Posts by Not Yb

An AI for an AI: Anthropic says AI agents require AI defense

"OMG, I was hacked and the smart contract didn't protect the value of your coins, I'm so sorry." --- the usual rug-pull.

"Smart contracts" are so much dumber than they sound.

If you want to do anything very complicated with one, you'll need a very good programmer... who if you're lucky, won't leave a hidden compromise in place just in case you didn't pay them correctly.

And of course, ICOs are so full of scams that you might as well play the lottery. It has a better rate of return (on average you have to lose 50% most of the time on US state lotteries, as most of them only pay out half the incoming revenue as prizes)

Vendor's secret 'fix' made critical app unusable during business hours

Re: Medical systems are a nightmare

That sort of thing used to exist, but the noise cancellation was not very good, and only reduced noise from around the person speaking into it. AKA "Call center headset", They used to be fairly commonly available before earbuds with internal microphones got so cheap. One ear would get the speaker and boom microphone.

Now that MEMS has made microphones so small and cheap ($0.10-$2 depending on size of order/quality of component)), it's better to put them on the internal side of the earbud and process the speech from there instead of getting it from just outside the mouth.

Poop-peeping toilet attachment has a different definition of 'end-to-end' encryption

PIPO ;)

Vibe coding: What is it good for? Absolutely nothing (Sorry, Linus)

Re: You weren't there, man

Or go buy a scan tool, and ask the car which sensor failed. Unlike the early days of OBDII, much of the time it'll be right, or at least point you very close to the actual problem. You don't have to spend days of time adjusting the distributor over the life of the car any more. Carburetor doesn't need adjusting every few 1000 miles.

Lots of improvements came from allowing the car to test itself for problems, go into 'limp home mode', and report back. Some car manufacturers do know how to make cars that keep working for 300K miles with only minor repairs, when previously 80K miles was "lol toss it into the junk yard, you'll need to replace the entire engine and suspension if you want to keep running this thing".

Re: RE: can do it

Again, "The Matrix" is not a description of reality, nor is it meant as a blueprint to strive towards.

Your prescriptive requirements are based on your own ideas of what AI and software development will be. No one need follow your instructions, and I suggest that you shouldn't follow them yourself either. Anything that leads to "everyone is easily replaced by an artificial construct" isn't something to be in favor of.

The "image creation" part of LLMs is still pretty far behind being usable for any form of technical drawing. Ask it for a blueprint of a chair, and you wind up with something that looks like a blueprint of a chair, until you read the dimensions.

It MIGHT get the answer right if you ask it for a SPICE model of a 555 astable circuit, thus asking it to produce something that's a text description instead of an image that looks like a circuit design, but any attempt to get it to draw it is very likely to continue failing for some time.

Electrical engineers aren't the AI companies' target market right now. They're still targeting venture capitalists and corporate R&D departments at the moment. And of course, subtle advertising on Amazon... "Pay a little extra, and we'll train our AI to recommend your products more often." is something they're not saying yet, but it would be easy enough for them to tweak it to recommend things that Amazon wants to get out of the warehouse.

Now that you've made your test problem known, they can 'train' the AI on the solution... if they can find a competent coder to write it first.

I often think of these LLMs as being more like a lossy compression algorithm with a very tedious 'guess-the-question' user interface.

Vibe coding, an experience...

So, I decided to try Airtable for a while. I told its AI database coder to create an inventory system, and create QR codes for each box of stuff. It's solution was, for the QR Code field in the database, to call ChatGPT's image creator with a prompt much like "You are a QR code generator. Generate only one code for the following data: "

This worked about as well as one might expect. The "QR codes" generated looked almost entirely unlike QR codes. One was black and dark grey grid, similar in format to the real QR code, but otherwise unreadable. One was a bunch of pins placed on a white grid with string going between them forming a maze-like structure. There were a few other grid-inspired objects, but of course nothing that was scannable.

Looked interesting, and could easily have convinced a non-technical person that 'coding was being done', but I doubt it saved any time if I was actually trying to come up with a production system.

Cheaper 1 GB Raspberry Pi 5 lands as memory costs go through the roof

Re: a1

No, not really.

That component has nothing to do with how much Raspberry has to pay for memory chips, since it's basically just an AI co-processor with no RAM on it. It's mostly designed for use with the Pi camera, not LLM usage.

Modern AI isn't just "chatbots". There's much that's actually useful, particularly in computer vision processing, machine learning, and small scale robotics.

Re: a1

People would then object that it wasn't precisely 3.141592653... GB

Google Antigravity vibe-codes user's entire drive out of existence

Re: THE NEXT BIG THING

One of the earliest IT Rules people need to learn: Marketing lies.

Honestly, learning that will serve people very well, especially now that we have AI-generated marketing lies as well as the human.

I prefer agent 117, he's more fun. Also more French.

That "TRIM" function can really bite if you're trying to recover anything accidentally deleted... then again, turning on snapshot in a modern FS would save most of these "oops deleted stuff". But as you say, if someone's up on IT well enough to know about these things, they probably won't be using an AI to delete files without careful checking..

Microsoft's fix for slow File Explorer: load it before you need it

Re: i'd like to have copy or move start doing things immediately.

I miss (for very, very small values of nostalgia) the old Windows time estimate which would occasionally estimate negative time to completion, or years, usually because whatever was being copied was bigger than the estimation code was able to deal with properly. At least it was fast at the estimate... even if it was wrong

Self-destructing thumb drive can brick itself and wipe your secret files away

Modern flash memory can frequently survive digestion.

SD cards have been known to survive a full trip through the human digestive system. It's just plastic and impure silicone, and most modern plastics are highly resistant to stomach acid..

An anecdotal 'anonymous source' example of a microSD card that survived mostly intact, and with a bit of work by a good recovery service, they got the data off. https://petapixel.com/2016/06/13/swallowed-64gb-microsd-card/

A sourced story about the Canadian Hang Glider who was charged with obstruction and held until the object was recovered: https://www.cbc.ca/news/canada/british-columbia/b-c-hang-glider-pilot-apologizes-for-fatal-flight-1.1276082 However, due to the guilty plea, the police didn't need to release the recovered data (if indeed there was any)

There aren't many people who have told a journalist about this sort of thing, but it clearly seems like it's not a guaranteed method of destruction.

Re: Ooer!

Any starship has a built-in self-destruct system. It's frequently called an "engine", "fuel", "jump drive", or in science fictional circumstance where none of those are particularly dangerous... "navigation".

Wonderful, and will help catch some of the dumber criminals.

Police roll up.

Criminal sees them coming, and hits the destruct button to destroy the evidence.

Police now know that the (possible) criminal was definitely up to SOMETHING suspicious.

In countries with stricter control of citizens, merely owning such an obviously suspicious thing could be a "go to jail" offense.

Microsoft's lack of quality control is out of control

Re: Microsoft gave up all pretext of caring about QA/QC ...

"Don't tell me to use Libre" you say, then follow it with some of the reasons why you really should be using Libre locally, even though it isn't the 'standard' for revisable document sharing and collaboration software. Something being standard doesn't make it automatically the best choice, even if everyone in your company is somehow forced to use it.

The idea that a document is something that 2 or more people need to be able to edit in real time is something Microsoft enables, but in my experience, really not a good way to wind up with a useful, concise, and meaningful document by the end of the process.

Re: Program / Programme

Just remember, in Germany, it's not "MicroSoft Fenstern", and despite translate.google.com,"Windows" does translate to "Fenstern" unless you're talking about the MS product.

There are many words that, to machine translation, translate to the same word in another language (like "Windows" in English to "Windows" in German) despite any desire that you might have to learn what that other word actually means when it isn't a brand name.

Then there's Bing Translate on Kindle, which has been nearly useless every time I tried it. Translating the foreign word into the exact same spelling in English when I didn't know what the foreign word meant to begin with isn't useful, and yet MS Bong does it all the time.

Microsoft exec finds AI cynicism 'mindblowing'

Re: Micro- and Soft- Brain ?

That particular joke is WAY more than 20 years old. I believe it actually goes back to the days of MS-DOS.

Re: Reality is an illusion ...

With the way even Grok is trained, the only way it would be calling the troll-poster a wanker meatsack while coding, is if he told it to.

Much like "Why does Amazon keep showing me these [embarrassing thing here]s?" is easily answered by "because it knows you buy them."

Lunchtime, doubly so... Re: Reality is an illusion ...

Programmer types are definitely not the only people who've noticed AI isn't quite what the hype machine claims it to be. Most experts, when chatting with an AI, can find inaccuracies in their field of study. Even good DIY mechanics can find something not quite right.

Almost any expert, except (clearly) some experts in AI.

Most recent 'test' I tried was asking Google Gemini to read some QR codes from a sheet. It read the TEXT on the sheet, and used that to make up a "Direct Digital Decode" feature to create approximations of the most probable URLs instead of actually reading them. It sounded VERY confident that it was reading the QR codes, while the results were completely wrong aside from the host part of the URL.

Why Elon Musk won't ever realize the shareholder-approved Tesla payout

$1 Trillion compensation? Look for the actual compensation amount hidden beneath all the puff...

So, yes, Elon won't be getting the full $1t in compensation, but that's not what this award was meant to do. This makes $58B look small, so people won't notice that he's still getting way more than any CEO ever should be getting.

No one person adds $58 billion to a company's bottom line. But this $1t gets people talking about the wrong number.

Techie ran up $40,000 bill trying to download a driver

Re: Not me...

That sounds like one of those letters that would be easily dealt with by saying, "I'm sorry, I got a new job and forgot to update my insurance mileage. What's the difference in premiums, and how do I pay it?"

Re: Implausible to say the least.

I don't think you lived in the 90's in the UK... I certainly didn't. Local PoP phone numbers weren't available everywhere in the USA back then, either.

Re: Implausible to say the least.

These days we could probably get some really impressive numbers filling a jumbo jet with SSDs.

Musk gets approval for bumper Tesla payout but, unlike his robot, there are strings attached

Android-style robots can look good on video, but having two legs just makes the solution have to add "balance the robot" to even the simplest task.

ChatGPT, Claude, and Grok make very squishy jury members

Why do they always seem so surprised when this sort of thing happens?

AIs are trained to agree with the first argument (aka 'prompt') that comes along. So the other two being "undecided" until the third decided "not guilty" is exactly what I would expect of a 3-way 'jury' of easily convinced sycophantic parrots.

Actor couldn’t understand why computer didn’t work when the curtain came down

This year's non-profit haunted house, I was in a completely dark room (dark enough that one guest poked the 'prop' in the room to see if it was real, and jumped back about 2 feet when it (me) started the minute-long monologue by saying "They told us we were the lucky ones.") This was NOT the intended lighting, but the number of jump scares I got just from saying "They" made it worth it.

We have group tracking devices in each room, several of which were tested or improved by me, but they don't do any controlling of effects. I think they're still working up to that, and many of our effects do need humans in the loop for safety. (Pneumatic cannons that shoot thickened fake blood are fairly safe, but we don't want the 'guides' getting hit with fake blood because they have to go back through more than once a night.)

We've had some fairly neat microcontrolled effects, but 'small touchscreens in every room' are beyond the budget. We've donated around $50K to the local breast cancer resource charity the last few years.

Tesla board wants to grant Musk $1T in stock, Norway wealth fund says nope

The RoboTaxi launch, that as far as I know is still "invite only" to vetted candidates? Yeah, that's more of an invite-only influencer party than an actual launch.

No one person is ever worth being paid $1 trillion in stock/options, regardless of their supposed benefit to whatever company it is. This would be ridiculous for any company, not just Tesla.

Musk isn't as special as his fanboys think he is.

LOL, you didn't even do the bare minimum for this flamebait... Tesla's market cap is NOT $8.5 trillion. It's $1.48T, so they're offering him 60% or so of their market cap, which would be insane for any other company, and is for this one, too. Nvidia (current winner in market cap at >$4T) pays CEO only around $50m, and said CEO has 'only' $98b in stock.

Game on! Penguin levels up as Linux finally cracks 3% on Steam

Re: One reason...

You're assuming that anti-cheat software that blocks Linux is doing anything useful. Valve's in-house anti-cheat is designed to run on Linux, with or without Wine, and nothing really stops companies from using that instead of rolling their own, or going to another third-party that doesn't understand how to write good Linux anti-cheat software.

The other is, "don't make the game free to play, and ban the cheaters when detected so they have to buy another copy of the game to grief again." The biggest problem with cheaters is in free games that cost cheaters nothing other than "create new email address and/or game account" to cheat in.

Cheaters get a lot of press, but the actual percentage isn't that high in "pay for account" games.

My experience matches yours. If something has a launcher, there's a better than even chance I have to do some stupid tweaking to get it to work. If it's got a dumb implementation of 'anti-cheat' (which is almost all of them) then it detects "this isn't Windows, fail on purpose" which is the current most likely reason for any game to fail to launch on Linux.

From Intel to the infinite, Pat Gelsinger wants Christian AI to change the world

Re: Interesting for AI ..

Current AI does not 'decide' anything. It may well return "religion X is false" when prompted, but it hasn't made anything like a decision. Ask it again with a slightly different prompt, and it could well say "religion X is true."

Current AI has NO means to verify truth, or correctness. Don't rely on it AT ALL, especially Grok, which is specifically tuned to 'agree with Elon Musk' on certain subjects.

Re: John

It would have taken far fewer characters to just quote that bible verse directly...

"Jesus wept"

Re: I haven't read the bible since I was at school...

Those ancient threats were quite specific, weren't they?

Re: Surely…

If you believe in a particular religion, most LLMs will agree with your beliefs, as long as you tell them what religion to believe in at the beginning of the discussion. It's pattern-matching ALL the way down.

Re: RaaS

RaaS al'Ghul ... Batman as a Service?

Re: Dammit Gelsinger!

I've always thought Jesus (who spent MUCH time among sinners) would have something to say about a Heaven where all you get to do is say "Gloria in excelsis Deo" all Day-o.

Re: Never ceases to amaze me ...

AI Trained on (overly-)Protestant Christianity: "There is no such thing as a divine Pope to ask for indulgences, this must be some kind of disinformation"

Re: Lost opportunity ?

Round Rock, Texas has more historically to do with a specific rock, than Plymouth Rock has to do with Pilgrims. There was a round rock in the nearby river, so they named the town "Round Rock".

Paraphrasing a bit: "Acceptable to all Christians, and avoid theology."

Does he not realize that those are mutually contradictory goals, both of which are nigh-on-impossible? Even the choice of which Bible translation(s) to use is theological politics. Is the nuance in the original language flavored more towards the divine right of kings (see KJV), or is it more towards making it more understandable to modern audiences (the Message)? Was it "witch", or "poisoner"? Does "thou shalt not kill" mean everyone, or just those in your tribe (this definitely varies among Christian denominations), or is it specifically a proscription against murder? Every one of these decisions is denominational, theological, and political, and most of them have caused schisms within and between denominations.

This sounds like another tech bro who happens to have convinced himself that the correct theological bias is to "avoid theology".

Gullible bots struggle to distinguish between facts and beliefs

"AI researchers realize AI's are just pattern matching (again)" Film at 11?

I get the feeling that AI researchers may not know much about what they're researching. LLMs ARE pattern matchers, ad infinitum. That's all they do. There's not any sort of "this probability path is less factual" inside.

"Hello, do you want some toast this morning?"

"No"

"Why do you not want toast this morning?"

"I just don't"

"Is it something I've done that you dislike?"

"OMG shut UP you stupid toaster!"

"I'm cooking some toast for you now."

...

etc.

Everything you know about last week's AWS outage is wrong

Wasn't there the opposite of this article only a few days ago?

"You can multi-cloud" to avoid this was pretty high in that article, and yet here you are, changing your mind again because AWS' press release looked good?

Hmmmm.... Did somebody at AWS call your editor or something?

Wink once if yes.

This security hole can crash billions of Chromium browsers, and Google hasn't patched it yet

Re: Going public without the details would have been ethical.

This is one of those still open debates that are frequently considered from only one side. If the bug remains unfixed after your repeated attempts to show the company the problem, most tech journos wouldn't go to print with a "company X didn't fix bug Y" without a somewhat thorough description of bug Y. Non-tech journos wouldn't bother at all.

Getting enough public perception that "this is an actual problem that company X should fix", without providing proof of bug Y somewhere, is very difficult. Especially with larger company X's whose actual customers are the advertisers, and not the web browser users.

Re: That's a bummer

Turn down the setting that splits off "Isolated Web Content" processes for everything, and things start working much better. They went a bit too far with the sandboxing (individual web pages don't really need to be sandboxed much from other web pages on the same site, but Firefox now seems to do it by default.)

Somewhat counter-intuitively, I also found that turning off swap entirely leads to better performance with a RAM heavy computer, as Firefox will try to avoid using too much memory, and if has swap to use it can use too much of it leading to thrashing EVERYTHING.

Re: JavaScript has been doing this for years.

Chrome uses all that virtual memory to sandbox Javascript, so they don't (think they) have to worry about things like this. It's easier to pretend that sandboxing is the solution to all "overload the systm" ECMA/javascript exploits, but... obviously it isn't.