AI: From fear to engineering
We hear two seemingly conflicting messages about artificial intelligence: America must move faster to “win the AI race,” while some AI experts warn that development may be moving too fast and should slow down.
Why the concern?
As AI becomes more capable, it may be allowed to operate other computer systems, communicate across networks and make decisions at tremendous speeds. If a powerful AI operates outside its intended limits, it could potentially disrupt banking, communications, utilities or other systems we depend upon. The concern isn’t that AI suddenly becomes “evil.” The concern is whether humans can remain in control of what it can do.
That’s a serious engineering problem – but humans have solved serious engineering problems before.
Consider the circuit breaker in your home. When something exceeds a safe limit, it shuts off power. If you don’t know why it tripped, you don’t keep resetting it. You call an electrician, find the problem and determine whether restoring power is safe.
Could we engineer an AI circuit breaker on the same principle?
Countries developing powerful AI systems could agree on measurable danger limits. If an AI system crosses them, an independent safety system could isolate it – interrupting its ability to communicate with or act upon outside systems. Qualified people would determine what happened, and independent human approval would be required before reconnecting it.
Perhaps “slow down” should be defined, not just stated. It could mean: time to understand the risks, agree on safety standards, and engineer the protections needed to move forward safely.
Moving forward matters. AI will help us discover, design, manufacture, educate, heal and solve difficult problems mankind struggles with.
America should aspire to lead in using AI for good as well as in engineering the safeguards that keep mankind safe.