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 |

Would you like to contribute a story tip? It's easy:
Get the URL of the story, and the related sf author, and add it here.

Comment/Join discussion ( 0 )

Related News Stories - (" Artificial Intelligence ")

A Bayesian Approach to Safe Imitation Learning For AIs and Robots
Um, how about that pension for the humans who serve as the models for robot behavior?

Our GodBot, Who Art In Cyberspace
Vaal hungers! We must serve him.

Will Robots Be Moral If We Raise Them Like Our Children?
'The birth of Machine, my robot child...' - Henry Slesar, 1958.

Rule Of Humans By Software Not Transparent
'The Council itself could be overridden by a superior power...' - Arthur C. Clarke, 1956.

 

Google
  Web TechNovelgy.com   

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.

 

 

 

 

 

Current News

3D Printed Artificial Muscles Are Stronger Than Yours
Bots don't need to work out.

Fog Computing (AKA Edge Computing) Ad Hoc Networks
'The tiny devices chirped their impulse codes at one another...'

Dubai Scorpion Police Hoverbike Ready To Pull Young Kirk Over
'Is there a problem, officer?'

HEXA Robotic Help For Plants
Then some unknown race had chanced upon the dreamers and decided to "help them out."

Korean Tesla Model S Video 'Excelsior' Is Indeed Our Motto
'Improving man by bringing him close to Nature, while they combine the sensations of coasting with the interest of seeing the country well...'

DIY Robot Shoots You In The Face
'...there were automatic guns that fired ligamine darts.'

A Bayesian Approach to Safe Imitation Learning For AIs and Robots
Um, how about that pension for the humans who serve as the models for robot behavior?

Qoobo Headless Robotic Therapy Cat Was Anne McCaffrey's Idea
'...used as surrogates in intense dependency cases.'

Autonomous Cars Talk To Each Other At MCity
'My cars talk to one another.'

PUFFER Robots - From Philip K DIck's Second Variety?
'Across the ground something small and metallic came, flashing in the dull sunlight of midday.'

Russian Space Garden
'We saw the gardens, flooded with artificial sunlight...'

Targeted Neuroplasticity Training For 'Downloading Skills'
'I know kung-fu.'

U of M's MCity To Feature Asimov's Automatobuses
Should you turn autonomous buses off?

Crazyflie Drone Swarm Technology
'...Programmed to hang in space in a hexagonal grid pattern.'

Our GodBot, Who Art In Cyberspace
Vaal hungers! We must serve him.

easyJet Short-Haul Electric Jets
Have a little faith, will you? They're working on it.

More SF in the News Stories

More Beyond Technovelgy science news stories

Home | Glossary | Invention Timeline | Category | New | Contact Us | FAQ | Advertise |
Technovelgy.com - where science meets fiction™

Copyright© Technovelgy LLC; all rights reserved.