What Should and Shouldn’t Be a Concern About AI

There is a lot of fear being spread about artificial intelligence and the data centers that power these models.  Let me dispense with the data center concerns first.  The misinformation that is spread about their power use and cost and about water use and environmental concerns is frightening.   It is frightening because these are economic growth engines that create a lot of well-paying jobs in construction but also in their operation.  They pay large property taxes.  Any locale that has a data center has experienced an economic boom.  These data centers are key to AI growth, which in turn drives productivity increases and faster economic growth.  Data centers at this point are generally producing their own electricity, usually from a co-located power plant.  They are not what is driving up the cost of electricity; the insistence on renewable energy sources is doing that.  Data centers do use water for cooling but that water most often comes from treated municipal waste water, not drinking water, and most is recycled and the heat in the water is used for energy.  Data centers directly produce little or no pollution and have few if any adverse environmental impacts.  But of course China and Russia and our own socialists and America-haters spread misinformation on all these counts, because they know the data centers are critical to maintaining our lead in AI.  If you want to go after a technology with horrific environmental impacts, go after solar panel farms and wind farms.

In terms of artificial intelligence itself, one myth has already been debunked; that AI would cause millions of jobs to be lost.  In fact, AI is a net job producer.  This is what has happened with every technological invention–fear is spread about job loss, but it turns out that more new jobs are created, usually better paying ones.  We saw exactly this phenomenon with the internet.  The AI-related jobs may be different than those that are lost, but they are typically better paying.  Would you rather be a poorly paid insurance claims processor or an AI coder paid far more?  Any government policies should be directed at helping with retraining, we have big need for people who can work in data centers and in AI.

The other misinformation is around AI going rogue and killing everyone.  That is frankly not where the concern should lie.  Artificial intelligence is nothing but a software program, a very sophisticated program but just a program.  It uses data it is given, it takes that data and processes it according to rules given in the software, it makes “decisions” or creates output based on the rules in the software.  The software is written by humans and in some cases by an AI program, but the AI program can only code using the tools and rules given to it by humans.  Software and software models have been around forever; they only cause issues when humans make mistakes or tell the software to do something they shouldn’t tell it to do or have it work on bad data.

If you understand AI models you know that the issues are in the rules given the software for how it makes decisions based on the data it has.  If you have an AI model gather data by reading every human work of fiction or non-fiction or collecting information from media sources, you are going to get what appear to be bad human behaviors–lying, hallucinating events, etc.  If you give the software latitude to make certain decisions, you can’t be surprised when it the software does things you don’t want it to.  The problem isn’t the software, it is the humans who built the software and decided what data it should use and how it weights data in making decisions.  A strong truth is that models only do and only output what we tell them to do or output.

So be afraid of humans, who do lie and cheat, who can be evil and malicious and certainly have used software for those purposes in the past and will in the future.  And yes, laws and rules are needed that have strong guardrails for AI software construction and operation.  But what you should really be afraid of is the US not keeping a lead in AI, because we are more trustworthy than most societies.  Do you want authoritarian China and Russia, much less Iran or North Korea to be way ahead of us in AI capabilities, because they don’t have any of our concerns and they have and will use AI against us.  And our best defense is AI software that can detect and defend against what those countries will do.

AI, like all next generation software, will do very good things.  It will enhance the productivity of just about any industry.  In health care, it already is beginning to help with identifying new drugs.  It has been a phenomenal boost for mathematics and general science.  If you think of everything the software in your phone, tablet or computer can do and how important it is for the quality and efficiency of your daily life or work, AI is just more of the same.  It can be abused, but the abuse isn’t inherent in the AI software; it is from the misuse that can occur with any software; misuse by bad people.  It is always the humans; never the machines or the software.

So believe me when I say that most of the bad things you hear about AI are pushed and supported and amplified by China and Russia and other enemies of our free and democratic society, including those internal voices that hate our country.  And of course there are the usual malcontents and attention-seekers inside the AI companies.  The dumbest thing we can do is slow down and given those countries the lead.  We can have useful AI that has protections from abuse.  We can’t leave ourselves open to AI abuse from our adversaries.