Many people are saying this (that my posts are good)duckzilla wrote:Ear-bot is at it again
European politics
- fightinfrenchman
- Ninja
- Posts: 23508
- Joined: Oct 17, 2015
- Location: Pennsylvania
Re: European politics
Dromedary Scone Mix is not Alone Mix
- harcha
- Gendarme
- Posts: 5141
- Joined: Jul 2, 2015
- ESO: hatamoto_samurai
Re: European politics
everyone is talking at the ear but is anyone actually using their s?
POC wrote:Also I most likely know a whole lot more than you.
POC wrote:Also as an objective third party, and near 100% accuracy of giving correct information, I would say my opinions are more reliable than yours.
Re: European politics
The impressive part is the way that it learned how to do that, not necessarily the feat itself (then again, it wouldn't have been possible otherwise).Dolan wrote:Well, what has the AI solved so far? Beating humans at Go and ...? Weak shit.fightinfrenchman wrote:Dolan wrote: And we live in the age of AI and other hipster "big tech" memes
Re: European politics
Yeah, it's impressive how millions of if statements and heavy use of memoization can beat a human at Go. But that's likely to happen for any problem that can be translated into maths.
I mean even multiplying 38404856 by 3453463 using a pocket calculator will beat a human too.
I mean even multiplying 38404856 by 3453463 using a pocket calculator will beat a human too.
Re: European politics
That's how our brains work too. It's neither embarrassing nor "weak shit" that they were able to replicate that
Re: European politics
Idk about that part of brains working in the same way. I mean, neurons aren't transistors or logical gates that only have two states.
Re: European politics
Your pocket calculator follows pre-defined instructions written by humans. The Go AI develop what can easily be called "intuition" about which moves are strongest without any human input, other than the rules of the game. This is why it's different from the old chess "AI" as well, which had complex decision trees pre-defined by humans and relied on brute-force move analysis to get an edge. Deepmind's AI was the first time computers defined their own instructions based on trial and error and memory, much like we do when we learn.Dolan wrote:I mean even multiplying 38404856 by 3453463 using a pocket calculator will beat a human too.
Re: European politics
Neural networks are called that for a reason; They are modeled after our brains. Anyway it doesn't matter if there's 2 states or 50. Anything can be rewritten to binary, it will just be less efficient. And yeah, current AI are certainly less efficient than our brains, but both can be rewritten as a bunch of if statements. Regardless, machine learning is obviously a big step in that field.Dolan wrote:Idk about that part of brains working in the same way. I mean, neurons aren't transistors or logical gates that only have two states.
Re: European politics
@Goodspeed
Yeah, but they defined their own moves based on instructions that were already baked into their initial state/program. It's like they had a victory condition specified in their initial algorithm and then even the function of storing previously tried moves, based on their outcome, was something programmed into them.
It's somewhat similar to that worm game, one of the easiest games you can write, in which the worm needs to only move in a 2D space that has no obstacles. You could generate new extra obstacles dynamically which will persist on the map, in addition to the ones that already exist. So the worm will have to "check" before making any new move if there's a new obstacle in front of it and store the result for that session. Such a worm game could be called AI, machine learning, etc.
Yeah, but they defined their own moves based on instructions that were already baked into their initial state/program. It's like they had a victory condition specified in their initial algorithm and then even the function of storing previously tried moves, based on their outcome, was something programmed into them.
It's somewhat similar to that worm game, one of the easiest games you can write, in which the worm needs to only move in a 2D space that has no obstacles. You could generate new extra obstacles dynamically which will persist on the map, in addition to the ones that already exist. So the worm will have to "check" before making any new move if there's a new obstacle in front of it and store the result for that session. Such a worm game could be called AI, machine learning, etc.
Re: European politics
The way we learn is equally baked into our systems. We, too, have a "baked in" ability to store previous trials and errors and use that memory to make better decisions in the future. The way our neurons interact is shaped by these trials and errors.
The difference is that the instructions to get to the end of the level are very easy to write yourself, whereas the "instructions" AlphaGo wrote for itself by playing against itself are not only impossible for a human to replicate, they even outperform the instructions embedded in human brains at a very complex task.It's somewhat similar to that worm game, one of the easiest games you can write, in which the worm needs to only move in a 2D space that has no obstacles. You could generate new extra obstacles dynamically which will persist on the map, in addition to the ones that already exist. So the worm will have to "check" before making any new move if there's a new obstacle in front of it and store the result for that session. Such a worm game could be called AI, machine learning, etc.
Re: European politics
Source for this?Goodspeed wrote:the "instructions" AlphaGo wrote for itself by playing against itself are not only impossible for a human to replicate, they even outperform the instructions embedded in human brains at a very complex task.
- fightinfrenchman
- Ninja
- Posts: 23508
- Joined: Oct 17, 2015
- Location: Pennsylvania
Re: European politics
Artificial intelligence is considered "cringe" to some peopleGoodspeed wrote:That's how our brains work too. It's neither embarrassing nor "weak shit" that they were able to replicate that
Dromedary Scone Mix is not Alone Mix
Re: European politics
Internet is cringefightinfrenchman wrote:Artificial intelligence is considered "cringe" to some peopleGoodspeed wrote:That's how our brains work too. It's neither embarrassing nor "weak shit" that they were able to replicate that
Re: European politics
AI is cringe, coz it's neither intelligent nor can it produce anything artificial of its own volition.
It's just another fancy name for software.
It's just another fancy name for software.
Re: European politics
They outperform humans at Go, and no human could create a decision tree that comes anywhere close to that level of performance. The only way to make that decision tree is indirectly: By having the program build its own decision tree.Dolan wrote:Source for this?Goodspeed wrote:the "instructions" AlphaGo wrote for itself by playing against itself are not only impossible for a human to replicate, they even outperform the instructions embedded in human brains at a very complex task.
Re: European politics
That's because human brains aren't adapted for massive parallel processing, like electronics are. That's why what AI/software does does not compare well to human cognition.Goodspeed wrote:They outperform humans at Go, and no human could create a decision tree that comes anywhere close to that level of performance. The only way to make that decision tree is indirectly: By having the program build its own decision tree.
Re: European politics
Yeah that's a great way to make up for the lack of efficiency. Whereas AlphaGo needs a million self-play games to get to beginner-intermediate level, we get there with a few hundred.
But the method of learning is still similar, and the fact that it's able to outperform a human brain at a very complex task, even if it's just specifically that task and even if it needs much more practice, is still a big deal. It's no general purpose AI but it's a big step in the field. Denying that just shows your lack of knowledge about said field.
I get the feeling you'd call any invention in this field "weak shit" until it's the end-all general purpose AI that outperforms us at everything. That's setting the bar a little high. The fact that you're even setting it that high shows how much progress was made in the field recently. Or that you've been watching too many movies I guess.
But the method of learning is still similar, and the fact that it's able to outperform a human brain at a very complex task, even if it's just specifically that task and even if it needs much more practice, is still a big deal. It's no general purpose AI but it's a big step in the field. Denying that just shows your lack of knowledge about said field.
I get the feeling you'd call any invention in this field "weak shit" until it's the end-all general purpose AI that outperforms us at everything. That's setting the bar a little high. The fact that you're even setting it that high shows how much progress was made in the field recently. Or that you've been watching too many movies I guess.
Re: European politics
And neither can they ever become a general purpose AI, since that requires having embodied cognition. They can only get good at tasks that lend themselves well to being automated.
So basically very specialised tasks that require no actual intelligence.
So basically very specialised tasks that require no actual intelligence.
Re: European politics
I think it's very possible there are limitations when it comes to resource efficiency and performance/speed, but I see no reason to believe our ability to learn any task cannot be replicated by computers.Dolan wrote:And neither can they ever become a general purpose AI, since that requires having embodied cognition. They can only get good at tasks that lend themselves well to being automated.
So basically very specialised tasks that require no actual intelligence.
- princeofcarthage
- Retired Contributor
- Posts: 8861
- Joined: Aug 28, 2015
- Location: Milky Way!
Re: European politics
Because computers don't actually learn. The intake data and process according to given parameters or instructions. Can AlphaGo for example learn to walk of given legs? No. It needs to be programmed to do so. And by programming you are just automating the task you otherwise would manually need to do. Yes the field has made lot of progress but it is still nowhere near to even 1% of true intelligence.Goodspeed wrote:I think it's very possible there are limitations when it comes to resource efficiency and performance/speed, but I see no reason to believe our ability to learn any task cannot be replicated by computers.Dolan wrote:And neither can they ever become a general purpose AI, since that requires having embodied cognition. They can only get good at tasks that lend themselves well to being automated.
So basically very specialised tasks that require no actual intelligence.
Fine line to something great is a strange change.
- scarm
- Howdah
- Posts: 1439
- Joined: Dec 7, 2018
- ESO: Malebranche
Re: European politics
Are Go and Chess actually "very complex tasks" though? They are mathematically solvable games in theory, humans just lack the computing power to do so.
Re: European politics
Yeah, show me an AI that one day decided on its own it wants to learn about why balloon sticks legprinceofcarthage wrote:Because computers don't actually learn. The intake data and process according to given parameters or instructions. Can AlphaGo for example learn to walk of given legs? No. It needs to be programmed to do so. And by programming you are just automating the task you otherwise would manually need to do. Yes the field has made lot of progress but it is still nowhere near to even 1% of true intelligence.
Re: European politics
This is not different from humans. We are programmed to mimic the behavior of our peers, which you could obviously program computers to do as well, in theory. I get that the field isn't there yet but the claim that it could never get there is shortsighted imo. Humans are not that special.
Re: European politics
Everything is mathematically solvable in theory. Go is plenty complex as a proof of concept for machine learning.scarm wrote:Are Go and Chess actually "very complex tasks" though? They are mathematically solvable games in theory, humans just lack the computing power to do so.
Who is online
Users browsing this forum: No registered users and 10 guests