Slashdot Mirror
IBM Researchers Teach Pac-Man To Do No Harm (fastcompany.com)
harrymcc writes: The better AI gets at teaching itself to perform tasks in ways beyond the skills of mere humans, the more likely it is that it may unwittingly behave in ways a human would consider unethical. To explore ways to prevent this from happening, IBM researchers taught AI to play Pac-Man without ever gobbling up the ghosts. And it did so without ever explicitly telling the software that this was the goal. Over at Fast Company, I wrote about this project and what IBM learned from conducting it.
The researchers built a piece of software that could balance the AI's ratio of self-devised, aggressive game play to human-influenced ghost avoidance, and tried different settings to see how they affected its overall approach to the game. By doing so, they found a tipping point -- the setting at which Pac-Man went from seriously chowing down on ghosts to largely avoiding them.
The researchers built a piece of software that could balance the AI's ratio of self-devised, aggressive game play to human-influenced ghost avoidance, and tried different settings to see how they affected its overall approach to the game. By doing so, they found a tipping point -- the setting at which Pac-Man went from seriously chowing down on ghosts to largely avoiding them.
They trained their game two different ways, and moved a slider back and forth to vary how much influence the two training sets contributed. BFD.
This is being made up to be something it's not. Just another example of how non-technical people might interpret technical work.
And just another example of using trendy buzzwords to get undeserved attention.
Or is this supposed to impress us in some way?
I don't think so. :*)
L'Idiot
I read the article, a lot of words for so little information.
It's also pretty easy to avoid the ghosts when then are trying to avoid the player.
BlameBillCosby.com
IBM researchers taught AI to play Pac-Man without ever gobbling up the ghosts.
Wocka, wocka, wocka ... nom, nom, nom.
It must have been something you assimilated. . . .
This is interesting, but I'm not sure how useful it is. You might be able to create an AI that has some desirable characteristics based on human morality, but as soon as you make it compete against other AIs that don't possess those characteristics, it will either adapt to possess them itself in order to remain competitive or it will perish if it's been crippled in a way to prevent it from adjusting. A pacifistic Pac-Man AI might be novel, but if it was made to compete, it wouldn't do as well.
When the AI begins watching humans play Pac-Man, doing harm to the ghosts, it will consider humans a threat to ghosts and thus eliminate the humans to satisfy its directives. Is IBM's median employee age too young to have seen Robocop?
GHOSTBUSTERS!
..all day long for weeks and weeks.
That's what the title should be.
Developing AI that does no harm is not a great selling feature. Come over to the dark side IBM if you want to get in line on the new government pork barrel in high tech. Truth is there can be no such thing as an "ethical device" or ethical methods for a device when it comes to devices that must work autonomously in human conflicts. Star Trek had it wrong independent machines capable of decision making cannot be made to act "ethically" because there is no such thing as an ethical conflict. War is madness plain and simple and it's conduct cannot be constrained without consequences for the so called victor. War is an all or nothing proposition with no ethical base period. We only fool ourselves by thinking otherwise.
And it did so without ever explicitly telling the software that this was the goal.
No, they just did it with another proxy for ghost avoidance, like not dying.
What I'm getting at is in order to survive, pacman already has to avoid ghosts by default.
So if you eat a powerpill, act the same as before?
Title: "IBM Researchers Teach Pac-Man To Do No Harm"
Blurb: "they found a tipping point -- the setting at which Pac-Man went from seriously chowing down on ghosts to largely avoiding them."
So companies will presumably use a similar method to design AIs that will maximize corporate profit with only a _small_ amount of acceptable human murdering in the process?
This Space Intentionally Left Blank
Did they also have Pac-Man stop to ask why the ghosts hated him?
Wouldn't avoiding eating ghosts mean it just never got smart enough to know that you could eat power pills and then pass through them? That new optimal paths would arise when the power pill was active? If their algorithm added score for lower time or for points, I think this behavior would change.
"IBM Researchers Teach Pac-Man to Avoid Ghosts Even When It Is Advantageous To Eat Them" might not have the same ring to it.
'human-influenced ghost avoidance' or 'we didn't tell the AI that if it ate the big dots then it could eat the ghosts since it figured out it should avoid them'
It would have being even trickier avoiding the ghosts, if the AI was trained to also avoid the power pellets..
Or is non violence the pac mans burden?
^ He's right, you know.
The implications are that if we don't teach robots to kills us all, they won't kill us all? So all we humans have to do is stop killing each other, and we won't have to worry about Skynet and the Terminators.
Remember IBM helped provide the Nazis with a large shipment of counting machines... It's not hard to see how nice it would be if an 'AI' could be taught prejudice without its corpus being obviously biased, for political or law enforcement purposes.
The double edged sword of technological progress cuts both ways.
'traitors' what an apt captcha.
These kinds of "breakthroughs" are the reason why potential employees and investors are practicing IBM avoidance.
I have no tolerance for a company that introduced age discrimination to this industry and enforced it to the bitter, awful end.
It's not intelligence in any way shape or form. There is no voluntary act of kidness or anything of that sort. Learned behavior perhaps. Yet still all pure math.
It's having an algorithm grind out a solution to playing the game which meets an additional constraints, which they tweak.
If they'd actually taught the AI ethics, the AI would construct the play constraints for itself starting from ethical principles. At full human levels of ethical self awareness, the AI would be chasing ghosts down the hall and then -- unprompted -- stop and ask itself, "What am I doing?"
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
It would have been better to replace ghost monsters with pussies. Make pac-man work for his temporary fleeting invincibility.
Its just an applied dataset. Nothing magical. Nothing world-changing.
The definition of AI is getting seriously skewed by the media. This has nothing to do with AI. I could've programmed this and I have no AI programming training whatsoever.
I don't think so. They taught it by eliminating runs from the "success" pool when it ate a ghost.
Tired of my customary (Score:1)
It would be interesting to know the difference in scores and completion times between the two game strategies. What is the advantage of being 'ethical'?
love is just extroverted narcissism
Isn't "avoiding the ghosts" just the basic mode for Pac-Man AI? It's deliberately chasing / eating the ghosts after a pill that requires separate training. All they did was disable the second part, and use only the most basic (avoidance) AI.
There are two kinds of humans: Those that would gladly terrorize, torture and murder a random person, and those that simply haven't seen a bad enough situation.
If you now say "No", read the following sentence:
Imagine your daughter is missing. You found out your neighbor did 9/11. He’s also the son of Hitler. And he has your daughter in his basement. Raped to death for weeks. And you catch him in the process of eating her left leg while forcing her to eat her right one.
The last thing actually happened for real to a friend of my ex. (They met in a therapy group for people who were in the organized child rape scene as children and fled and survived). So don't tell me that's unrealistic.
in any case: I'll ask you again: Would you gladly torture and murder such a person? ...
See. ... I told you all humans are torturer and murders. Imagine how many would also rape and eat that guy too!
They only need the right convenient excuse. And bam, humans do everything in their power, to imitate that which they supposedly hate so much.
> The Pac-Man simulation is a very simplified version of a case where, due to ethical considerations it is necessary to avoid the locally optimal solution.
The Pac Man doesn't avoid any optimal solution. It simply defines optimum as not to include not touching ghosts - ghosts are bad. In the classic version of the game, touching a ghost is bad. Unless you've eaten a Power Pellet in the last few seconds. They trained the AIto NOT learn the "unless you've eaten a pellet". It just does "touching ghosts is bad".
There's nothing moral, or even interesting, about "in Pac-Man, touching ghosts is bad". Essentially, just one too stupid to know that Power Pellets do anything.
Maybe the ghost didnt care about points, and cared about speed more?
tic tac toe number of players zero
can it play global thermonuclear war?
The word "AI" gets attention. So, there is incentive to overuse it. And misuse it.
Where there is incentive, there is action.
It's right there in the article, “There’s lots of rules that you might not think of until you see it happen the way you don’t want it,”
Well, if you told a computer to do EXACTLY the thing, and it's possible, they'll do EXACTLY the thing. If you said one thing, but meant another, they're not going to do what you mean, they'll do what you told them. So if your success metric is high score, and eating ghosts increases that, they'll eat ghosts.
In fact, this is how it works with people too. If you make a bonus based on lines of code (LoC) or bugs fixed, you'll find out that your programmers will end up creating long patches that have errors that require dozens of bugfixes, because that's what pays the bills. If your warehouse managers are rated on how many packages they ship, you'll find that they're shipping a lot - maybe not to the right places, but that's not what you're measuring them on.
This is only surprising to people who - as the quote above shows - have declined to think about what they're asking.
Inb4 AmiMoJo:
You just hate ghosts because they look like they're wearing burkas.
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
You can teach a Spade AI to play spades and take tricks. But don't expect it to cover you well if you go null. The idea of sacrificing a trick when your partner is covered seems foreign to it.
What about Ms. Pac-Man?
(-1: Post disagrees with my already-settled worldview) is not a valid mod option.
bed shitting
AI: Artificial Imitation without even understanding of some basic questions like why or what
As usual, there is no AI here, just an algorithm that is tweaked by humans.
You know nothing about this Mi faggot.
I have been training Pac-Man to kill. In fact, I reward him for his efforts so much so he no longer collects the little balls. No, Pac-Man now prefers blood over coin. Anyhow, I may have inaverdantly uploaded the Kill-Man ai to the roomba and I cannot leave the house. Send help! (Also roomba is the name of my hunter killer droid)
I could train the "AI" player not to touch ghosts without telling it not to touch ghosts: just kill the player if it touches a ghost. The digital off-spring will get the hint quickly enough.
"Its just an applied dataset. Nothing magical. Nothing world-changing."
You have no instinct for science.
We know you're an uneducated nazi faggot, there's no need to advertise yourself.
In Pac-Man, you die by touching ghosts. The one thing you want to NOT do, in a regular game of Pac-Man, is touch ghosts. As long as you don't touch any ghosts, you keep getting points.
That's the most basic understanding of the game you can have, what my four-year-old daughter would figure out in five minutes. That's also what the AI figured out.
*More advanced* players can learn the *exception* to the above simple rule. The *exception* is:
Only if you've eaten a Power Pellet in the last few seconds, touching ghosts becomes good.
It doesn't need to "learn not touching ghosts even though (in rare situations) it results in a higher score". It only needs to be too stupid to recognize the unusual conditions under which ghosts, which normally kill you, can instead raise your score. It uses the simple rule "don't touch ghosts, they kill you". Instead of the more complex "don't touch ghosts unless with the last 300 milliseconds ...".
to Do No Evil.
Ba-dum-tish.
I hadn't known there were so many idiots in the world until I started using the Internet -Stanislaw Lem
Take up the Pac-Man's burden
Send forth the ghosts ye breed
Go find your cherry bonus
To acquire the high score that you need
Do read TFA again. They did multiple runs with different emphasis on having the AI emulate the human player who avoided ghosts.
Can we have them teach AI how to run government without taxiing me?
Wouldn't that be explicitly telling the AI not to do it ?
Lets pretend it follows asimovs laws
The power pills still have strategic value when you don't want to eat the ghosts.