How Can We Make Artificial Intelligence Safer?


“Artificial Intelligence” conjurs images that range from Wall-E to the Terminator, but this growing field has spawned some of our most helpful devices from The Nest, to control our home environment, to Siri’s ubiquitous snark-filled responses to our sports questions. For those interested in pursuing this growing field, OpenAI is the cutting-edge, nonprofit research hub supported by some of the stars in the tech field, like Tesla’s Elon Musk and venture capitalist Peter Thiel.

OpenAI’s mission is to develop open-source A.I. software that will benefit all of humanity without the need to worry about the bottom line and profit margins, allowing for creatively and motivation-driven software development. In order to help developers, they created OpenAI Gym, the reinforcement learning (RL) algorithms toolkit. OpenAI addressed the next set of A.I. challenges when they issued a paper with Google Brain outlining how machine-learning systems can experience failure or frizzle into bad patterns.

“Many of the problems are not new, but the paper explores them in the context of cutting-edge systems,” OpenAI’s website explains.  “We hope they’ll inspire more people to work on A.I. safety research, whether at OpenAI or elsewhere.”

Some of the greatest challenges OpenAI and Google suggest developers could face include:

Safe Discovery

How does the A.I., no matter the form (robot, software, etc.), investigate it’s environment without accidentally causing itself harm?

Management of Distributional Shift

Does the A.I. pivot as needed without inducing system chaos when it encounters new and unanticipated data?

Impact of Negative Side Effects

How does the A.I. impact the surrounding world without causing harm or chaos? For example, can it locate, clean and stack all the glasses from a party without breaking the priceless art on the table tops?

Avoiding Gaming the System

Much like a kid figures out that if they fake an illness they get to stay home from school, how do you keep the A.I. from tricking its environment so it gets unearned rewards

Scalable Direction

Does the A.I. require less direction and feedback as it learns?

A.I. is an emerging field with deep impact for society. At Consultant Specialists, Inc. we can help you make your mark. Contact us for more information about openings and placement services, and working with a top IT recruiter in San Francisco!

Leave a Reply

  • (will not be published)