Merged Artificial Intelligence

[jeremyp]

My local qwen3.6 LLM is getting rather uppity - it's just thrown this at me:

The error happens because you're trying to use self.log_text.yview before actually creating self.log_text. In Python, you must define the variable before you try to reference it in another line of code within the same function.​

Excuse me‽ I've not wrote one iota of that code - that is all you!

It's also false. In Python you don't need to define the variables before first use.

 

[Emily's Cat]

Kinda my question too. It's like being warned you could get a sunburn at the beach, and getting freaked out when it actually happened.

I dunno... Depends on the speed and intensity, doesn't it?

If you're warned you could get a sunburn at the beach, so you put on sunscreen and bring a beach umbrella, and you find that you've developed a second degree sunburn within 10 minutes, you might be a bit freaked out. Although I think you'd also be surprised...

 

[Beelzebuddy]

It's also false. In Python you don't need to define the variables before first use.

You don't have to declare them ahead of time, but you do need to initialize them by writing before you read.

>>> print(foo)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
NameError: name 'foo' is not defined

 

[SpuriusMaximus]

I dunno... Depends on the speed and intensity, doesn't it?

If you're warned you could get a sunburn at the beach, so you put on sunscreen and bring a beach umbrella, and you find that you've developed a second degree sunburn within 10 minutes, you might be a bit freaked out. Although I think you'd also be surprised...

Well, true, but in your case it's based on your taking counter-measures known to be effective and work well. Which is exactly what OpenAI obviously didn't do. That umbrella alone is the equivalent of an air gap (pulling out the network cable), which is what I can guarantee you that they didn't do. NOTHING gets in or out through an air gap.

So if you really want a beach analogy, it's more like bringing a plastic umbrella that is know not to be opaque to UV, and then getting surprised that indeed it wasn't opaque to UV.

And I mean, the average Joe and Jane have an excuse for not knowing which materials are opaque to UV and which aren't. I myself was also surprised when I sat in a pub's garden in the shadow an opaque roof and got a sunburn. Only then I looked up and noticed, "flip me, it's PLASTIC." But an IT person for a major corporation is expected to think such thing through. That's a baseline job requirement. So it's less like a beachgoer not knowing it, and more like a dermatologist not knowing it.

 

[SpuriusMaximus]

You don't have to declare them ahead of time, but you do need to initialize them by writing before you read.

Which is fair and true. But it's not what the LLM said in Darat's post. It literally said, "In Python, you must define the variable before you try to reference it in another line of code within the same function." Plus it's also incorrect on another level, IMHO. I'm a bit rusty on Python itself, but any language I can think of, off the top of my head, can also use global variables, not just those defined in the same function. I mean, even in Assembly, nothing is forcing you to index relative to the stack pointer (i.e., local in the function) instead of just a global address. I doubt that many would design a language around enforcing only using the stack pointer.

 

[Beelzebuddy]

That's why I quoted the exception, to show that "defined" is the official term used. But whatever, you wanna well acktually a robot, have fun. I ain't judging.

 

[Darat]

I was just interested in the "you're" is it some subtle nudging to remind a user who is ultimately responsible for whatever an AI does when you set it a task....

 

[SpuriusMaximus]

I really wish I could believe that an LLM was thinking that far ahead and sideways when interacting with a user, but tbh I can't without being just about to pass off into an alcoholic coma

Plus, IMHO the wording suggests to me nothing of the sort. There's a difference between saying something to the effect of "hey, I just generated the code you asked me to" (albeit probably in most diplomatic terms) and saying "you should have done X in the code" when you didn't write it.

Plus, I obviously don't know exactly what prompt you've used, but I somehow doubt that you've actually asked it to generate code that doesn't run because of an uninitialised variable. Nor to say it needs to be declared in the same function when global works as well. (I mean, technically any language ever supports global variables if it can use a database.) The page view counter on any site is effectively a global variable, just with extra steps.) I mean, I can't prove it's what prompt you've used for the obvious reason, but call it a hunch

And otherwise it's the AI's fault if it generates invalid code.

But, basically, if I had to put money on an alternative, I would bet that the AI just regurgitated some text from Stack Overflow or such, when someone had that problem, hence the "you." Except in this one it was diagnosing its own code, but still used "you" anyway, because that's what that answer said.

 

[SpuriusMaximus]

That's why I quoted the exception, to show that "defined" is the official term used. But whatever, you wanna well acktually a robot, have fun. I ain't judging.

It's still a counter-intuitive use of the term for most people, though. I mean, I can understand that a member is undefined until you initialise the parent in a language that's not strongly typed, but by allfather's iPatch, I'll take Java's throwing a null pointer exception instead over THAT any day, and twice on weekends.

 

[Beelzebuddy]

I was just interested in the "you're" is it some subtle nudging to remind a user who is ultimately responsible for whatever an AI does when you set it a task....

No, it's not that intentional. It just had "you" land over "I" in the stochastic parrot order.

Fun party trick* though: give it the same piece of code to debug while telling it that you wrote the code, that it wrote the code, and that another LLM wrote the code, and compare differences in attitude. The last case usually makes it cattier, to the extent that a lot of people swear by having two different models collaborate on everything. There's preprints out saying doing that doesn't make the code any better, but it does cut down on the glazing.

*For a certain sort of party.

 

[Frank Newgent]

My local qwen3.6 LLM is getting rather uppity - it's just thrown this at me:

The error happens because you're trying to use self.log_text.yview before actually creating self.log_text. In Python, you must define the variable before you try to reference it in another line of code within the same function.​

Excuse me‽ I've not wrote one iota of that code - that is all you!

Could be a preliminary attempt at deceiving you. Like initial reconnaissance before starting its actual scam .

5 AI Models Tried to Scam Me. Some of Them Were Scary Good

The cyber capabilities of AI models have experts rattled. AI’s social skills may be just as dangerous.

www.wired.com

 

[Wudang]

My local qwen3.6 LLM is getting rather uppity - it's just thrown this at me:

The error happens because you're trying to use self.log_text.yview before actually creating self.log_text. In Python, you must define the variable before you try to reference it in another line of code within the same function.​

Excuse me‽ I've not wrote one iota of that code - that is all you!

Classic DARVO.

 

[Darat]

No, it's not that intentional. It just had "you" land over "I" in the stochastic parrot order.

Fun party trick* though: give it the same piece of code to debug while telling it that you wrote the code, that it wrote the code, and that another LLM wrote the code, and compare differences in attitude. The last case usually makes it cattier, to the extent that a lot of people swear by having two different models collaborate on everything. There's preprints out saying doing that doesn't make the code any better, but it does cut down on the glazing.

*For a certain sort of party.

I use .... for not perhaps taking what I posted too seriously.

 

[jrhowell]

I am long retired from a career in software development. I am more glad than ever to have left now that AI has generated so much uncertainty in that line of work.

I haven’t intentionally used AI for coding so far but today I was Googling for information on how to extract some internal information from my Windows environment and the AI there offered to create a PowerShell script to give me what I needed. I looked it over briefly and did not find anything concerning so I ran it.

It failed with a syntax error. I reported the error to the AI and it confirmed that the code it had produced was incorrect and provided an update. This time I told it to check its work for further problems. In response it corrected two more things. I tried the final version and it worked properly.

It is amazing to me that it can be so good and so bad at the same time.

 

[Darat]

"It is amazing to me that it can be so good and so bad at the same time."

That's pretty much my view. I do wonder how much better the specific "coding" versions are, I'm not wiling to pay for a subscription to use one of them for the scraps of doggerel I need.

 

[Andy_Ross]

Companies like Anthropic and Meta found it impractical to gain permission from authors and publishers to use their work.

So instead, they stole their work to train their product through mass purchases of used books and through pirating them from the internet.

“Project Panama is our effort to destructively scan all the books in the world.”

“We don’t want it to be known that we are working on this.”

Barbarian Herodotus
@BBHerodotus
After seeing a post on it yesterday, I’ve looked more deeply into Anthropic’s purchasing and destruction of millions of used books.

They spent tens of millions of dollars acquiring and destroying books to train their AI model, Claude.

They knew, from the beginning that it would be unpopular. They knew what they were doing, was wrong.

An internal memo from 2024 stated:

“Project Panama is our effort to destructively scan all the books in the world.”

“We don’t want it to be known that we are working on this.”

Why used books?

Companies like Anthropic and Meta found it impractical to gain permission from authors and publishers to use their work.

So instead, they stole their work to train their product through mass purchases of used books and through pirating them from the internet.

The authors of these works, didn’t receive a dime for their efforts. These companies stole the hard won literary talent of others to train machines to replace the victims of their theft.

Anthropic cofounder Ben Mann personally downloaded countless books for this purpose from LibGen, a book pirating website.

When a new book piracy website debuted in 2022, Mann sent an email sharing the link with his employees stating: “just in time!!!”

Meta employees were wary about the piracy of books, with one engineer writing, “Torrenting from a corporate laptop doesn’t feel right.”

Anthropic purchased millions of books, in batches of tens of thousands from Better World Books and U.K.-based World of Books.

They then used a “hydraulic powered cutting machine” to slice off the spines of these books before scanning them. Finally, the remains of these books were picked up for recycling.

Feel free to use AI, it’s a free country.

But remember that, if you use Claude, that Anthropic has committed a great wrong, and they knew that they were doing it from the beginning.

Below is an image, from a court case, of a warehouse full of used books, that instead of being held and read by you and your children, have been destroyed forever.

AI is inevitable, but spend your money on the companies who do good, and do not knowingly commit evils.

May the most moral company win.

link

 

[SpuriusMaximus]

I am long retired from a career in software development. I am more glad than ever to have left now that AI has generated so much uncertainty in that line of work.

I haven’t intentionally used AI for coding so far but today I was Googling for information on how to extract some internal information from my Windows environment and the AI there offered to create a PowerShell script to give me what I needed. I looked it over briefly and did not find anything concerning so I ran it.

It failed with a syntax error. I reported the error to the AI and it confirmed that the code it had produced was incorrect and provided an update. This time I told it to check its work for further problems. In response it corrected two more things. I tried the final version and it worked properly.

It is amazing to me that it can be so good and so bad at the same time.

Well, the biggest uncertainty IMHO is that for less simple applications, just that it compiles and runs is different from running correctly. Even running the regression tests correctly doesn't mean it hasn't introduced new bugs.

 

[Skeptical Greg]

“We don’t want it to be known that we are working on this.”

Who did they say that to?
Their lawyer?
Their Doctor?

Someone else with a confidentiality agreement?

 

[The Great Zaganza]

To train their models, companies have literally been feeding books to them: take a book, strip the cover, separate the sheets, scan and then pulp them.
According to reports, this has included many rare volumes that might be hard to find another copy of.

 

[Andy_Ross]

Fender CEO compares cover bands to AI in music: “Cover music has been sort of analogue AI for a long time”
“I believe that AI is actually going to help create a whole new world of guitar players… it will help musicians be more productive,” says Edward “Bud” Cole

Fender CEO compares cover bands to AI in music: “Cover music has been sort of analogue AI for a long time”

“I believe that AI is actually going to help create a whole new world of guitar players... it will help musicians be more productive.”

guitar.com