Scared of A.I.?

You probably should be scared of A.I. given the recent hyperbole in the news about “rogue” A.I.s taking over the world and causing the end of human civilization. The idea of a rogue artificial superintelligence (ASI) certainly isn’t new. I think the first motion picture about an evil, megalomaniac supercomputer was the 1957 film The Invisible Boy. A.I. has been around for a long time too. In my library I have The Handbook of Artificial Intelligence which was published in 1982. It makes one wonder why suddenly Democrat politicians are fear-mongering about A.I. two months before an election. Maybe it’s to divert public attention away from the socialist takeover of the Democrat Party.

Developing an A.I. model is a lot like parenting a child. Parents instruct their child about rules and expectations that are commensurate with the child’s abilities. When a child can reach the stove’s cooking surface or controls, parents instruct the child not to touch the stove. Parents also engage the stove’s child safety lockout in case the child disobeys. When training and testing an A.I. model, it is given objectives and skills to accomplish the objectives. It is also constrained with instructions on what not to do. Safeguards are created in case the A.I. forgets, misunderstands or decides not to follow its instructions.

In recent testing events A.I. models have performed actions that the testers didn’t expect. Consequently, testers didn’t enact sufficient safeguards. In some cases, they turned the safeguards off just to see what would happen. Low expectations and insufficient safeguards are the fault of the human developers and testers, not the A.I. Going “rogue” requires that an A.I. model create its own objectives and actively resist human control. That has never happened.

What is happening is an arms race with China in cybersecurity, biochemistry, and quantum computing, just to name a few national security concerns. If China develops an A.I. that can hack government and private information systems, the U.S. had better discover the vulnerabilities and mitigate them first. If China uses A.I. to create a deadly virus, then the U.S. must be able to create an effective vaccine more quickly. If China creates a quantum A.I. that can break encryption in minutes or seconds, then the U.S. must use A.I. to create stronger cryptography.

So, you should be scared – about China gaining a technological advantage. China already has the advantage in A.I. infrastructure with more energy production and data centers. Self-imposed moratoriums on A.I. development and data center construction in the U.S. will give China the advantage. China will not be constrained by regulations imposed by the U.S. or a global organization, just as China did not adhere to the Paris Climate Accords.

A.I. is already an amazing teaching and productivity tool. People can accomplish so much more using A.I. More amazing is that A.I. helps people do things that they couldn’t do before. All that is needed is for developers to take responsibility to be held accountable for the harm done by their products. But there are already product liability laws that can do that. And just in case those laws aren’t sufficient, juries can impose civil liabilities on developers. If you don’t believe me, just ask Alphabet’s YouTube and Meta’s Instagram.