Follow Slashdot stories on Twitter

 



Forgot your password?
typodupeerror

Submission + - Microsoft's update also breaks USB scanners (slashdot.org) 2

shanen writes: Not requesting a dupe, but the story died too soon for me to add the data point. I don't use the scanner often enough, eh? Not sure how to revive interest, but I actually think some stories should move down the top page more slowly, and this is one that probably deserved more than the standard one-day lifetime. Not sure when I remembered that some "USB topic" had gone past on Slashdot recently...

So I searched google and for some strange reason the google didn't return any links to the horse's mouth at Microsoft. What?

That provoked me into trying Bing and then Copilot, which produced a rather hilarious and infuriating discussion. Some of it might be amusing here, but I don't need the headache of getting sued by Microsoft if I dared to quote what their AI said. Much of the discussion involved "regression testing" and how little anyone should trust Microsoft. Confessions from the jackass's mouth?

Comment Re:Simple answer [to the complicated question] (Score 1) 162

Do I have to complete the quote? It would be additional evidence. [And the moderators rated that FP "Informative"? Really?]

Also me thinks though dost not followeth the link and readeth the story. The original is on the clickbait-ish side, but the Slashdot summary version much more so. The original starts with videos, and I think the complicated answer is more along those lines. So here's a short summary of my latest thinking on the topic of human intelligence, such as it is.

Long time ago some general purpose computing units evolved. They were built out of neurons. Before the mammals showed up, though I'm not sure where to draw the line. Some of them got incorporated into specialized systems for processing certain types of data such as smells and sounds and images. There was no real design process involved in those times, just lots of trials and lots of errors, but Ma Nature is kind of efficient and the sooner the errors died, the sooner their amino acids were recycled back into food and fertilizer. (Going at that level because I'm currently reading "Amino-San No Himitsu", Volume 222 in the series of "good" manga secrets.)

But recently humans came along and subverted the hardware with strange new software. We couldn't do much with the smell stuff, but we added language programming on the sound system and wound up with stories and then implemented concepts like self-awareness and consciousness on top of that "new OS". Complicated part was integrating some of visual stuff into it, but I think the example of cats can help clarify things. A cat can hear and see a lot of stuff and can learn all sorts of complicated behaviors when it comes to catching birds and mice, but there is no language and therefore no internal stories of the cat itself and no external stories to teach to other cats. But working on the level of the sound system its basically a linear thing, like the tape in Turing's machine. From an evolutionary perspective, strings of sounds made sense, but at the level of good and bad patterns. The steady sound of the wind in the grasses versus the changing sound of an approaching prey or predator.

Then we invented written language and things became really perverse as we subverted the visual system. The evolved hardware did stuff like recognize lines and colors at the low level and recognize more complicated constructs at higher levels. Originally we could recognize the foods and predators and take appropriate actions, but when we recognized words we compressed concepts to a whole new level--but the general purpose neural hardware didn't care that it was solving much more complicated problems at much higher levels. Higher level processing of complex words, not simple corners. Very flexible programming indeed.

Now things have gotten really perverse in bad ways. The kids are now saturating their video channels with garbage like cute cat videos instead of complicated books. And the AIs are evolving by design to handle many kinds of problems that really strained our human hardware. "Simple" example from the LLMs. Humans are basically evolved to handle two to four dimensions at a time, but the AI chips are designed to handle thousands and millions of "dimensions" at the same time. We humans have to shift our levels of abstraction and reference frames to see relevant aspects of various problems, but the AIs just swallow the entire problem whole without even needing "meaning" for most of the dimensions...

Comment Re:In other news.... [But no solutions?] (Score 1) 23

Mostly thanks for the relatively good FP, and it deserved the Interesting mod, but... Insight would have called for deeper consideration of the problem and Slashdot has no mod to encourage solution-oriented thinking. My anecdote below might qualify for interesting, but first a word about the solution approach I still favor...

I think someone (perhaps the police) should be collecting targeting data and trying to find the money. Cutting off the money would stop a lot of the scamming. Many examples, but I'm going to use email because it's the easiest one. In this story's scenario the victim probably has an email message from the criminals. I imagine a website where the email (preferably the full source with headers) would be submitted into a webform and then analyzed (presumably with AI these days) and the donwannabe (don't-want-to-be (but perhaps-already-a)) victim would confirm the details. Might involve several iterations of analysis and confirmations and the victim might even have to check into some details. For example, maybe the victim has forgotten about an earlier phishing email that triggered the attack, but could dig it up with a bit of guidance? Adding that email to the report might provide more clues for the investigation. However the important thing is that such a website could collect lots of data and direct targeting at the biggest criminals and focus on cutting the criminals' money flows.

Now the anecdote. Already a couple of months old, but I'm sure it was an AI GAIvator of a famous person. The email was beautifully written and really had me fooled, even after a couple of exchanges. The style was that good, but the only suspicious part was how quickly the responses came back. Not instantly, because that would have been too obvious, but within a day or two, and I became suspicious that our discussion could have such a high priority for such a busy person. So I checked via the celebrity's website and got confirmation that it was a scam.

Meanwhile the cops are chasing their tails and the criminals continue laughing all the way to the bank. Except when there are crooked cops who are wetting their beaks--but I better not cite that book I'm currently reading. I've already got enough enemies, thank you.

Comment Re:Skipping the fundamentals destroys the future (Score 1) 55

I'm still updating my Tripod page. But you're still failing to motivate curiosity or interest or civility or whatever. You do have a reason to post, right?

Me? I'm certainly not expecting much in the way of intelligent discussion on Slashdot these years. Mostly I'm trying to clarify my own thinking on whatever topic has managed to perturb my ADHD.

Comment Re:only use less gasoline if you actually charge t (Score 1) 112

Productive FP thread though I couldn't find much on the angle I was interested in, the battery capacity of hybrid vehicles versus pure electric. However it seems there should be a fairly direct optimization problem for the engineers lurking around here. Weight of the engine versus the weight of the batteries versus typical driving behaviors and refueling versus recharging times... Sources of the electricity for recharging are also important. My basic guess is that the optimum solution for each driver is mostly controlled by that driver's typical driving pattern.

Slashdot Top Deals

Optimization hinders evolution.

Working...