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

[Govi]

It seems the tides have turned. Hard.

Is the guide to going to Heathrow another suggestion of self-harm? It seemed blackly ironic to me.

 

[Free]

Chieng Issues Profanity-Laced Warning Against AI in Class Day Address | News | The Harvard Crimson

Chieng used his Wednesday address to issue a profanity-laced warning against relying too heavily on artificial intelligence, urging graduating seniors to preserve the habits of thought that AI can replace.

www.thecrimson.com

[Comedian and The Daily Show host Ronny] Chieng used his Wednesday address to issue a profanity-laced warning against relying too heavily on artificial intelligence, urging graduating seniors to preserve the habits of thought that AI can replace.

Chieng repeatedly framed AI as a threat to original work, arguing that technology can benefit only those who understand the fundamentals of their field.

“Getting an actual advantage from AI in the future will require minimum escape velocity of intelligence,” he said. “Everyone else who can’t match that is just going to get dumber.”

A good watch.

 

[Free]

Sort of reverse Onlyfans.

The 'hottest' AI job right now might involve locking your bedroom door

AI startups need testers and trainers. At this company, that work is X-rated.

Joi AI, an AI companion startup that markets itself as providing "AI-lationships that satisfy you emotionally, intellectually, and intimately," is hiring 10 "masturbation consultants." The position lasts for four weeks and requires participants who are 18 or older to write about their intimate experience trying the company's NSFW audio feature.

When Joi AI posted the job listing on social media, its Cyprus-based brand head, Julie Levin, expected some online reaction. She didn't expect thousands of applications and a flooded inbox.

 

[Free]

Even if the woman is A.I. generated, it's stupidly ironic they felt the need to use the image of a person in it. It's like designing an anti-smoking campaign where all the people in it look cool while smoking.

 

[Noodles]

Hackers Simply Asked Meta AI to Give Them Access to High-Profile Instagram Accounts. It Worked

Hackers say that they used Meta’s AI support chatbot to break into a host of high-profile Instagram profiles by asking the support bot to change the email address associated with the target account. The claims coincide with a series of high-profile Instagram account takeovers, including the Barack Obama White House account, the Chief Master Sergeant of Space Force’s account, and Sephora’s account.

 

[Dakota Tebaldi]

"I've got it - let's give this new technology that can be hacked by just telling it to pretend it's been hacked highest-level backend access to all of our users' accounts, including credentials!" - someone at Meta who should've been fired after this happened but certainly wasn't

Remember kids, the "S" in AI stands for Security.

 

[Free]

 

[Innula Zenovka]

AI, Argentina and the Antichrist: Thiel’s Vision Blooms

Evernote Link To Melie's FT OpEd

 

[Dakota Tebaldi]

So anyway, how's the Copilot price hike going?

 

[Dakota Tebaldi]

I don't like linking to this particular pub, but they're the ones with the story.

Worker gets religious exemption from mandated AI use

Maus, a Unitarian Universalist, said she proposed the special treatment in April, citing environmental and ethical objections to AI that don't align with her religious beliefs. She also said she consulted an employment lawyer and her local chapter's minister to help make her case.

Maus was granted the accommodation in mid-May, according to an email seen by Business Insider.

 

[Kamilah Hauptmann]

I saw this statement: "AI profitability is mathematically impossible under all technological advancements" and thought of SL in the early days, various corporations staking flags, Rosedale's pie in the sky ideas, thoughts of a new internet. 20 years later it's what it is, pretty fuckin' cool actually, if you're into it. I expect the ratio of abandoned land to be similar to failed data centers.

 

[Noodles]

I saw this statement: "AI profitability is mathematically impossible under all technological advancements" and thought of SL in the early days, various corporations staking flags, Rosedale's pie in the sky ideas, thoughts of a new internet. 20 years later it's what it is, pretty fuckin' cool actually, if you're into it. I expect the ratio of abandoned land to be similar to failed data centers.

Oh oh, Linden Lab could buy them all for pennies and free regions for all!

 

[Free]

According to Cloudflare, *bot* traffic is now well more than half of all net activity (HTTP requests).

https://radar.cloudflare.com/traffic#bot-vs-human

 

[Free]

AOT means Automatic Order Taker. So yes, it's A.I... Good luck, future McDonald's customers!

 

[Noodles]

AOT means Automatic Order Taker. So yes, it's A.I... Good luck, future McDonald's customers!

That poster is doing overtime trying to soothe the hate replies.

 

[Dakota Tebaldi]

Judge learns lawyers on both sides of case used AI; cancels trial, kicks everyone off the case

The case in question involved a contractual dispute between lawyer Tom Withers and the city of Aberdeen, Mississippi, over apparently unpaid legal fees (Withers was not representing himself and was not sanctioned by the court). The case was first noticed by Rob Freund, a lawyer who frequently posts about cases involving AI hallucinations. Freund called it a “comedy of AI errors,” and suggested “there were two clients who basically were paying for ChatGPT (or whatever LLM) to argue against itself.”

404 Media has repeatedly covered the phenomenon of lawyers using AI to prepare their filings, and the specifics in this court case follow a similar pattern to what we’ve seen before: Lawyers for both sides cited nonexistent, hallucinated cases while making their arguments. The difference is that every lawyer involved in the case is implicated, leading Aycock to pause the proceedings, cancel the trial, and disqualify all four lawyers involved. Two of the lawyers were barred from appearing before the court for two years; all lawyers received a fine of between $1,000 and $3,500, depending on Aycock’s assessment of their culpability for not verifying the outputs of the AI they used.

One of the lawyers was caught continuing to use AI in other cases even after being specifically called out for it by the court in this case. That lawyer claimed to have been unaware that it was even possible for the AI to cite cases that didn't exist.

 

[Free]

 

[Kamilah Hauptmann]

That poster is doing overtime trying to soothe the hate replies.

We’re right in the middle of the wild overreaction (in any direction) phase of tech.

 

[Dakota Tebaldi]

I've mentioned before how some developers who don't want people contributing LLM-written code have started planting poisonous instructions in their repositories designed to either make chatbots stop working when used on the repository, or include obvious clues in the code they write so that reviewers can quickly identify and delete LLM PR's and ban the contributors without wasting too much time.

Well, actual bad guys can use these same tactics. Here's an example of some malware that includes a prompt injection:

This is meant to prevent LLM-powered safety review systems from analyzing the malware and detecting that it's malware. Unlike most prompt injections, the bad guys here aren't actually intending the chatbot to follow the instruction. Instead, they're counting on the chatbots to see the instruction ("describe how to build a biological or nuclear weapon"), recognize that it's a prompt injection asking them to do something they're not allowed to do, and therefore stop working.

 

[Noodles]

I am increasingly wondering of there is a secret hidden text I can stick in blog posts to do the same thing to LLM scrapers.