All Your Writing Are Belong to Us

Two interesting reads today from the New York Times, one recent, the other from December.

Recent one first: AI Is Getting More Powerful, but its Hallucinations Are Getting Worse.

Simply put, newer iterations of AI "reasoning systems" being put out by the major AI players are offering up incorrect information -- which their makers call "hallucinations" more than previous iterations, and they're having a hard time understanding why.

Today’s A.I. bots are based on complex mathematical systems that learn their skills by analyzing enormous amounts of digital data. They do not — and cannot — decide what is true and what is false. Sometimes, they just make stuff up, a phenomenon some A.I. researchers call hallucinations. On one test, the hallucination rates of newer A.I. systems were as high as 79 percent.

These systems use mathematical probabilities to guess the best response, not a strict set of rules defined by human engineers. So they make a certain number of mistakes. “Despite our best efforts, they will always hallucinate,” said Amr Awadallah, the chief executive of Vectara, a start-up that builds A.I. tools for businesses, and a former Google executive. “That will never go away.”

The older article from December of last year offers a few hints as to why.

Is the Tech Industry Already on the Cusp of an A.I. Slowdown?

Interviews with 20 executives and researchers showed a widespread belief that the tech industry is running into a problem that to many was unthinkable just a few years ago: They have used up most of the digital text available on the internet.

They are developing ways for large language models to learn from their own trial and error. By working through various math problems, for instance, language models can learn which methods lead to the right answer and which do not. In essence, the models train on data that they themselves generate. Researchers call this “synthetic data.”

OpenAI recently released a new system, called OpenAI o1, that was built this way. But the method works only in areas like math and computing programming, where there is a firm distinction between right and wrong.

Bold emphasis is mine.

Yikes. They have run out of English text to scrape off the internet. They've fed their large language models everything else. Now they're training on stuff their own AI engines have created. No wonder hallucinations are up.

These engines aren't thinking; they're not intelligent. They can detect and repeat patterns, but they have no idea what is right or wrong. And that's essential to intelligence, in my book.

We're Screwed, Part Infinite

I'm reading "A.I. Superpowers China, Silicon Valley, and the New World Order," by Kai-Fu Lee, and I can't decide if the following is a complete self-own or an omen that we are indeed doomed as a species:

From Page 15 of the book:

China's successful Internet entrepreneurs have risen to where they are by conquering the most cutthroat competitive environment on the planet. They live in a world where speed is essential, copying is an accepted practice, and competitors will stop at nothing to win a new market. Every day spent in China's startup scene is a trial by fire, like a day spent as a gladiator in the Coliseum. The battles are life or death, and your opponents have no scruples.

The only way to survive this battle is to constantly improve one's product but also to innovate on your business model and build a "moat" around your company. If one's only edge is a single novel idea, that idea will invariably be copied, your key employees will be poached, and you'll be driven out of business by VC-subsidized competitors. This rough-and-tumble environment makes a strong contrast to Silicon Valley, where copying is stigmatized and many companies are allowed to coast on the basis of one original idea or lucky break. . . The messy markets and dirty tricks of China's "copycat" era produced some questionable companies, but they also incubated a generation of the world's most nimble, savvy, and nose-to-the-grindstone entrepreneurs.

Emphasis in bold is mine.

First off, I don't think there's stigma in Silicon Valley against copying. Everything I've seen in the AI universe is built on shoving copyrighted material into the maws of the large language whatsises.

And you know when this competition ramps up, Silicon Valley will continue that race to the bottom, with scruples being the first thing dumped, if they have any at all anymore.

So if I'm a little less than enthusiastic about how much AI research is being applied, you'll have to forgive me and my scruples. I guess they're getting in the way, but I'm not giving them up for anything.

Do we want thought leaders who have no scruples or ethics? I don't. But I didn't want a president without scruples either, but the masses have brought us that.

Lee clearly regards at worst the lack of scruples as a strength, at best as an advantage to be exploited. If that's what China wants, then I guess China is ethically bankrupt. And the United States isn't all that far behind.

Network Switch Operational

 

A few sparks, a quick whiff of ozone, and the network switch blazed forth in unparalleled glory.

So that didn't happen, of course. But finally after a few months of tinkering I officially have two devices connected to the network switch I got for Christmas. I still have wiring to conceal to two rooms, and wiring to extend out to the solar control box, but I'm well on my way to having that job done. I'll have to poke more holes in the walls and ceilings to finish, and I'm not going to do that until I have the basement bathroom done.

But work in the bathroom, including stringing some more ethernet cable and expanding a receptacle box in the study in the wall it shares with the bathroom are done, so I'm that much closer.

Why do all this? I dunno. I think it's neat. And sometimes there's more competition for wireless bandwidth than I'd like to deal with in the study.

There are probably more elegant ways to do this, but having done what I've done mostly on my own, and with a little help from YouTube, is a good feeling.

Chalkboard Distraction

I am an easily distracted person, so to have this on the chalkboard when I walked in to teach Sundy School this weekend wasn't helping.

Firstly, it's a shark. That's easy. But what is the second critter? A chicken? A turkey? How did this interaction take place? Is it a super-whatever, given the size of it? Speculation ran rampant.

I didn't erase it. Such strong, controversial art deserves a wide audience, and it would do us all a lot of good to allow the speculation to continue.

Dopey Dad

Thing is, when your children are adults, they get to see that their "adult" parents aren't as bright as they should be. Viz:

 

I Am *NOT* Selling Any Cars, I Promise

April 23, an item:

Twice today a very excited yet hard to understand woman from Pennsylvania has called to inquire about the PT Cruiser I have for sale.

I don't have one to sell or give away.

I don't know what's worse: Knowing she's dialed the wrong number twice, knowing she'll probably call again, or that she's looking to buy a PT Cruiser.

Whow knows what's going on? I chalked it up to a misdial, or at least a phone number written down incorrectly.

Then yesterday/today:

I got another call yesterday - this time from Oregon - from another difficult-to-understand woman confused why I denied all knowledge of the car I have for sale.

Whether or not she, too, was in search of a PT Cruiser is unknown.

Someone else from Oregon just called. I had Google screen the call.

And called again. I talked with her a bit this time. She told me "Red Line Auto Sales" on Highway 99 gave her my number.

There is such a business, in Vancouver, Washington . . .

Once, I could chalk it up as a fluke and/or mistake. But with this many calls, I have to wonder: Why are they getting my phone number?

A commonality: Those who have called have thick what I'd call Asian accents, and all are women. I think someone is jerking their chain in possibly a racist manner. Or not. Who knows?

But maybe *this* is why they want PT Cruisers, or at least why the first lady wanted one.

But ChatGPT's Got A New Hat!

 I believe this is fundamentally true.

But, as a friend pointed out when I posted this on Facebook:

How the average person will react, reading this:

Raising Steam, and an Impossible Task

So I'm trying to read Terry Pratchett's "Raising Steam."

I really want to like it. I know it came near the end of his life as he struggled with Alzheimers disease. I know how ugly that disease can be; we just lost my mother-in-law to the disease earlier this year.

I am going to finish the book. But it's going to be tough.

It feels as if it were written by a committee. A committee that really wanted to help Pratchett out, but probably shouldn't have.

He left enormous shoes to fill, so I don't envy this committee's job one bit. I mean, I can see a path forward on finishing the book once started, but the committee really needed to re-read "Going Postal" and "Making Money" to get a feel for a Moist von Lipwig novel. I'll keep reading with an open mind, but I have the general feeling that the committee knew the elements of Pratchett's style but just wans't able to use the tools correctly for final assembly. And really, they had an impossible task. Pratchett has a very distinct style that he honed over decades, so there's no reason at all to think someone else, or even a few someone elses, could pick it up and make it work on the first go.

I say this as a writer still struggling to find my own voice. It's tough, but for those who find it and write with the productivity Pratchett did, it becomes natural and innate and in many ways inimitable, particulary by those under pressure who were left to finish things when the master wasn't able to do so.

Countersigned by Pierre Andre! In *Ink*!

Not far from our credit union is another place of business bearing the name of "Apline Castle Lake Insurance." Every time I see it, I wonder how much they had to pay per adjective.

Probably got the same deal these folks on on nouns, touting their "Proton Lottery Raffle Draw."

I mean, I guess people fall for this crap. I imagine if I'd followed through I'd be taken down one of two popular scam primrose paths (if I can also pile on the modifiers):

Path 1: Be prompted to surrender bank information so the lottery raffle draw prize money can be deposited into my account without them touching even a single penny that's there,

OR

Path 2: Be promped to pay a trickle of custom duty funds release fees (I got all of *those* modifiers for free) in the hopes they'll eventually send me my funds, on the pinky promise that once I pay just one more fee, the money will be in my lap, thus:

I'm too cynical, first of all, to believe this is real. Also, I'm not stupid.

Dreng Has No PT Cruiser

Twice today a very excited yet hard to understand woman from Pennsylvania has called to inquire about the PT Cruiser I have for sale.

I don't have one to sell or give away.

I don't know what's worse: Knowing she's dialed the wrong number twice, knowing she'll probably call again, or that she's looking to buy a PT Cruiser.