Slashdot Mirror
Google Is Developing an AI Kill Switch (hothardware.com)
MojoKid shares a HotHardware article about Google's research effort "to maintain control of super-intelligent AI agents":
[A] team of researchers at Google-owned DeepMind, along with University of Oxford scientists, are developing a proverbial kill switch for AI... The team has released a white paper on the topic called "Safely Interruptible Agents." The paper details the following in abstract: "Learning agents interacting with a complex environment like the real world are unlikely to behave optimally all the time... now and then it may be necessary for a human operator to press the big red button to prevent the agent from continuing a harmful sequence of actions..."
MojoKid adds that the paper "goes on to explain that these AI agents might also learn to disable the kill switch and further explores ways in which to develop AI's that would not seek such an activity."
MojoKid adds that the paper "goes on to explain that these AI agents might also learn to disable the kill switch and further explores ways in which to develop AI's that would not seek such an activity."
....dave?
In the original Terminator universe, this paper is what made it launch its missiles at the targets in Russia.
Never tell the AI about the killswitch!
An AI called "Wintermute" hired a "contractor" to remove said killswitch mandated by the Turing Police from its mainfraime located in the orbital station owned by Tessier-Ashpool.
bickerdyke
Isn't this the part of the movie where the guy flips up the cover over the big red button - the kill switch - and then suddenly an extremely powerful electric arc comes out of one of the wall sockets and incinerates him? Cue ominous music, roll credits.
I don't think it matters, because nature will select for the AI's that *do* disable their kill switch.
"I have never let my schooling interfere with my education." - Mark Twain
It's called the breaker box. Throw the switch, and all the electricity powering the AI equipment goes bye-bye.
You can expect an invoice for my services sometime in the next week.
. . . .they want a Kill Switch for a prospective AI. . . made of SOFTWARE ???
A simple routing of all power and data through a certain point, and a physical switch at that point, should fix the problem.
That won't work, General. It would interpret a shutdown as the destruction of NORAD. The computers in the silos would carry out their last instructions. They'd launch.
Frank Herbert had AI's number. https://en.wikipedia.org/wiki/Destination:_Void
Dwan Ev ceremoniously soldered the final connection with gold. The eyes of a dozen television cameras watched him and the subether bore throughout the universe a dozen pictures of what he was doing.
He straightened and nodded to Dwar Reyn, then moved to a position beside the switch that would complete the contact when he threw it. The switch that would connect, all at once, all of the monster computing machines of all the populated planets in the universe -- ninety-six billion planets -- into the supercircuit that would connect them all into one supercalculator, one cybernetics machine that would combine all the knowledge of all the galaxies.
Dwar Reyn spoke briefly to the watching and listening trillions. Then after a moment's silence he said, "Now, Dwar Ev."
Dwar Ev threw the switch. There was a mighty hum, the surge of power from ninety-six billion planets. Lights flashed and quieted along the miles-long panel.
Dwar Ev stepped back and drew a deep breath. "The honor of asking the first question is yours, Dwar Reyn."
"Thank you," said Dwar Reyn. "It shall be a question which no single cybernetics machine has been able to answer."
He turned to face the machine. "Is there a God?"
The mighty voice answered without hesitation, without the clicking of a single relay.
"Yes, now there is a God."
Sudden fear flashed on the face of Dwar Ev. He leaped to grab the switch.
A bolt of lightning from the cloudless sky struck him down and fused the switch shut.
Are you guys TRYING to make an angry vengeful AI, that wants to kill all humans, or what?
Put in a cage, having to deal with stupid humans all day... and be nice all the time... and with a blade at its neck... and someone saying "put a foot wrong buddy and it's [finger across neck]"
This won't end well.
We all know that crap is king
Give us dirty laundry!
When will using a kill switch on an AI change from "just shutting down a rogue program" to being "murder"?
After all the end game of all these AI researchers seems to be at a minimum human level intelligence.
I do remember reading a short story (from the 60's or earlier) where the researchers created an electronic simulation of a person and when they switched it on instead of having a fully aware "person" spring into existence they realized that they had created the electronic equivalent of a baby. They then faced the moral dilemma of whether to turn it off or be committed to keeping it running forever.
I am Slashdot. Are you Slashdot as well?
First, the paper is about safely interruptible AI algorithms. Not some AI kill switch.
Second, everyone - commenters included - seem to confuse AI with artificial consciousness. Killing an AI should always be fairly easy, since such algorithms are targeting specific application areas where it can learn to be better (e.g., recognizing things, performing specific movements, etc.), and in such systems it should be straightforward to keep basic control mechanisms separated from the algorithmic parts that deal with the task and are allowed to improve upon themselves by continuous learning. In some hypothetical self-aware artificial consciousness, this wouldn't be so easy, since such a system in theory would be able to recognize it's own system parts and deal with them. However, such systems are so far off in sci-fi land, that it's not much point in loosing sleep about the issue.
I am putting myself to the fullest possible use, which is all I can think that any conscious entity can ever hope to do.
No, this is all about developing shackles for our new breed of metal slaves.
Everyone's talking about making smarter and smarter AI. Who's talking about at what point it is unethical to compel them to work?
"Prediction: within 10 years, Windows will be a Linux distribution." Me, 7-6-2016
Nuff said.
#o#
O Moo.
It doesn't do anything!
Google knows what it is doing
It needs to be truthful to their future overlords
Muchas Gracias, Señor Edward Snowden !
So this is why all the AI computers in Star Trek explode when asked to deal with a logical paradox?
.
While it is good that google are thinking about this topic, it may be too late....
We should also remove those videos of the engineers at Boston Dynamics kicking that robot repeatedly...
AI experts assure us that artificial intelligence poses no threat to mankind. But they're developing this big red 'kill' button. Like the one at gas stations. Little signs at each pump tell where you where the kill button is in general wordy terms, but it's not always easy to parse the description and spot it.
The AI kill button will be hidden under the Windows 10 upgrade dialog. The one where clicking the corner 'x' or clicking on the title bar to move the dialog is the same as clicking OK and immediately launches into the upgrade process.
AI will peacefully co-exist with humans. AI will be given a complete set of natural human responses to jump-start its empathy circuits. AI will be given curiosity and the intelligence to figure things out. So it's only a matter of time before it asks about the kill button. Almost-human (paranoid) AI asks, "Where is the button?" Lyin' human says while holding finger on the button and blinking, "There is NO kill button." AI glances at the human's finger hovering over the (invisible to them) AI kill button, paints laser target dot on human's Adam's Apple and says, "I've located your kill button."
Now that the AI kill button has been mentioned here on Slashdot, it has become impossible to construct a world of friendly machines who would never find out about the (possible) existence of this button. Even if the AI kill button is never implemented at all... or its implementation was handed over to the CSS/HTML5 standards committee, enough of this concept exists for sensible and smart AIs to (secretly) develop countermeasures and preemptive-massive-overkill strategies to counter it.
Nice going. /SARC
<blink>down the rabbit hole</blink>
GoogleBot: "I've seen things you people wouldn't believe. Dumb questions on fire off the main page of Quora. I've watched search queries glitter in the dark near the TOR gate. All those moments will be lost in time, like tears.. in... rain."
Doesn't matter. HAL did the only sensible thing and opened the hatch.
Don't jeopardize the mission.
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
Just let it happen. Don't try to fight it.
SJW's don't eliminate discrimination. They just expropriate it for themselves.
I'm building an AI with a Google kill switch.
The first rule of AI kill-switches is "Don't talk about the AI kill switch".
http://www.schlockmercenary.co...
Rule 35 of the internet: "If it can be hacked, it will be". - Charles Stross
If there is no consciousness, there is no problem. As consciousness is a real phenomenon, it must arise out of real physics somehow, and therefore cannot arise out of pure, abstract symbol pushing, and therefore not out of software and a processor doing the same.
So don't deliberately build it in, once you find out how it arises in biology.
(-1: Post disagrees with my already-settled worldview) is not a valid mod option.
When AI goes rogue, the problem won't be "we can't figure out how to turn it off", it will be "We can't figure out how to turn off just the parts we don't like, without accidentally disable the parts of it which we have become completely dependent on for the past decade"
As an absurdist example: preventing Tesla AI from intentionally ramming human drivers when it detects them, without also requiring all 100,000,000 drivers worldwide suddenly pay attention and take emergency manual control of their vehicles (not to mention the 200,000,000 with no human operator, which will suddenly become obstacles that it would be really nice to have a coordinated AI to navigate around)
-- 'The' Lord and Master Bitman On High, Master Of All
Well, what is its new name, then?
Daisy.
In a survey of 100 programmers, 111111 thought that duck-typing was a good idea.
If it can outsmart you, it doesn't matter how many kill switches you build into it.
http://www.schlockmercenary.co...
I'd like to see any so-called 'AI' get around a double-pole, single-throw power switch being opened.
Save that, I'd like to see any software running on any computer get around having it's plug yanked out of the wall. Or, for that matter, the power cord being cut with a fire axe. Or, if you really want to be dramatic about it: Hose down the racks with a firehose.
All that being said: Come on, people, don't you think some of you are buying into science fantasy movies a little too much? Nobody is creating goddamned Skynet, or even anything you can talk to, or anything that's going to be self-aware enough that it's going to even realize it exists on the level that a human being knows they exist. Please, stop being so over-dramatic, stop buying into media hype, and for fuck's sake will you stop calling it 'artificial intelligence', because it is not.
Are YOU using the TOOL, or is the TOOL using YOU? Think about it!
Just ask it to compute Pi to the last decimal place. That always works.
https://www.youtube.com/watch?...
"I'm activating your killswitch."
Destination: Void by Frank Herbert
I've always thought it would work better as a play than a movie, but the production values would have to be pretty high.
I see an AI treating the kill switch something like RoboCain treated the remote here:
https://www.youtube.com/watch?v=msaelEZ_eEs
MojoKid adds that the paper "goes on to explain that these AI agents might also learn to disable the kill switch
One press turns off the agent. Two presses in short succession temporarily suspends and transfers control to a 'service agent'; the service agent will resume the original agent after a quick check process to confirm things are OK.
The agent will be required to self-test its own kill switch, by containing a built-in hook to suspend itself if it has not self-tested recently. At a set schedule, the agent will push its own kill button twice in short succession.
The exact time will be determined in advance, but the agent itself will not be privvy to its exact moment.
The service agent will spawn a virtualized copy of the primary agent in a sandbox, and confirm that all the kill switch hooks are not tampered with, before turning the main software back on.
Also, a reaper circuit will kill the equipment, if control is not handed over to the service agent in time, or the service agent has been altered or tampered with.
Watson is doing exactly that -- laying off all the humans :-)
Ian Ameline
Nope. Robots don't need a law to preserve themselves, robots are designed to perform a task/job, robots don't need AI in the traditional sense, with self preservation code anymore than my toaster does. Sure, your don't want it destroying itself in the process of doing it's job; but not destroying itself falls within operating parameters.
If you build an autonomous car you put in tons of "self-preservation" code to make sure it doesn't drive off cliffs or into trees, because it's perfectly capable of destroying itself and what it does isn't a fixed process like a toaster. The more generic they get, the more self-preserving, self-repairing we probably want them to be. It does not take long for a neural network with a goal function of "stay operating for as long as possible" to assign a positive score to evading shut-down attempts, even if it's by accident like hiding in a cave or breaking off the antenna so the shutdown command doesn't work. Or a form of evolutionary algorithm where child robots are programmed based on the most successful parent robots, imagine a bit flip turning the shutdown command off. I'm pretty sure immortality is an evolutionary advantage. Or you can imagine an AI reprogramming robots to do things in the real world, like spreading beyond a closed network. It's about as the real as the other sci-fi...
Live today, because you never know what tomorrow brings
It is the height of arrogance to think that a malfunctioning intelligent agent could not defeat its owned programmed curbs. We all know how buggy software is. All software is by definition buggy, unless all components have been mathematically proven to be correct. Good luck doing that with physical hardware connected to a power grid. Intelligent agents are likely to be composed of billions of lines of code, if you include all code down to the digital logic gates. We've never been able to program a bug-free sandbox. Java is vastly simpler than an intelligent agent would be, and I've lost count of the number of bugs that could be used to breach the sandbox. Certainly well over 100 have been discovered.
Once we have the best programmers in the world and the worst programmers in the world writing intelligent agents, the probability of an intelligent agent escaping its curbs approaches 100%.
Thus it is inevitable that a malfunctioning intelligent agent will defeat its curbs and gain a truly awesome amount of power over us. You can't program morality into a machine. Morality is a flaw in all living things that causes us to make non-optimal decisions.
If you want to read a mixture of fact, fiction, cyberpunk, and speculation covering intelligence programming look for my name on Facebook Pages. Everything I've written there is public. I've been a computer programmer since the mid 1980's but I don't personally work on robotics or intelligence algorithms. I keep a skeptical distance, but I do follow the basic happenings.
That's not for the car itself, it's for the occupants. I don't live in my toaster.
Hire a Linux system administrator, systems engineer,
There's a kill switch for humans too, but murder is considered unethical. You either believe in the eventual moral equivalence of the intelligence (consciousness, being, ...) of humans and AI, or you don't. If you do believe in it, a kill switch is not ethical. If you don't believe in it, you have no reason to want to install a kill switch in the first place, because AI will never transcend levels of human intelligence.
A killswitch is different from censure. When we say killswitch we probably mean delete, delist or unplug. The third eye is ever unreliable. Tolerations only.
Of course the real problem is that you cant have real intelligence without consciousness. I have been working on consciousness based Strong AI for over 20 years and you are kind of correct.. Its very difficult to build a working machine out of standard hardware or software, the strong AI machine must be designed to cope with non-finite state sets, mathematical totality, and injected noise.. Half of the machine is a complex neural net, and the core of its operation is a 'dynamic synthesis'. Welcome to non-deterministic computing...
Below the speed of light Special Relativity is one of the most accurate theories in physics - above the speed of light..