An Aikido Approach to Chatbots

Tools can be useful,
but don’t count on them to think.
Use them mindfully.

One of the things I’ve noticed is that the discussion of guns for self defense all seem to start — and end — with the purchase of said gun. Perhaps a few of those who hold the view that “an armed society is a polite society” (to quote Robert Heinlein) also advocate serious training, but it’s easy to get the impression that too many people think owning the gun is all you need to protect yourself.

I wrote a story about this called “Survival Skills.” In it an Aikido sensei told the protagonist that no tool is ever ultimately the answer. The protagonist had to learn the core truth of that the hard way, though.

I bring this up because all the furor about the AI chat bots has skipped over analyzing them as a tool that has both benefits and flaws. Some people are already using them to replace humans, without paying any attention to some of their significant flaws. (A writing program that makes up facts and cites non-existent articles is not a tool to rely on.)

And the scammers are already out in full force: people are submitting chatbot written stories to magazines. The biggest problem from the magazine POV is not separating them out from real stories — that’s pretty easy — but the fact they flood the inbox, exhausting the editor who has to deal with them.

Nobody’s going to make any money sending chatbot stories to magazines, but someone’s probably making money teaching people how to do that.

My Aikido teacher used to occasionally say, “I teach philosophy,” meaning that Aikido is so much more than a physical practice. I try to apply the principles of Aikido to other aspects of life.

I just applied two Aikido principles to the discussion of chatbots: relying on a tool when you don’t understand what you’re doing with it and acting without integrity. Aikido teaches you to avoid both of those things.

A lot of Aikido is at heart about paying attention, not jumping to conclusions. So as an Aikido practitioner analyzing a new and disruptive technology, I think we should look at the bigger picture. Those who believe in the tech but fear it might create an existential crisis for humanity have leapt without looking.

Others who are looking at what these large language models are trained on and pointing out that they are picking up a lot of bad information have a grasp of the bigger picture. These people are not opposed to using LLM tech to develop what is called AI, but they are interested in doing it responsibly, ethically. And slowly.

That is, they’re coming at it with integrity and learning how and when the tool might be appropriate.

I’ve noticed a couple of “freak out” topics in the media about modern tech in general of late. One is that kids who have grown up using direction apps don’t have a good sense of where they are. Paper maps work better for teaching our relationship to space.

But even paper maps are flawed. Both of these things are flawed in different ways. We need to remember that the map is not the territory.

I looked that idea up. It comes from a mathematician named Alfred Korzybski. Reading about it reminds me that engineers often say, “All models are flawed. Some are useful.”

Likewise people freak out that kids don’t know how to write cursive. But those freaking out are putting too much importance on cursive. It is a tool. It can be a useful tool, but it also has flaws and isn’t useful to everyone.

Tools can be useful, but no tool is useful in all circumstances.

I think the issue with all tech tools that might be worthy of concern is whether people are learning to pay attention.

I always think paying attention is the first and most important lesson of Aikido. And while it’s inherent in most martial arts teaching, it’s taught best in Aikido. One of the things that we do regularly in Aikido classes is work on situations with multiple attackers. That teaches you to be aware of a lot of factors at once, because if you just focus on one attacker, one of the others will hit you over the head.

In using digital tools, paying attention includes developing the ability to see whether the tool is giving you a useful response. Thinking back to the first computer-like tool I ever used regularly – a pocket calculator – I remember that I was very glad I had a good grounding in some basic arithmetic. I knew whether the answer I was getting was logical. (This is very useful with spreadsheets as well and I suspect with many other programs that calculate things.)

Paying attention also requires remembering that the LLMs are not, in fact, on the verge of being sentient beings. They are software that can predict reasonable responses relying on a great deal of information that has been reduced to digital formats. They can’t work with any subject that cannot be so reduced.

We must pay attention so that we don’t let anyone or anything, including the chatbots themselves, convince us that they are something more than they are.

Of course, we also need to pay that kind of attention to other human beings. We have plenty of people in the tech world  and also in politics who spend a lot of time trying to convince us that they understand the world and the future better than the rest of us. They may even believe that.

But we don’t have to.

Leave a Reply

Your email address will not be published. Required fields are marked *