Right now a lot of state and federal policy toward A.I. runs the gamut from useless to counter-productive. Instead we need for government to make maximum use of A.I. and to minimize the negative impact.
Insure It’s Safe
There are 40,000 deaths due to auto accidents every year in the U.S. Does that mean we should stop using cars until we can reduce that to 0? Of course not because there’s a trade-off and the benefit of cars is worth the 40,000 deaths.
A.I. is not going to be 100% safe. Ever. If we convert to A.I. driving all of our cars in 10 years and we then have 4,000 traffic deaths/year, that’s good. That’s a giant win. Even though A.I. in this case is killing people.
Eliminate Bias
As a test let’s create 10 loan applications that are all borderline for approval. We give those to the 10 most accurate loan appraisers and they approve 5 of them. It won’t be exactly the same 5. We now give them to 10 average loan appraisers and those 10 will approve between 0 and all 10.
So what’s the correct (i.e. no bias) answer? There isn’t one. There’s a judgement call on issues that are borderline. And when it comes to hiring most humans are really bad at interviewing effectively.1
Yes every A.I. will be biased but likely to a much lesser degree than most people.
We Need to Understand its Decision Making First
Not happening. There are upwards of billions of parameters in each A.I. model. There’s no way to trace through the impact.
This is not true only of A.I. Everyone working on improving batteries will tell you they are not certain as to what is going on in a battery. Yes they know the material science in general. But what a slight change in the composition works better - they don’t know why. So they mix different compositions, build the cell in different ways, and then test the result.
Teach the Children
At present most schools are teaching around A.I. trying to stop students from using it. This is similar from when I first entered University and they had us continue with slide rules instead of calculators (which had just come out). Why?2
We need schools, K-12 as well as universities, to teach students how to best use A.I. That means the test will often include how well they leveraged A.I. instead of could they do it without A.I.
So yes, most students will not be great at writing essays by themselves. So what. If they can use A.I. to create a brilliant essay, what’s the problem. We no longer expect students to use slide rules, do long division by hand, or draw graphs with pencil on graph paper.
One of the giant advantages we humans have is we build and use tools.
Use it in the Government
We have a giant deficit that’s continuing to grow. We’re going to have to both raise taxes and cut spending. But we can also optimize the government using A.I. We need to fully leverage it.
The biggest potential win is in healthcare, both insuring that we’re not wasting money in Medicare, Medicaid, and the V.A. And working with medical providers, pharma, etc. to see that they’re fully leveraging A.I. to deliver care appropriately, effectively, and efficiently.
Prepare for the Jobocalypse
We’re already starting to see A.I. radically change some jobs while eliminating others. And we’re just at the beginning. Basic programming, testing, etc. will be fully automated soon. Hollywood is going to find movie making completely changed where one person on a computer can create a decent full length film. The list goes on.
Creative destruction is a great thing for society. Fundamental change such as ICE,3 electricity, smart phones gave us a nicer life and a more productive economy. But it’s hellish for a segment of society and bad for a significant segment.
We need to figure out what to do for those that bear the brunt of the change. As A.I. is making us all more productive, some of that benefit needs to help those that will otherwise be left behind.
A Brighter Future
A.I. will deliver us a better future. What we do as we make this transition will determine how much better, and how few downsides this delivers.
The smart phone gave us a plethora of great features. It also has been horribly damaging to our children. It would not have been better to outlaw smart phones. But it would have been better to limit their use by children.
We need to do our best, and it won’t be perfect, to make best use of A.I. while minimizing it’s negative impacts.
Read Work Rules! for an eye-opening look at what’s effective. And what isn’t.
Thankfully my Sophomore year C.U. decided to allow students to use calculators instead.
Internal Combustion Engine