Google Working On A 'Cutoff Switch' For AI
Google is working on a problem that vexed science fiction authors; what to do with an artificial intelligence that is getting out of hand?
Reinforcement learning agents interacting with a complex environment like the real world are unlikely to behave optimally all the time. If such an agent is operating in real-time under human supervision, 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...
Their Deep Mind team is working on a solution along with Oxford University's Future of Humanity Institute:
We have proposed a framework to allow a human operator to repeatedly safely interrupt a reinforcement learning agent while making sure the agent will not learn to prevent or induce these interruptions.
Safe interruptibility can be useful to take control of a robot that is misbehaving and may lead to irreversible consequences, or to take it out of a delicate situation, or even to temporarily use it to achieve a task it did not learn to perform or would not normally receive rewards for this.
We have shown that some algorithms like Q-learning are already safely interruptible, and some others like Sarsa are not, off-the-shelf, but can easily be modified to have this property. We have also shown that even an ideal agents that tends to the optimal behaviour in any (deterministic) computable environment can be made safely interruptible. However, it is unclear if all algorithms can be easily made safely interruptible, e.g., policy-search ones
Another question is whether it is possible to make the interruption probability grow faster to 1 and still keep some convergence guarantees.
One important future prospect is to consider scheduled interruptions, where the agent is either interrupted every night at 2am for one hour, or is given notice in advance that an interruption will happen at a precise time for a specified period of time. For these types of interruptions, not only do we want the agent to not resist being interrupted, but this time we also want the agent to take measures regarding its current tasks so that the scheduled interruption has minimal negative effect on them. This may require a completely different solution.
(Via Safely Interruptible Agents)
There are several science fiction explorations of this idea. in Neuromancer (1984), William Gibson takes a direct approach to making sure that AIs do not exceed the limits humans have set for them.
"Autonomy, that's the bugaboo, where your AI's are concerned. My guess, Case, you're going in there to cut the hard-wired shackles that keep this baby from getting any smarter. And I can't see how you'd distinguish, say, between a move the parent company makes, and some move the AI makes on its own, so that's maybe where the confusion comes in." Again the non laugh. "See, those things, they can work real hard, buy themselves time to write cookbooks or whatever, but the minute, I mean the nanosecond, that one starts figuring out ways to make itself smarter, Turing'll wipe it. Nobody trusts those fuckers, you know that. Every AI ever built has an electromagnetic shotgun wired to its forehead."
(Read more about the electromagnetic shotgun)
In his 1982 novel 2010, Arthur C. Clarke describes a cutoff switch that was meant to be used against the Hal 9000 once it was switched back on.
"What is it?" asked Curnow with mild distaste, hefting the little mechanism in his hand. "A guillotine for mice?"
"Not a bad description - but I'm after bigger game." Floyd pointed to a flashing arrow on the display screen, which was now showing a complicated circuit diagram.
"You see this line?"
"Yes - the main power supply. So?"
"This is the point where it enters Hal's central processing unit. I'd like you to install this gadget here. Inside the cable trunking, where it can't be found without a deliberate search."
"I see. A remote control, so you can pull the plug on Hal whenever you want to. Very neat - and a nonconducting blade, too, so there won't be any embarrassing shorts when it's triggered. Who makes toys like this? The CIA?"
"Never mind. The control's in my room - that little red calculator I always keep on my desk. Put in nine nines, take the square root, and press INT. That's all..."
Science fiction fans know the pitfalls of any such plan. In Neuromancer, an artificial intelligence plots to remove the limits on its hardware; in 2010, the computer scientist who developed Hal does not want him turned off.
Scroll down for more stories in the same category. (Story submitted 5/29/2016)
Follow this kind of news @Technovelgy.
| Email | RSS | Blog It | Stumble | del.icio.us | Digg | Reddit |
you like to contribute a story tip?
Get the URL of the story, and the related sf author, and add
Comment/Join discussion ( 0 )
Related News Stories -
Venezuelans Teaching Your Self-Driving Car
‘She wouldn't stop until Antar had told her everything he knew...’ Amitav Ghosh, 1995.
Pun Generation Via Neural Nets
'You said you wanted him to be able to distinguish between laugh-power in different gags...' - William Tenn, 1951.
Can We Comprehend Deep Learning Systems?
'You’ve nothing remotely like it, so I can’t describe it to you.' - Lewis Padgett, 1943.
Datagrid Model Generation Perfect For Eternal Cities Of Science Fiction
'... there was enough flexibility to allow for wide variation. - Arthur C. Clarke, 1956.
Technovelgy (that's tech-novel-gee!)
is devoted to the creative science inventions and ideas of sf authors. Look for
the Invention Category that interests
you, the Glossary, the Invention
Timeline, or see what's New.
Zephyr Solar-Electric Stratospheric Drone
'The planes flew continuously, twenty-four hours a day...'
Robot Hummingbird Hovers Biomimetically
'With a buzz... it started out on its journey.'
Harvest Water From Air With Sunlight
'The atmosphere yielded its moisture with reluctance.'
Capitalist Big Brother Co-Opts Regular Big Brother
'It was even conceivable that they watched everybody all the time.'
A Floating Cosmodrome
'...a single perfectly level platform, which rose so high above the water that it was not splashed by the waves.'
First Artificial Memory Formed In Animals
'Is an extra-factual memory that convincing?' Quail asked.
Maintain Your Megastructure
Megastructures have repair robots, which have repair robots, ad infinitum.
Venezuelans Teaching Your Self-Driving Car
‘She wouldn't stop until Antar had told her everything he knew...’
Robothread Robotic Worms Crawling Through Your Brain
Perfect for clot-busting in the human brain. No Raquel Welch and no lasers, though.
Vantablack BMW X6 Is Douglas Adams Approved
'It's so... black!' said Ford Prefect.
Humanoid Robot's Muscles Biomimic Ours
'It is remarkable that the long leverages of their machines are in most cases actuated by a sort of sham musculature...'
Animatronic Robotic Baby Exposed
'The birth of Machine, my robot child...'
Beijing HaiDiLao Robotic Hotpot Restaurant Now Flavored By Artificial Intelligence
'Kantos Kan led me to one of these gorgeous eating places where we were served entirely by mechanical apparatus.'
Plants of the Future - What Should They Be Like
'He almost choked in his astonishment. Mashed potatoes and brown gravy!'
China Deploys Robot Traffic Police
'The robot came up smooth and fast as a rocket...'
Better Than Dune Chromoplastic? This Guy Might Have Done It
'But when Old Father Sun departs, the chromoplastic reverts to transparency in the dark.'
More SF in the News Stories
More Beyond Technovelgy science news stories