Radiation House - Vol. 7 Ch. 52 - Memories in the Mist (2)

[MangaDex]

 

[damakuno]

While no doubt the narrow AI described here has the potential to be useful, in real life AGI remains in the realm of fiction. And even narrow AI has its problems and can't be reliably used to replace a diagnosis from a qualified radiologist . See linked paper for more discussion on this topic by people way more qualified than me, as my expertise is only in data engineering and computer vision, I have no experience whatsoever in medicine. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6038229/

 

[Batias]

I hate this AI design. It's the worst.

 

[sexysolaire]

While no doubt the narrow AI described here has the potential to be useful, in real life AGI remains in the realm of fiction. And even narrow AI has its problems and can't be reliably used to replace a diagnosis from a qualified radiologist . See linked paper for more discussion on this topic by people way more qualified than me, as my expertise is only in data engineering and computer vision, I have no experience whatsoever in medicine. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6038229/

yeah its wierd how they had what so far seems like really accurate for the most part medical stuff and must've done a ton of research on that.

then they wrote up this weird scifi bullshit with next to no basis in reality. like wtf singularity soon?! fuck no thats at least a few hundred years away unless we suddenly go through something like a second industrial revolution soon.

 

[osmium1010]

People in general don't understand what AI is at all and so it leads to BS like this. Great.

 

[JasonBourne117]

That's not what "AI" (machine learning systems" are or how they work but oh well... plot I guess.

 

[rudezki]

take your AI where the sun don't shine

 

[CowHudson]

yeah its wierd how they had what so far seems like really accurate for the most part medical stuff and must've done a ton of research on that.

then they wrote up this weird scifi bullshit with next to no basis in reality. like wtf singularity soon?! fuck no thats at least a few hundred years away unless we suddenly go through something like a second industrial revolution soon.

Actually it’s believed that the singularity MIGHT happen soon (the big names believe around 2030). For reference Google: Ai 2027. Most arguments against the report is the time frame. The report originally predicted later 2027 but one of the authors said they now believe around 2030 and most experts who don’t fully agree with the report disagree on the time frame

 

[sexysolaire]

Actually it’s believed that the singularity MIGHT happen soon (the big names believe around 2030). For reference Google: Ai 2027. Most arguments against the report is the time frame. The report originally predicted later 2027 but one of the authors said they now believe around 2030 and most experts who don’t fully agree with the report disagree on the time frame

i doubt the singularity is gonna happen when ai cant even get the right number of fingers down.

also what big names and when did they say it? you dont mean the rich guys who run/have a stake in the companies that profit from people thinking ai some kinda magic compared to what it actually is?

cause as someone who knows alot about tech in general current ai isnt really that impressing compared to other tech we've had in general for a while. its just overhyped by lazy idiots hoping to get rich quick like the internet did. of course they forget the huge bubble burst that still happened. the only thing really interesting, new or impressive about ai is how the learning models can funnel such a huge amount of stuff together to do its stuff. however realistically we are still about a good 60 full years of hard work if im being a big optimist to make something like a more 5 times more advanced version of siri.

also thats very very very optimistic view. theres a limit to how fast progress can be made before it becomes mostly diminishing returns. for example moores law which was basically a law made up by a guy about how fast circuit boards speed increased each year in 65. at the time it was doubling every year and he made a pretty accurate prediction that it would double every other year which was pretty accurate for a while. now though? dead we can only improve so fast with current tech without some kinda breakthrough somewhere else.

and funny enough it seems we are already seeing that with ai starting to plateau. what are they trying to do to advance it now? they're just trying to brute force it of course by buying insane amounts of parts to try to make it more powerful with the power of money ofc. after all they gotta make sure their golden bubble doesnt burst all over their face and wallets

but man im talking about us getting to the point that ai isn't a stupid gimmick they throw into everything. to a point where its actually useful in daily life. but self upgrading tech in 2030? sentient fucking self upgrading robots the motherfucking singularity in 2030? fucking lololololololoololol

btw can you ask the all the early 1930s-2026 scifi novels where the fuck my jetsons/back to the future style flying car is already? but anyway yeah sure singularity soon lol. by that timeline id better go get my shit done or the sun might explode in 2031!

 

[Sonny6155]

Well, I guess they sort of had something with the 3 outlined issues, and I suppose it's good that the author briefly covered how a broader model backing a specialist model can help address some key limitations of a naive narrow AI approach.

It's just that... I can hardly call the "so we are gonna run it through this super deep neural net and feed it a quadrillion-jillion images" solution a satisfying nor nuanced one. And character writing-wise, I'm pretty sure handing your student a GPT-7-powered tamagotchi on your debut chapter doesn't help your "leading expert" image, so much as it frames you as the next big Silicon Valley grifter on TikTok...

 

[CowHudson]

i doubt the singularity is gonna happen when ai cant even get the right number of fingers down.

also what big names and when did they say it? you dont mean the rich guys who run/have a stake in the companies that profit from people thinking ai some kinda magic compared to what it actually is?

cause as someone who knows alot about tech in general current ai isnt really that impressing compared to other tech we've had in general for a while. its just overhyped by lazy idiots hoping to get rich quick like the internet did. of course they forget the huge bubble burst that still happened. the only thing really interesting, new or impressive about ai is how the learning models can funnel such a huge amount of stuff together to do its stuff. however realistically we are still about a good 60 full years of hard work if im being a big optimist to make something like a more 5 times more advanced version of siri.

also thats very very very optimistic view. theres a limit to how fast progress can be made before it becomes mostly diminishing returns. for example moores law which was basically a law made up by a guy about how fast circuit boards speed increased each year in 65. at the time it was doubling every year and he made a pretty accurate prediction that it would double every other year which was pretty accurate for a while. now though? dead we can only improve so fast with current tech without some kinda breakthrough somewhere else.

and funny enough it seems we are already seeing that with ai starting to plateau. what are they trying to do to advance it now? they're just trying to brute force it of course by buying insane amounts of parts to try to make it more powerful with the power of money ofc. after all they gotta make sure their golden bubble doesnt burst all over their face and wallets

but man im talking about us getting to the point that ai isn't a stupid gimmick they throw into everything. to a point where its actually useful in daily life. but self upgrading tech in 2030? sentient fucking self upgrading robots the motherfucking singularity in 2030? fucking lololololololoololol

btw can you ask the all the early 1930s-2026 scifi novels where the fuck my jetsons/back to the future style flying car is already? but anyway yeah sure singularity soon lol. by that timeline id better go get my shit done or the sun might explode in 2031!

Hey all I’m saying is some experts believe it MIGHT. AI 2027 is a report made by former open ai employees who resigned from the company and claimed that open ai was being reckless in its pursuit of AI. Sam Altman himself said they have an internal goal at open ai to have an “automated AI research intern by September of 2026 running on hundreds of thousands of GPUs, and a true automated AI researcher by March of 2028.” Now of course this is all speculation and as you said in your example these predictions could be completely wrong.

However, in your original comment you were saying it’s absolute BS and sci-fi and that it shouldn’t belong. All I’m trying to say is people actively in the AI industry are claiming there’s a potential for it. So it’s not really sci-fi bs and doesn’t really go against the usual facts they’ve shown in this manga especially since they are also only claiming that it will happen soon. Plus they aren’t using anything sci-fi they’re just talking about it.