certain that if the AI bot predicts Muslims are likely to commit terrorist acts then there’s a decent chunk of people on the left that would be horrified of the Islamophobic AI bot. But if it predicts Christian fundamentalists are likely to bomb an abortion clinic it will be job well done.
|
On May 19 2025 01:37 BlackJack wrote:Show nested quote +On May 19 2025 00:12 Nebuchad wrote:On May 18 2025 13:27 BlackJack wrote:
That goes to Rayzda’s point. Sure you can program google gemini to give you ridiculous responses. You can also program an AI to predict the Asian grandmother to be equally likely to nick your bike as anyone else. But if you’re going to program the AI to exist in a clown world then it’s not very useful to the real world. Presumably when the AI is involved there would already be a context, and then you could be discriminatory within that context. If some random dude is robbing a store and the AI says send two units, and then some random black dude is robbing a store and the AI says send four units, probability of violence increased, then that would be an example of a discriminatory AI. There's no scenario that matters in which the AI is to be asked if an old lady is more dangerous than a young man. We can sort of see that you understand the problem with your own argument because you should have a comparison between an old asian lady and an old black lady in order to test discrimination by race, but of course the old black lady isn't particularly likely to steal your bike either. Edit: and we could even extend that to have a reverse clown world example, for example an AI that says that an old black lady is more likely to steal a bike than a young white guy because black people steal more bikes. You can change my example to an old black lady if you want. You’re still discriminating on the basis of sex and on the basis of age. I’m not sure why discriminating on those immutable characteristics instead of race is a counter to my argument. The only reason you think this changes the equation at all is because you’re more tolerant of discrimination on the basis of sex/age in this matter. As you should be. Men commit overwhelmingly more crime than women. An AI pre-crime bot that doesn’t judge men differently from women would also be pretty useless. I’m quite certain that if the AI bot predicts Muslims are likely to commit terrorist acts then there’s a decent chunk of people on the left that would be horrified of the Islamophobic AI bot. But if it predicts Christian fundamentalists are likely to bomb an abortion clinic it will be job well done.
But there's nothing happening in what you're saying, there are only "judgments" and "predictions". What is happening based on it?
|
|
|
|
|
|
|
|
EPT
