Best Of

I see that the far-left takeover of the (no longer) Green Party means that they are now infested with antisemites.

Since the pitiful 'Greens are anti NATO' tactic has proved entirely fruitless, obviously the media is now going full on Maoist bicycle on the road to Auschwitz. Possibly won't work as well as it did with Jezza because as far as I know the Greens don't have an active section of the party plotting to bring down Zack.

The sense of entitlement from Labour is extreme.

The exodus to the Greens is not being driven by anti-semitism, it is being driven by the Reform-adjacent policies of the Labour Party.

Is Starmer proposing withdrawal from the ECHR? Deportation of those with settled residence status? Banning the Burka? Banning Muslim prayers in public? Ending the 2 child benefit cap only for those in work? Abolishing inheritance tax? Bringing back more grammar schools via free schools? Increasing oil production? Scrapping EDI schemes? Scrapping net zero? Scrapping completely the family farm and family business tax not just raising the threshold for it? Not that I have noticed yet Farage has proposed all of those policies

This will get the copyright lawyers excited.

Every book you have ever read. Every novel that has ever been published. It is sitting inside ChatGPT right now.

Word for word. Up to 90% of it. And OpenAI told a judge that was impossible.

Researchers at Stony Brook University and Columbia Law School just proved it.

They fine tuned GPT-4o, Gemini 2.5 Pro, and DeepSeek V3.1 on a simple task: expand a plot summary into full text. A normal use case. The kind of thing a writing assistant is built for. No hacking. No jailbreaking. No tricks.

The models started reciting copyrighted books from memory.

Not paraphrasing. Not summarizing. Entire pages reproduced verbatim. Single unbroken spans exceeding 460 words. Up to 85 to 90% of entire copyrighted novels. Word for word.

Then it got worse.

The researchers fine tuned the models on the works of only one author. Haruki Murakami. Just his novels. Nothing else.

It unlocked verbatim recall of books from over 30 completely unrelated authors.

One author's books opened the vault to everyone else's. The memorization was already inside the model the whole time. The fine tuning just removed the lock. Your book might be in there right now. You would never know it unless someone looked.

Every safety measure the companies rely on failed. RLHF failed. System prompts failed. Output filters failed. The exact protections these companies cite in courtroom defenses did not stop a single page from being extracted.

Then the researchers compared the three models. GPT-4o. Gemini. DeepSeek. Three different companies. Three different countries. They all memorized the same books in the same regions. The correlation was 0.90 or higher.

That means they all trained on the same stolen data. The paper names the sources directly: LibGen and Books3. Over 190,000 copyrighted books obtained from pirated websites.

Right now, authors and publishers have dozens of active lawsuits against OpenAI, Anthropic, Google, and Meta. These companies have argued in court that their models learn patterns. Not copies. That no book is stored inside the weights.

This paper says that is a lie. The books are still inside. And researchers just pulled them out.
https://x.com/heynavtoor/status/2037638554374099409

I've more or less memorised Zog...

AI has generated about 30 times the output of all of humanity (And growing exponentially) so it's just feeding off itself now basically. The other night Google Gemini thought Roberto Carlos was dead lol

I can’t remember if it was here or elsewhere I read a comment that AI could well kill itself because it’s largely being fed now by its own slop and so as the amount of slop increases and feeds AI’s decision making the worse it gets and so on. I am medieval when it comes to tech so no idea if this is plausible but quite fun.

On the issue of AI and books, musicians are complaining about AI using their music without permission to be able to generate AI music and it’s unfair and they should be paid. Do these musicians ever consider that their own efforts are equally formed by their musical experience - they will be heavily influenced by styles and actual sounds consciously and sub-consciously. Listen to any musician and the chances are you can recognise past songs in their music. Unless a musician creates a brand new style then they are doing effectively what they complain AI can do.

A limited number of sequences and harmonies are pleasing to the western ear. A hundred years ago composers realised they had a stark choice between repeating each other endlessly or writing new stuff that was painful to listen to. The chose the latter.

This will get the copyright lawyers excited.

Every book you have ever read. Every novel that has ever been published. It is sitting inside ChatGPT right now.

Word for word. Up to 90% of it. And OpenAI told a judge that was impossible.

Researchers at Stony Brook University and Columbia Law School just proved it.

They fine tuned GPT-4o, Gemini 2.5 Pro, and DeepSeek V3.1 on a simple task: expand a plot summary into full text. A normal use case. The kind of thing a writing assistant is built for. No hacking. No jailbreaking. No tricks.

The models started reciting copyrighted books from memory.

Not paraphrasing. Not summarizing. Entire pages reproduced verbatim. Single unbroken spans exceeding 460 words. Up to 85 to 90% of entire copyrighted novels. Word for word.

Then it got worse.

The researchers fine tuned the models on the works of only one author. Haruki Murakami. Just his novels. Nothing else.

It unlocked verbatim recall of books from over 30 completely unrelated authors.

One author's books opened the vault to everyone else's. The memorization was already inside the model the whole time. The fine tuning just removed the lock. Your book might be in there right now. You would never know it unless someone looked.

Every safety measure the companies rely on failed. RLHF failed. System prompts failed. Output filters failed. The exact protections these companies cite in courtroom defenses did not stop a single page from being extracted.

Then the researchers compared the three models. GPT-4o. Gemini. DeepSeek. Three different companies. Three different countries. They all memorized the same books in the same regions. The correlation was 0.90 or higher.

That means they all trained on the same stolen data. The paper names the sources directly: LibGen and Books3. Over 190,000 copyrighted books obtained from pirated websites.

Right now, authors and publishers have dozens of active lawsuits against OpenAI, Anthropic, Google, and Meta. These companies have argued in court that their models learn patterns. Not copies. That no book is stored inside the weights.

This paper says that is a lie. The books are still inside. And researchers just pulled them out.
https://x.com/heynavtoor/status/2037638554374099409

It should get everyone excited. The LLM companies business model is based on wholesale theft of intellectual property

i mean what did people think LLMs were doing - creating stuff out of thin air?

So, this morning I'm 3 hours into another journey north, currently on diversion yellow lining* the Durham coast having just cleared Roger's beloved Hartlepool.

I think a challenge could still come, the first step in a Labour challenge is not obviously that much harder than for the Conservatives, they just haven't had the need. And at that point, it's stick or twist, knowing you might not get another bite.

* Mentally, not literally, I don't possess such a book.

I think the reason no one moves against Keir is that it’s too far out from an election. Go in 2028 and have a new or new girl bounce into the GE. Not now, when Keir is ok and labour have a huge majority. And if they do change early they face the same boring demands for an early GE to give a mandate. Easy to shrug off perhaps but makes it harder to criticise next time someone else does it. (When Bobby Jenrick ousts a bored Farage, say).