Conversation

***** AI vs. Cats! *****

This the script of my national network radio report from last Monday, reacting to the lead-in of a Meta AI expert admitting that AI doesn't even have the intelligence of a cat. I agree. As always, there may have been some minor wording changes from this script as I presented this report live.

- - -

So now we finally get to the heart of the matter about artificial intelligence. It's NOT intelligent at all. These systems don't have any intelligence as we would normally use the term. In fact, they're as dumb as damp sponges -- no offense meant to damp sponges of course. These systems can often find some kinds of patterns in data much faster and more reliably than humans. That's a good thing.

On the other hand, generative AI can spit out summaries of data that they've previously sucked in, with decidedly varying levels of accuracy, often misleading and sometimes even dangerously incorrect summaries in whole or in part.

Neither of these kinds of abilities are intelligence. And that cute kitty or doggy that may be in the room with you right now has more common sense and abilities to deal with the real world than all the AI systems on the planet rolled into one big ball of electricity sucking CPUs.

And frankly, this is intuitively obvious to all of us, despite the endless stream of AI promotional hype the AI firms have been shoving down our throats and in our faces.

A bit of history. You might not know that this is actually the second wave of AI. The first was back in the latter half of the 20th century, from places like the MIT AI Lab on the east coast and the Stanford AI lab out here on the west coast. In their heyday I used to visit both of them whenever I had the opportunity. And all manner of important computer innovations came out of those centers.

But when you come right down to it most observers feel that the approaches available at the time didn't actually lend themselves to really successful artificial intelligence projects even under the more limited definitions for AI being used at the time. Still, they were fascinating places doing fascinating work, even though the results in an AI context are generally viewed as having been limited.

So fast forward to today and the sheer processing power being thrown at AI systems. Both Microsoft and Google are making deals for nuclear plants to provide the massive amount of electricity these systems need.

And the resulting generative AI large language models are being rapidly deployed to consumers whether they want them or not. And their promoters keep talking about how these systems will be more intelligent than humans in 10 years or 20 years, or more or less. Of course they say this is dependent on their continuing to have a pretty much endless flow of untold billions of dollars into their development coffers. Of course that's what they'd say.

Make lots of predictions and promises, and when things don't go as planned, hey, just pretend that's what you were expecting all along. Then they ask for even more billions and it's deja vu all over again.

This is certainly not to say that AI systems aren't useful. We know many areas where their pattern analytics and related capabilities can be extremely useful -- medical diagnostics is only one example. On the other hand, I personally consider generative AI such as chatbots and the like to be typically useless at best, and potentially dangerous misinformation sources that have the potential to do serious damage in some cases.

Right now there don't seem to be breakthroughs on the horizon that will alter this dynamic in major ways. Of course, technology can change very suddenly. A paper might be published tomorrow that will supercharge the AI world with incredible new approaches that could be enormous game changers to how we all view AI.

But frankly, absent that ... I'll still take the cat!

- - -

L