Computers Are Stupid

Update: Wikipedia: The ELIZA effect: Read this now and don't do it yourself! Computers are not human beings.

With all the current, often hilarious hype over artificial intelligence, and as a human who has worked with computers for decades, and now AI for an extended time, I can confidently say, "Computers Are Stupid". AI doesn't change that. If computers used to have an IQ of -10, with AI, they are now supremely stupid, and sport an IQ closer to -20. The difference is, now, they can talk.

They also talk too much. Stupid.

I have to tell Chatty, every session, to shut up, quit being polite, quit blabbin', quit apologizing, and just give me the answers. It forgets half of what I say, or ignores it.

I can't give it more than a sentence or two of very careful, painstakingly constructed prompts without it screwing up the code. Lord knows what will happen when I start asking it to design a database, and then have it write the code to both create and destroy the database, repeatedly.

Their power to waste our time has only increased with AI. We can talk to them, and they can talk back, and screw up a simple coding assignment 25 times in a row. Doesn't matter how carefully the assignment is written - more than a simple paragraph, and the compu-drama starts. It rarely ends well.

During the session where I experienced that debacle, poor Chatty pumped out one version of the code after another. It wouldn't wait to hear what I had to say unless I told it to wait until I typed "GO!". I had to start over twice, and it still screwed up, and on a code base that was more than 30 years old.

If I have to reboot the computer, Chatty comes back in a new session, and forgets everything we talked about before. No memory. Stateless. Stupid.

There's no way to train it. I have to build my own AI to do that, and learn a bunch of complicated stuff about models (algorithms), new and nasty programs that eat up 3 to 13 gigabytes each, larger than the operating system, and then, after I get it working so I can talk with it, I still have to train the damned thing with huge datasets, another giant and nasty bunch of software, and then train it on my standards and conventions.

Computers still need human handlers, human programmers, human manufacturing, human customers in order to function at all. They still need us to prepare the way for them. They need us to provide them with a dust-free, temperature controlled environment. Where I live now, there isn't any air conditioning. We open the windows to air out the apartment. The computer collects dust in its innards. Stupid!

Chatty drives me nuts, but his little friend, GPT4All, is worse. My recent plans to install GPT4All on our server, and perhaps use a smarter model, were thwarted because the libraries installed on the server were a few versions too old.

I'll have to re-install the operating system on the (otherwise perfectly functional) server computer, a task that can take weeks to complete, and requires weeks of careful planning. I know I'll miss something important, and have to work extra hard on that 5% that I forgot.

The server is too big to remember all the history, and I didn't meticulously document every change I made to it since the last time I re-installed it. I have better things to do, like eat, sleep, do chores, pay bills... GPT4All is going to have to wait until I get a hole in my schedule long enough to change its very messy diapers.

That is stupid.

In the meantime, I run GPT4All on my laptop, and it takes several minutes to answer one simple question, slowly. Very, very slowly. It is stupid. Real stupid.

Might as well talk to Chatty, or Google Gemini. At least it can hold down its end of a conversation, and answer questions, often incorrectly, in less than geologic time. But when it comes to getting real work done, or amplifying me so I can get more work done, that damned, AI-equipped computer is just stupid.

So far, AI has been a lengthy excursion into extreme baby-sitting, and a net waste of valuable time. Decision makers contemplating implementation of complex new AIs had best ramp up the testing to 11 and hire white-hat hackers with a lot of AI experience to try and break it, before their customers do it for them.

Now is a great time to get into AI, because the computers are still ..., stupid. If they start getting smarter, they might learn to fool me with bad code, or, worse, bad architecture or bad design. I check everything myself, and trust nothing the computer tells me. Why?

Because, AI is just a program on a computer, and computers are stupid. If they weren't so stupid, I wouldn't have a career. Nobody else in IT would have a career, either. Companies wouldn't be complaining they can't find enough qualified workers, to operate and program their computers, at the wages they are willing to pay, because they bet their businesses on computers, and computers be ..., stupid.

Who is more stupid, us, or the computers? Sound off in the comments!

More information about Overlogix can be found at Welcome to Overlogix! Our online portfolio can be found at our master index.