Nobody cares about "AI" (Chatbot: I disagree.)

[Free]

It's like having a brush with Kafka. Only you've mistaken the brush for a painting utensil.

(No, I don't mean this.)

 

[Noodles]

An experiment to see if you can identify AI writing.

So ... is AI writing any good? .... PART 2!

So ... is AI writing any good? Two years ago I asked the same question and pitted 4 good authors against ChatGPT 4 in a flash fiction head-t...

mark---lawrence.blogspot.com

There are some links to get to the stories, and interim results, which you should check after reading.

I got them all correct.

Spoiler: The Biggest Tell

The AI ones don't really finish or have any real plot, they are all just overly descriptive and meandering with basically zero wrap up. Tell tale of the AI "Would you like to explore more...." I use AI to make dumb short stories (not published anywhere) like this and it always Lways wants to continue. I only ever got a "The End" out of it once ever.

 

[GoblinCampFollower]

An experiment to see if you can identify AI writing.

So ... is AI writing any good? .... PART 2!

So ... is AI writing any good? Two years ago I asked the same question and pitted 4 good authors against ChatGPT 4 in a flash fiction head-t...

mark---lawrence.blogspot.com

There are some links to get to the stories, and interim results, which you should check after reading.

I got them all correct.

Spoiler: The Biggest Tell

The AI ones don't really finish or have any real plot, they are all just overly descriptive and meandering with basically zero wrap up. Tell tale of the AI "Would you like to explore more...." I use AI to make dumb short stories (not published anywhere) like this and it always Lways wants to continue. I only ever got a "The End" out of it once ever.

Yeah.... as you said, AI knows what words go together, but has no real grasp of what a human would consider important. It's very hard for AI to "learn" such things without physical bodies and related experience. This is also related to how AI generates terrible summaries because the summaries will be well written but have no idea if a real person would think a detail is important or not. It still has no real grasp of "plot".

 

[Innula Zenovka]

BBC Verify Live: How we know this image of 'leaders' waiting for Trump is AI

Latest updates from the BBC's specialists in fact-checking, verifying video and tackling disinformation.

www.bbc.co.uk

 

[Free]

Is Meta's Superintelligence Overhaul a Sign Its AI Goals Are Struggling?

The company is dismantling the division it built two months ago, and is looking to downsize after a remarkably expensive hiring spree.

gizmodo.com

Meta is splitting its AI division Meta Superintelligence Labs less than two months after the company announced its formation in June.

The group will be split into four smaller groups, according to a New York Times report. One group will focus on AI research, another one on infrastructure and hardware projects, one on AI products, and another one on building out AI superintelligence, a hypothetical AI system that could outperform human intelligence on any and all scales.

Facebook did not respond to a request for comment.

I wouldn't expect them to.

Along with the restructuring, Meta is also looking at downsizing its AI division completely, although no final decision has been made on that. That may not be too surprising given the multi-billion dollar hiring spree summer Meta has been having, which is likely to cause some shareholders concern when the company next releases spending.

First their was the vast amounts of money Zuck flushed down the toilet on "metaverse" and VR stuff. Now this...

Nelson-Muntz-ha-ha.gif

 

[Free]

BBC Verify Live: How we know this image of 'leaders' waiting for Trump is AI

Latest updates from the BBC's specialists in fact-checking, verifying video and tackling disinformation.

www.bbc.co.uk

First thing I noticed was it was lacking Zelenskyy. Seems a bad miss on the prompter's part.

 

[Dakota Tebaldi]

It Took Many Years and Billions of Dollars, But Microsoft Finally Invented A Calculator That Is Wrong Sometimes

Think of it. Forty-five hundred years ago, if you were a Sumerian scribe, while your calculations on the world's first abacus might have been laborious, you could be assured they'd be correct. Four hundred years ago, if you were palling around with William Oughtred, his new slide rule may have been a bit intimidating at first, but you could know its output was correct. In the 1980s, you could have bought the cheapest, shittiest Casio-knockoff calculator you could find, and used it exclusively, for every day of the rest of your life, and never once would it give anything but a correct answer. You could use it today!

But now we have Microsoft apparently determining that "unpredictability" was something that some number of its customers wanted in their calculators. I will admit to envisioning a thrilling frisson as I review the Defector P&L statement, not knowing if it will ensure our sustainable financial future or land us all in prison for fraud. I will enjoy it as much as the next fellow when I read Kathryn's proprietary new evaluation model that identifies Giancarlo Stanton as MLB's fastest runner, going first-to-third in 0.4 seconds on average. One can only hope that when you point out one of Copilot's errors, it apologizes in that wheedling, I'm-just-a-smol-bean LLM register that makes me want to pour a glass of water into its mainframe.

 

[Free]

"Hey, the industry I make a living with may be bringing about the end times. Let's have lunch!"

These are the people INVOLVED in AI.

The AI Doomers Are Getting Doomier (archive.today)

Nate Soares doesn’t set aside money for his 401(k). “I just don’t expect the world to be around,” he told me earlier this summer from his office at the Machine Intelligence Research Institute, where he is the president. A few weeks earlier, I’d heard a similar rationale from Dan Hendrycks, the director of the Center for AI Safety. By the time he could tap into any retirement funds, Hendrycks anticipates a world in which “everything is fully automated,” he told me. That is, “if we’re around.”

The past few years have been terrifying for Soares and Hendrycks, who both lead organizations dedicated to preventing AI from wiping out humanity. Along with other AI doomers, they have repeatedly warned, with rather dramatic flourish, that bots could one day go rogue—with apocalyptic consequences. But in 2025, the doomers are tilting closer and closer to a sort of fatalism.

 

[Free]

It's never a good sign when an industry tries to gin up legal barriers to protect them against lawsuits.

Politicians Are Trying to Make It Illegal to Sue AI Companies

A new bill in Colorado would prevent individuals from suing AI companies under the state's consumer protection laws.

futurism.com

 

[Dakota Tebaldi]

"Hey, the industry I make a living with may be bringing about the end times. Let's have lunch!"

These are the people INVOLVED in AI.

They feel like they have no choice. There's this ideology - for all intents and purposes a religion, really, calling itself Rationalism (not to be confused with 'rationality' dear god don't make that mistake). It's the R in TESCREAL, if you've heard of that term before. Rationalists believe as an article of faith that an advanced AI superintelligence must eventually, inevitably come about that has the potential to solve all of the world's problems overnight and advance humanity into a utopian golden age, but that last part can only happen if the super-AI is properly "aligned" to work for the greater good of humanity. So they say that it is the most important thing in the whole world for THEM - rationalists - to make a properly-aligned AI superintelligence as soon as possible, before someone else makes a detrimentally aligned one.

It all became a problem because they're stupid. They gambled on finance hype-bros to make AI happen but they neglected to account for the fact that finance hype-bros don't actually care about all this culty nerd-wank, they just want as much money as they can make as fast as they can make it, so they have absolutely no incentive to go "slow and safe" and "ensure alignment". On top of that, Rationalism isn't a monolithic faith, there's denominations - there are disagreements about what exactly what "proper alignment" looks like, and what values a superintelligence ought to prioritize for the greater good. It was all silly hypothetical navel-gazing when it was just people "debating" each other on internet boards, but LLMs exist now and these guys are dumb enough to believe that LLMs are the thing that will evolve into the superintelligence, so to them suddenly the "singularity" is imminent and the people closest to creating it have not properly aligned it, and it's all "real" and a very dire, immediate crisis with the fate of humanity and possibly the cosmos at stake.

 

[Dakota Tebaldi]

AI doesn't seem to do well with original riddles.

It's neat that these examples kind of show how the LLMs are failing though. It looks through its training data and finds a very famous riddle with a very famous solution that is constructed similarly to the prompts, and it assumes the solution is probably the same. It sees that the prompted riddles are different, but doesn't understand that they are different in a way that makes the "original" solution nonsense, so it just insists it's still the solution, even if it has to claim that the prompt riddle is flat out lying in places, i.e. the mechanic in his workshop is "actually a surgeon".

 

[Dakota Tebaldi]

TIL: You can upload images to a chatbot and it will automatically downscale them, and if you have designed the images so that when they are downscaled text appears and that text is in the form of a prompt, the chatbot will silently execute the prompt.

For something like ChatGPT it probably doesn't matter, but since Google and Microsoft have chatbots that are granted the power to look through all of your account data and autonomously take actions like create calendar events and send emails, a trick like this can be used to steal confidential information.

 

[Argent Stonecutter]

It's neat that these examples kind of show how the LLMs are failing though. It looks through its training data and finds a very famous riddle with a very famous solution that is constructed similarly to the prompts, and it assumes the solution is probably the same.

Try asking it anything involving any number of combinations of wolves, goats, boats, rivers, and cabbages.

Though some products now have system rules specifically dealing with avoiding the river crossing puzzle.

For something like ChatGPT it probably doesn't matter, but since Google and Microsoft have chatbots that are granted the power to look through all of your account data and autonomously take actions like create calendar events and send emails, a trick like this can be used to steal confidential information.

Prompt-inject Copilot Studio AI via email, grab a company’s whole Salesforce

Copilot Studio is a Microsoft AI agent for the enterprise. An ordinary non-techie user can make a chatbot search front-end for your company data. It’s a chatbot agent, so it’s full of h…

pivot-to-ai.com

 

[Argent Stonecutter]

Oh, because this happened in Australia where unions are actually a thing they had to back down when they were caught trying to use "AI" as an excuse for breaking contracts and outsourcing jobs overseas:

Commonwealth Bank fires, un-fires call centre staff over AI

The Commonwealth Bank, Australia’s largest bank, is going all-in on AI! In June, it put in a new AI voice bot to sort calls to customer service! [ITnews] The new voice bot was so efficient, the ban…

pivot-to-ai.com

 

[Innula Zenovka]

404 Media: Citizen Is Using AI to Generate Crime Alerts With No Human Review. It’s Making a Lot of Mistakes

Crime-awareness app Citizen is using AI to write alerts that go live on the platform without any prior human review, leading to factual inaccuracies, the publication of gory details about crimes, and the exposure of sensitive data such as peoples’ license plates and names, 404 Media has learned.

The news comes as Citizen recently laid off more than a dozen unionized employees, with some sources believing the firings are related to Citizen’s increased use of AI and the shifting of some tasks to overseas workers. It also comes as New York City enters a more formal partnership with the app.

 

[Free]

404 Media: Citizen Is Using AI to Generate Crime Alerts With No Human Review. It’s Making a Lot of Mistakes

Article through archive.today.

 

[Essence Lumin]

Article through archive.today.

I don't get why it's so important for sites to want to be indexed by the archive sites they let them through but is quite useful.

 

[Free]

I don't get why it's so important for sites to want to be indexed by the archive sites they let them through but is quite useful.

archive.today is more about getting around subscription walls than anything else. But in regards to its archival element, it's completely self-selective: only the urls/pages people save to it are "archived." I suspect many of the sites with pages archived on it are not necessarily wanting that to happen.

 

[Free]

Good, I can stop watching the Weather Channel.

Google's AI Weather Model Nailed Its First Major Storm Forecast

They're cooking in the Weather Lab.

gizmodo.com

 

[Innula Zenovka]

BBC reveals web of spammers profiting from AI Holocaust images

The BBC has found an international network of content creators profiting from AI generated posts on Facebook.

www.bbc.com

Mr Sawicki said the tsunami of fake images was undermining the Auschwitz Memorial's mission to raise awareness of the Holocaust.
"We already started getting comments on our Facebook posts that 'oh, this is an AI-generated photograph'," he said.
Survivors and families are also disturbed by the surge of Holocaust AI slop, according to an organisation promoting Holocaust education and research.
"They don't quite understand what they're seeing," said Dr Robert Williams, from the International Holocaust Remembrance Alliance.
He said Holocaust survivors were feeling a "certain sense of sadness this has been allowed to happen" despite government and philanthropic investments into awareness campaigns.
"They feel like their efforts haven't been enough," he said.
"That's a very sad thing to consider because the last of the survivors will soon leave us."