Geek Stuff category archive

An automated parrot is still a parrot.

It does not think. It merely parrots.

What big tech calls “AI” is indeed automated.

But it is not intelligent.

It is a parrot dressed up in Sunday go-to-meeting clothes.

Addendum:

At Psychology Today Blogs, John Nosta notes that “GPT models appear intelligent but fundamentally rely on pattern recognition from extensive training data.” Follow the link for details.

Methinks Atrios has a point nailed it.

Getting reporters to call the type of interactions people have been having with Alexa for years, even if marginally improved, “AI,” has been a neat trick.

His entire post is worth your while.

The operating system of the VPS (virtual private server) provided by my hosting provider has reached EOL (end of life). Sometime soon, my hosting provider will migrate this site to a new, currently supported OS.

This site may be unavailable for a bit, hopefully for no more than a few hours. But it will not go away.

Update 2023-12-23:

I have been notified that migration is scheduled for December 28. Expect little or no posting on that day.

If all goes well, normal (?) activity will resume with minimal interruption.

Read more »

Gwynne Dyer doesn’t seem to believe they hype about AI. A bit from his piece about the recent “AI Safety Summit” at Bletchley Park:

The “large language models” the chatbots are trained on make them expert in choosing the most plausible next word. That may occasionally produce random sentences containing useful new data or ideas, but there is no intellectual activity involved in the process except in the human who recognizes that it is useful.

There is plenty to worry about in how “smarter” computer programs will destroy jobs (now including highly skilled jobs), and also in how easy it has become to manipulate opinion with deep fakes and the like.

Addendum:

Bruce Schneier suggests “ten ways in which AI will change democracy.”

Debian Sid with the Fluxbox window manager. (Sid is officially “Debian Unstable.” It’s more stable than most distros’ “stable.” Indeed, Ubuntu is based on Sid.) Firefox, KeepassXC, and Konsole are shaded. Xclock is in the upper right and GKrellM in the lower right. The wallpaper is from my collection.

Sam and the crew parse the perfidy.

(Wait for Emma’s comment at about the 10:56 mark. It’s a gem.)

Aside:

I submit that this is a classic case of psychological projection. It’s not San Francisco that is suffering from a “mind virus.”

Like Elon Musk cares about San Francisco. Or anything or anyone other than himself.

You too can be assimilated by the Zuckerborg. But it’s likely easier if you are young and impressionable, and not a grumpy cynical old man like me.

I have a Facebook account, but I hardly ever use it any more, ever since i realized that I wan’t using Facebook.

Facebook was using me.

___________

*With apologies to H. P. Lovecraft.

William Pounstone, writing at Psychology Today Blogs, explains why AI lies. Given all the who shot john about AI, it is a valuable read. A nugget:

Why do LLMs (Large Language Models–ed.) often spin convincing but bogus answers (a.k.a. “hallucinations”)? The short answer is that LLMs have no innate conception of truth or falsehood.

Follow the link for context.

And, in more news of the not as smart as they think they are . . . .

I logged into my VPS tonight, as I do several times a week, to do routine maintenance. (Indeed, the last time I had to call my hosting provider’s most excellent tech support, the support tech was mildly surprised to find that my server was quite up-to-date. Having briefly worked in hosting provider tech support, I am not surprised that he was mildly surprised.)

I ran a virus scan (it came up clean as usual), then I ran an software update. The update updated the SQL database engine.

As I commonly do after an update, I went back to the backend of the blog and clicked on a random link as a test and got the dreaded “Error connecting to database” error. But, a moment later, everything was working again. I’m guessing that the database engine update hadn’t quite fully taken effect, but that’s just a guess. Anyway, just for grins and giggles, I restarted the server.

Everything seems to be working okay now.

Aside:

I’ve been using the same hosting provider for over a decade, ever since I stopped self-hosting. I went with it because someone I found trustworthy recommended it. I haven’t looked elsewhere because of the quality of its support techs. Fortunately, I don’t need them often, but they have never let me down.

Its support techs set the example.

You think they’re your “friends” on “social” media until you find out that they aren’t.

A 22-year-old man is accused of using Snapchat’s location-sharing feature to try to kidnap a woman and another person on the Blue Ridge Parkway in North Carolina, according to an indictment filed in federal court, WSOC reported.

The good news is that you can turn that “feature” off.

Much more at the link.

Aside:

I keep “location services” turned off on my cell phone unless I have a positive need for them, which is almost never, because I know how to read a map. Remember maps?

Big Tech doesn’t need to know when I take the cats to the vet.

I watched an episode of Monarch of the Glen tonight on Tubi.

Susan Hampshire, as Molly, did web search using a search engine named “Ogle” (I’m sure for legitimate naming issues).

Methinks that the pseudonym may be more accurate than the actual nym.

At Psychology Today Blogs, William Poundstone looks at the current hoopla over AI and puts it in context by recalling the Turing test and, later, a computer program called ELIZA, which was capable of carrying on limited conversations via text.

One passage in particular caught my eye. Poundstone cites a comment by MIT computer scientist Joseph Weizenbaum, the creator of ELIZA, reacting to how willing persons were to think of ELIZA as sentient:

Yet Weizenbaum was astounded at how many people found its canned statements convincing. “What I had not realized,” he wrote, “is that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people.”

One wonders what Weizenbaum might say today.

Follow the link for context.

If you are whelmed by the hyperbole about AI Chatbots and Large Language Models, I commend the segment on AI from the October 15, 2023 episode of Harry Shearer’s Le Show.

In it, Dr. Gary Marcus–er–annotates the recent discussion between Scott Kelley and Geoffrey Hinton, which took place on 60 Minutes.

I think you will find it enlightening.

Coincidentally, I participated in a Zoom meeting last night. It was quite good fun, thank you very much.