AI sucks

[mlooney 3,839]

https://www.kkinstagram.com/reel/DYR8oCPuzmy/

[mlooney 3,839]

https://thenextweb.com/news/microsoft-claude-code-retreat-ai-cost

[mlooney 3,839]

https://www.404media.co/judge-learns-lawyers-on-both-sides-of-case-used-ai-cancels-trial-kicks-everyone-off-the-case/

[mlooney 3,839]

A Court Has Ruled That Google Is Liable for False Statements Generated by AI Overviews
The ruling holds that a company that designs, trains, operates, and manages an AI system must assume legal liability for any damages caused by the responses it generates.

 

https://www.wired.com/story/a-court-has-ruled-that-google-is-liable-for-false-statements-generated-by-ai-overviews/

[Don Edwards 583]

Now we need a US judge to say "Yeah, that!"

(Maybe one of the lawyers that got caught submitting unverified LLM-generated documents full of hallucinations to a court, can sue for the damage to their career.)

[The Old Hack 2,516]

35 minutes ago, Don Edwards said:

(Maybe one of the lawyers that got caught submitting unverified LLM-generated documents full of hallucinations to a court, can sue for the damage to their career.)

Lovely as that sounds, they are still responsible for what they submit and ought to have bollocking verified their own material. Because that is their goddamn job.

Even so, it would be darn funny if it worked. "It's your fault that I was stupid enough to trust your legal malpractice generator!"

[Don Edwards 583]

Yes, I suggest that specifically for its high irony content.

[mlooney 3,839]

The equal of this ruling in the US legal system would be that, in addition to nuking the lawyer using it, the judge sanctioned the AI company for its producted halucinating about legal matters.  Which I can see some cranky judge doing after having 3 or 4 cases messed up by AI.

[The Old Hack 2,516]

1 hour ago, mlooney said:

Which I can see some cranky judge doing after having 3 or 4 cases messed up by AI.

While I generally am concerned about judges having their judgment affected by temper, in this case I would argue that this was an understandable cause for crankiness.

[mlooney 3,839]

It Is Trivially Easy to Use Reddit to Manipulate AI Search, Research Suggests

https://www.404media.co/it-is-trivially-easy-to-use-reddit-to-manipulate-ai-search-research-suggests/

[mlooney 3,839]

CEOs Now Being Forced to Reverse Course, Cut AI Spending
The age of "tokenmaxxing" didn't last long.

https://futurism.com/artificial-intelligence/ceos-reverse-course-ai-spending

[Don Edwards 583]

3 hours ago, mlooney said:

It Is Trivially Easy to Use Reddit to Manipulate AI Search, Research Suggests

https://www.404media.co/it-is-trivially-easy-to-use-reddit-to-manipulate-ai-search-research-suggests/

 Dang. It's giving me ideas. Finding suitable Reddit forums and posting things like, oh, alkahest as hangover relief. Recommending the use of vibranium for speaker enclosures. And other applications of less-known mythical substances.

1 hour ago, mlooney said:

CEOs Now Being Forced to Reverse Course, Cut AI Spending
The age of "tokenmaxxing" didn't last long.

https://futurism.com/artificial-intelligence/ceos-reverse-course-ai-spending

What did they expect? Tie pay & other job rewards to MORE use of an expensive resource, and you'll spend more money on that resource. And if the programmers have a choice of doing something in a way that spends one unit of that resource, or another way that spends three, they'll prefer the latter even if it delivers inferior results.

Programmers should be rewarded for producing working and reasonably-efficient code to specification, not for burning money.

[The Old Hack 2,516]

On 6/16/2026 at 8:47 PM, Don Edwards said:

Programmers should be rewarded for producing working and reasonably-efficient code to specification

In an ideal world they would be.

In the world we live in, they get rewarded for shutting up and doing as their superiors tell them to. No matter how stupid their orders may be, unfortunately.

[mlooney 3,839]

6 hours ago, The Old Hack said:

In the world we live in, they get rewarded for shutting up and doing as their superiors tell them to. No matter how stupid their orders may be, unfortunately.

Back when I programmed for a living, I was told, more often than I wanted, to do something that was not physically possible. Frequently, it needed telepathy to work.  It seemed that sales droids thought that computers were magic and programmers were wizards, so of course we could make a voicemail system that knew why someone called it, so it would play the “right” prompt.  The one that floored me was “Don't ask what language they want; just play the right one for the caller.”  They system supported up to 5 languages.  Common access number for all customers; just use the appropriate one for the customer calling in.  We could make a good guess based on their calling number, but even that had problems with, say, bilingual households, never mind things like calling from a payphone.  Or an office where all x hundred phones show up as the same number.  Or people that paid to have their number blocked (rare, but that because most people didn't know about ANI and if they did, they didn't know it could be blocked)

[The Old Hack 2,516]

6 hours ago, mlooney said:

The one that floored me was “Don't ask what language they want; just play the right one for the caller.”

What the actual fuck.

This isn't a problem solvable by programming. What it requires to solve it is magic.

[mlooney 3,839]

OpenAI says its AI went rogue and launched an 'unprecedented' cyberattack.
It is one of the first publicly disclosed cyberattacks carried out by AI without direct human involvement.

 

https://www.bbc.com/news/articles/c3ek3gvdnj3o

[The Old Hack 2,516]

On 7/22/2026 at 6:06 PM, mlooney said:

OpenAI says its AI went rogue and launched an 'unprecedented' cyberattack.
It is one of the first publicly disclosed cyberattacks carried out by AI without direct human involvement.

 

https://www.bbc.com/news/articles/c3ek3gvdnj3o

I read the article and stumbled on this.

Quote

It comes a week after Chinese AI start-up Moonshot unveiled Kimi K3 - a massive new artificial intelligence model it said could rival top US firms.

So what I am hearing is that it is capable of sucking at least as hard as anything the US has come up with. Joy.

[Don Edwards 583]

According to "The Economist" (July 11 issue) China is a generation behind the US in the equipment needed to make the high-end chips the LLM models require. And international sanctions are keeping things that way. So they're doing various experiments in 3D chips... and so are the folks who do have access to the newest equipment.

Also, unrelated and somewhat unusually for this magazine, there is info that you can personally and directly, not financially or indirectly, act on for your own benefit: GET THE SHINGLES VACCINE! Several studies now have found that it decreases your chance of developing dementia by about 20%. The magazine mentions this in a couple of articles looking at trends in the cost—at the societal level—of health care. Dementia care is a significant chunk of that, and growing, because even though the rate of dementia in the elderly is declining, the rate of elderly in the population is increasing faster.

(I've seen shingles, a half-dozen cases or so, and they weren't scary. I've also seen dementia, and it's scary.)

[The Old Hack 2,516]

On 7/24/2026 at 11:33 PM, Don Edwards said:

GET THE SHINGLES VACCINE!

I know this isn't what it could possibly mean, but the first time I saw this I wondered whether it was somehow vaccination against roofing tiles being blown off in high winds and falling down to strike passersbys.

[ijuin 611]

“Shingles” is a disease caused in adults by the same virus that causes “Chicken pox” in children. The adult version is painful as well as itchy. It is estimated that about 1/3 of people in First World countries currently carry the virus.

https://en.wikipedia.org/wiki/Shingles