|
Should I step down as head of twitter This poll is closed. |
|||
|---|---|---|---|
| Yes |
|
420 | 4.43% |
| No |
|
69 | 0.73% |
| Goku |
|
9001 | 94.85% |
| Total: | 9490 votes | ||
|
shyduck posted:https://twitter.com/drmistercody/status/1727503522416271835?s=20 Big "I'll take up smoking and give that up" energy
|
#
?
Nov 23, 2023 04:06
|
|
|
|
| # ? May 30, 2024 19:32 |
|
|
burger
|
|
#
?
Nov 23, 2023 04:07
|
|
|
The Saddest Rhino posted:burger hot dog
|
|
#
?
Nov 23, 2023 04:07
|
|
|
|
|
#
?
Nov 23, 2023 04:08
|
|
|
To me there is a really obvious reason why Ai companies are clearly full of poo poo, they are selling the technology. Tesla self-driving robotaxis, ChatGPT, nVidia Ai accelerators, everything is for sale. It doesn't make any sense, if this technology was as amazing as you say it is, you would keep it to yourself. Think about how much money you could make on relatively simple problems if this technology was so amazing, super accurate weather predict, stock performance, drug discovery. They don't even have the technology to serve us good ads. Shovel sales in a gold rush. Three Olives fucked around with this message at 04:17 on Nov 23, 2023 |
|
#
?
Nov 23, 2023 04:08
|
|
|
much aesthetically please m'lusk. tip of the fedora
|
|
#
?
Nov 23, 2023 04:12
|
|
|
Also AI is impossible and stupid at even the slightest amount of thinking about it
|
|
#
?
Nov 23, 2023 04:12
|
|
|
euphronius posted:Also AI is impossible and stupid at even the slightest amount of thinking about it Sounds like someone hasn’t seen the documentary “Short Circuit”. I think you’ll find Johnny 5 was quite alive
|
|
#
?
Nov 23, 2023 04:13
|
|
|
euphronius posted:Also AI is impossible and stupid at even the slightest amount of thinking about it  Getting computers to not be stupid pieces of poo poo is cool and good, but a piece of code comparing bit values between two jpegs isn't going to hunt down humanity with Hunter-Killer robots
|
|
#
?
Nov 23, 2023 04:16
|
|
|
shyduck posted:https://twitter.com/drmistercody/status/1727503522416271835?s=20 Cheong was cool enough to have the redemption arc of a clown liker into clown incarnate
|
|
#
?
Nov 23, 2023 04:20
|
|
|
truly innovative
|
|
#
?
Nov 23, 2023 04:21
|
|
|
euphronius posted:Also AI is impossible and stupid at even the slightest amount of thinking about it colleague Henry Farrell wrote an excellent essay on how looking at the current AI movement in religious/eschatological terms makes far more sense than viewing it from a scientific standpoint, since we are dealing with people who are talking about immanentizing the eschaton rather than just making a new computer program mannerup fucked around with this message at 04:30 on Nov 23, 2023 |
|
#
?
Nov 23, 2023 04:28
|
|
|
if you can dodge a wrench you can extinguish humanity
|
|
#
?
Nov 23, 2023 04:36
|
|
|
mannerup posted:the problem with this is when most of those people talk about 'AI', they are just talking about 'programs' and 'algorithms' which can be very, very sophisticated but are merely just a series of heuristics. AGI (Artificial General Intelligence) is what most people think of when they hear the term 'AI', and we are as close to solving that as we are to being able to solve the hard problem of consciousness (not loving close at all) It’s all so incredibly stupid Maybe more stupid than crypto
|
|
#
?
Nov 23, 2023 04:40
|
|
|
euphronius posted:It’s all so incredibly stupid
|
|
#
?
Nov 23, 2023 04:51
|
|
|
"AI" in its current form is probably just going to cut a lot of labor out of things like graphic design and data entry/transcription. Nothing is going to fundamentally change, it's a new excuse to pay people less. If we get lucky a bunch of middle managers whose main contribution is regurgitating information through powerpoint are collateral damage.
|
|
#
?
Nov 23, 2023 04:54
|
|
|
mannerup posted:AI bros would've been better off spending their time reading at least Kant, Hume, Spinoza and Wittgenstein instead of thinking the humanities were a waste of time if they wanted to take a serious approach to the problems they are attempting to solve Their whole goal is to get an AI to read them and summarize it for them. Reading and thinking is a lot harder then just letting the computer do it while you oppress the masses and jerk off to furry porn
|
|
#
?
Nov 23, 2023 04:57
|
|
|
mannerup posted:the problem with this is when most of those people talk about 'AI', they are just talking about 'programs' and 'algorithms' which can be very, very sophisticated but are merely just a series of heuristics. AGI (Artificial General Intelligence) is what most people think of when they hear the term 'AI', and we are as close to solving that as we are to being able to solve the hard problem of consciousness (not loving close at all) I'm sure I have cobbled this opinion from many, much smarter people, but as I understand it basically everything we have done on self-driving technology has been a complete waste of time if our goal is completely autonomous self-driving. Basically, we could never come up with enough rules to cover everything and that it would actually require something closer to consciousness to accomplish because things that seem simple actually require a real fundamental understanding of how the world and other humans work than "rules". Like, it's not enough to recognize fire, you need a really fundamental understanding of what fire is and how it interacts with other things/people. Something that came up with Cruise is their cars couldn't understand gestures from emergency responders and gestures are actually super difficult because you have to understand the intent of the human making them on a really fundamental way, some random crazy on the street holding their palm out could mean nothing or it could be someone warning you of serious danger and that is really difficult, if not impossible for AI to understand that without understanding that humans have very different intents that we regularly have determine in really important ways.
|
|
#
?
Nov 23, 2023 04:58
|
|
|
Three Olives posted:Basically, we could never come up with enough rules to cover everything and that it would actually require something closer to consciousness to accomplish because things that seem simple actually require a real fundamental understanding of how the world and other humans work than "rules". Like, it's not enough to recognize fire, you need a really fundamental understanding of what fire is and how it interacts with other things/people. quote:Something that came up with Cruise is their cars couldn't understand gestures from emergency responders and gestures are actually super difficult because you have to understand the intent of the human making them on a really fundamental way, some random crazy on the street holding their palm out could mean nothing or it could be someone warning you of serious danger and that is really difficult, if not impossible for AI to understand that without understanding that humans have very different intents that we regularly have determine in really important ways. 'AI' shares a similar issue with mankind in reporting data: it doesn't know what it doesn't know, but will make up bullshit to cover the gaps
|
|
#
?
Nov 23, 2023 05:16
|
|
|
Calling it now the AI found a proof for P=NP and now there's no such things as encryption that works.
|
|
#
?
Nov 23, 2023 05:28
|
|
|
Calling it now they convinced some Saudi prince that chatgpt can predict the stock market and the hitmen are pulling into their driveways as we speak
|
|
#
?
Nov 23, 2023 05:32
|
|
|
mannerup posted:algorithms can only do tasks on what they have learned from their dataset and make interferences from that dataset to situations outside of their training data to provide heuristics; they can't extrapolate beyond that with any true accuracy (this is generally when they start to break down and provide you garbage data back to hilarious results) It's all about shapes. It's designed to interpret everything as vectors if you ask it a 'question,' it turns that into a shape, and then it responds with another shape. So you get things shaped like 'answers' but as you say, no understanding or comprehension.
|
|
#
?
Nov 23, 2023 05:32
|
|
|
The AI figured out all 11 secret herbs and spices in KFC and now the tech bros are afraid it'll figure out their browser history next.
|
|
#
?
Nov 23, 2023 05:33
|
|
|
but i was using incognito mode!!
|
|
#
?
Nov 23, 2023 05:35
|
|
|
ashpanash posted:It's all about shapes. It's designed to interpret everything as vectors if you ask it a 'question,' it turns that into a shape, and then it responds with another shape. So you get things shaped like 'answers' but as you say, no understanding or comprehension.
|
|
#
?
Nov 23, 2023 05:36
|
|
|
ashpanash posted:It's all about shapes. It's designed to interpret everything as vectors if you ask it a 'question,' it turns that into a shape, and then it responds with another shape. So you get things shaped like 'answers' but as you say, no understanding or comprehension. tbf that what I do
|
|
#
?
Nov 23, 2023 05:39
|
|
|
mannerup posted:using your example, this is a fair assessment. you can provide a program a good heuristic of what 'fire' is like including its interactions with other objects/situations and how to react in those situations, but it will never fundamentally understand the concept of 'fire' I guess the example for me is, if you are going under an overpass and see a car on fire on top, it's probably safe to pass underneath, if you see overpass with a Exxon fuel truck on it and there is a liquid pouring down onto the road there are a lot of assumptions that a human can make about the safety of the situation that would be really difficult for AI, you have to understand what a fuel truck is, you have to understand why one would be labeled Exxon, make a leap to that it is probably carrying a very flammable chemical, that the tank is probably punctured if there is a large amount of fluid draining onto the road, that the fluid is likely from the truck because overpasses don't usually have large amounts of fluids coming off of them when it is not raining and that fluid is likely now covering the area you are intending on passing over and is highly likely to be suddenly engulfed in flames that could kill you. That is a lot of loving stuff to infer that is easy for a human and incredibly difficult for AI.
|
|
#
?
Nov 23, 2023 05:45
|
|
|
Calling it now the AI determined the proper way to measure ur dick and Elon is super mad about the result
|
|
#
?
Nov 23, 2023 05:47
|
|
|
ashpanash posted:It's all about shapes. It's designed to interpret everything as vectors if you ask it a 'question,' it turns that into a shape, and then it responds with another shape. So you get things shaped like 'answers' but as you say, no understanding or comprehension. Its just a Chinese Room AI  reads the questions and knows what to respond, but it doesn't actually read anything. Its looking at the hexadecimal number values of the input query and calculating the best solution. No idea if its math is right or wrong, it just punches in the number into an impossibly complex calculator and spits out the sum/bullshit LinkedIn post
|
|
#
?
Nov 23, 2023 05:48
|
|
|
Three Olives posted:Something that came up with Cruise is their cars couldn't understand gestures from emergency responders and gestures are actually super difficult because you have to understand the intent of the human making them on a really fundamental way, some random crazy on the street holding their palm out could mean nothing or it could be someone warning you of serious danger and that is really difficult, if not impossible for AI to understand that without understanding that humans have very different intents that we regularly have determine in really important ways. And lots of drivers currently driving now can't, or just won't be bothered to, understand gestures from first responders, or even simpler that they should get out of their way if they are in their way. Like I'm not saying we're even close to being there with self driving cars, and I get there is just a whole bunch of real world stuff that using a simple rules based approach would be almost impossible to design for, but the bar is set pretty loving low.
|
|
#
?
Nov 23, 2023 05:50
|
|
|
dr_rat posted:And lots of drivers currently driving now can't, or just won't be bothered to, understand gestures from first responders, or even simpler that they should get out of their way if they are in their way. actually self-driving robo taxis are going to change the world any day now
|
|
#
?
Nov 23, 2023 05:55
|
|
|
I know I shouldn't wade into GBS philosophy but if you even believe it's possible to build an artificial material intelligence without some form of dualism then it's going to be expressible as an algorithm and some form of mathematics. Neural networks aren't AGI but their ability to create heuristic solutions blow previous attempts out of the water, and that's basically all we've got for trying to mathematically define 'understanding.' I know all the annoying techbros are all over it but it is an interesting technology.
|
|
#
?
Nov 23, 2023 05:55
|
|
|
Nice Van My Man posted:I know I shouldn't wade into GBS philosophy but if you even believe it's possible to build an artificial material intelligence without some form of dualism then it's going to be expressible as an algorithm and some form of mathematics. Neural networks aren't AGI but their ability to create heuristic solutions blow previous attempts out of the water, and that's basically all we've got for trying to mathematically define 'understanding.' I know all the annoying techbros are all over it but it is an interesting technology. Concerning
|
|
#
?
Nov 23, 2023 05:58
|
|
|
Nice Van My Man posted:I know I shouldn't wade into GBS philosophy but if you even believe it's possible to build an artificial material intelligence without some form of dualism then it's going to be expressible as an algorithm and some form of mathematics. Neural networks aren't AGI but their ability to create heuristic solutions blow previous attempts out of the water, and that's basically all we've got for trying to mathematically define 'understanding.' I know all the annoying techbros are all over it but it is an interesting technology. Exactly
|
|
#
?
Nov 23, 2023 06:03
|
|
|
ben shapino posted:actually self-driving robo taxis are going to change the world any day now We probably will get self-driving robo taxi's in places with lovely regulations getting rolled out on mass well before the AI is up to it. But more worrying is it happening with transport, as transport companies would love to be able to get rid of drivers, and yeah those big lorries going around without someone behind the wheel is a bit of a scary thought. I believe some mining companies already have automatic trucks going on their own properties (which being mining companies is usually a fair bit of land).
|
|
#
?
Nov 23, 2023 06:03
|
|
|
Nice Van My Man posted:Neural networks aren't AGI but their ability to create heuristic solutions blow previous attempts out of the water, and that's basically all we've got for trying to mathematically define 'understanding.' I know all the annoying techbros are all over it but it is an interesting technology. I think people who underestimate the potential of these algorithms/heuristics should see how it has done in games like Poker, long through to be a domain where humans would reign supreme because of our unpredictability even if AGI is a pipe dream, specialized AI in specific domains can do a hell of a lot in the mean time on problems that aren't the hard problem of consciousness
|
|
#
?
Nov 23, 2023 06:04
|
|
|
euphronius posted:Also AI is impossible and stupid at even the slightest amount of thinking about it Someone isn't being very ftw
|
|
#
?
Nov 23, 2023 06:07
|
|
|
mannerup posted:this is a fair point, and there are absolutely certain applications that blow humans out of the water (I honestly think operating motor vehicles is one of these areas, as much as people like to think that is a unique domain to humans or we are somehow superior to a machine with multiple sensors to detect collisions) I have the newest Ford system and it is really amazing because I don't have 5 radar sensors covering all directions around me. Something I have thought about though is it can make things more complicated overall, there is no question I could drive drunk tons better with these systems, lane keeping works really well, you might not even be able to tell I was drunk, I still might plow into something my car couldn't recognize that I would see if I was sober. dr_rat posted:We probably will get self-driving robo taxi's in places with lovely regulations getting rolled out on mass well before the AI is up to it. Realistically we will probably see something like bus lanes everywhere for self-driving cars where they are considered exempt from capabilities a human is capable of, making the whole system worse for everyone but you get to be driven to work in a $80k Audi instead of a bus like a filthy poor.
|
|
#
?
Nov 23, 2023 06:22
|
|
|
mannerup posted:lmao that is so, so much worse than what media matters found lol
|
|
#
?
Nov 23, 2023 06:31
|
|
|
|
| # ? May 30, 2024 19:32 |
|
|
mannerup posted:this is a fair point, and there are absolutely certain applications that blow humans out of the water (I honestly think operating motor vehicles is one of these areas, as much as people like to think that is a unique domain to humans or we are somehow superior to a machine with multiple sensors to detect collisions) The problem isn't operating a motor vehicle in and of itself (which any well-trained computer algorithm should be able to exceed the skill of most humans in ideal conditions), but rather in the environment in which we use motor vehicles, which is an absurd mish-mash of non-deterministic, somewhat ill-defined and occasionally ad-hoc systems. Humans are comfortable with absurdity - you might even say we thrive on it. Computers don't do absurdity well.
|
|
#
?
Nov 23, 2023 06:32
|
|