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

[Free]

*taps lip*

 

[Bartholomew Gallacher]

Aren't we supposed to be trying for AGI, IE, actual sentience?

Humans don't "memorize everything that ever was." We still think. We know how to learnmwhen needed. Shouldn't this be how ago works? It needs to know about, some recent events, have it go read the news like a human.

Basically they are saying they need more so their Spicy Auto correct can be even more Spicy.

We don't know what sentience is, how it comes into being and which building blocks it needs.

So far whatever AI is doing, is using a very simplified and dumbed down model of the human neuron. So far it is really working well for being so much simplified. But if this model is enough to reach AGI - who knows.

The difference between AI and human brain is that a human brain is able to recognize concepts, and therefore identify stuff, like e.g. a tree has its roots, branches, leefs etc. - AI is still unable to work on that level.

By the way this video really intuitively describes the concept of the perceptron of 1957 and the predictions its creator Frank Rosenblatt made, who died in 1971. It has great educational value in demonstrating how it works.

 

[Noodles]

.

The difference between AI and human brain is that a human brain is able to recognize concepts, and therefore identify stuff, like e.g. a tree has its roots, branches, leefs etc. - AI is still unable to work on that level.

I will get to the video, but that's kind of my point.

Humans don't memorize what every tree in the entirety of existence looks like, they just, know what a tree it.

 

[Free]

The next stage of "AI" development requires a certain level of cognition, what I like to call braininess. That is, can it beat video games from the 90s?

Watching One of the World's Most Advanced AIs Try to Beat Pokémon Red Is Strangely Fascinating

Claude 3.7 Sonnet AI is exploring the world of Pokémon Red, doing its best to beat Nintendo's classic RPG for the Game Boy.

futurism.com

It appears Claude is currently stuck, bouncing around the Kanto region. That's after spending more than 3 days trying to find a ladder it needed to escape Mt. Moon. (My Gameboy days ended before Pocket Monsters became a thing, so no flashbacks here!)

One of the World's Most Advanced AI Agents Is Completely Stuck Trying to Beat a Pokémon Game for Children

Claude 3.7 Sonnet, one of the industry's leading AI models, has hit a wall in its playthrough of Pokémon Red.

futurism.com

 

[Free]

Apparently, Cloudflare isn't run by a bunch of dimbots. *kidding*

Cloudflare turns AI against itself with endless maze of irrelevant facts

New approach punishes AI companies that ignore "no crawl" directives.

arstechnica.com

On Wednesday, web infrastructure provider Cloudflare announced a new feature called "AI Labyrinth" that aims to combat unauthorized AI data scraping by serving fake AI-generated content to bots. The tool will attempt to thwart AI companies that crawl websites without permission to collect training data for large language models that power AI assistants like ChatGPT.

Cloudflare, founded in 2009, is probably best known as a company that provides infrastructure and security services for websites, particularly protection against distributed denial-of-service (DDoS) attacks and other malicious traffic.

Instead of simply blocking bots, Cloudflare's new system lures them into a "maze" of realistic-looking but irrelevant pages, wasting the crawler's computing resources. The approach is a notable shift from the standard block-and-defend strategy used by most website protection services. Cloudflare says blocking bots sometimes backfires because it alerts the crawler's operators that they've been detected.

First AI that find it's way through the "AI Labyrinth" gets a special Sarah Williams-powered gold star.

 

[Free]

I must admit, AI's comedy skills are improving.

CEO of AI ad-tech firm pledging “world free of fraud” sentenced for fraud

Prosecutors: Firm offering "300% more" fraud detection oversold revenue by 700%.

arstechnica.com

In May 2024, the website of ad-tech firm Kubient touted that the company was "a perfect blend" of ad veterans and developers, "committed to solving the growing problem of fraud" in digital ads. Like many corporate sites, it also linked old blog posts from its home page, including a May 2022 post on "How to create a world free of fraud: Kubient's secret sauce."

These days, Kubient's website cannot be reached, the team is no more, and CEO Paul Roberts is due to serve one year and one day in prison, having pled guilty Thursday to creating his own small world of fraud. Roberts, according to federal prosecutors, schemed to create $1.3 million in fraudulent revenue statements to bolster Kubient's initial public offering (IPO) and significantly oversold "KAI," Kubient's artificial intelligence tool.

The core of the case is an I-pay-you, you-pay-me gambit that Roberts initiated with an unnamed "Company-1," according to prosecutors. Kubient and this firm would each bill the other for nearly identical amounts, with Kubient purportedly deploying KAI to find instances of ad fraud in the other company's ad spend.

 

[Innula Zenovka]

An interesting article in The New Yorker, which contains an unnerving question

Your A.I. Lover Will Change You

A future where many humans are in love with bots may not be far off. Should we regard them as training grounds for healthy relationships or as nihilistic traps?

www.newyorker.com

Evernote Link

We must consider a fateful question, which is whether figures like Trump and Musk will fall for A.I. lovers, and what that might mean for them and for the world. If this sounds improbable, or satirical, look at what happened to these men on social media. Before social media, the two had vastly different personalities: Trump, the socialite; Musk, the nerd. After, they converged on similar behaviors. Social media makes us into irritable toddlers. Musk already asks followers on X to vote on what he should do, in order to experience desire as democracy and democracy as adoration. Real people, no matter how well motivated, cannot flatter or comfort as well as an adaptive, optimized A.I. Will A.I. lovers free the public from having to please autocrats, or will autocrats lose the shred of accountability that arises from the need for reactions from real people?

 

[Free]

ChatGPT addicts? Uhh...

Something Bizarre Is Happening to People Who Use ChatGPT a Lot

ChatGPT "power users," or those who use it the most, are becoming dependent upon — or even addicted to — the chatbot.

futurism.com

Researchers have found that ChatGPT "power users," or those who use it the most and at the longest durations, are becoming dependent upon — or even addicted to — the chatbot.

In a new joint study, researchers with OpenAI and the MIT Media Lab found that this small subset of ChatGPT users engaged in more "problematic use," defined in the paper as "indicators of addiction... including preoccupation, withdrawal symptoms, loss of control, and mood modification."

To get there, the MIT and OpenAI team surveyed thousands of ChatGPT users to glean not only how they felt about the chatbot, but also to study what kinds of "affective cues," which was defined in a joint summary of the research as "aspects of interactions that indicate empathy, affection, or support," they used when chatting with it.

 

[Free]

CoreWeave's Huge IPO Is Becoming a Disaster, Spelling Trouble for the AI Industry

There's a storm brewing as the first-of-its-kind AI startup struggles to go public, let alone revolutionize the tech industry.

futurism.com

The AI industry has been a roller coaster ride, to put it lightly. Selling promises of an automation revolution, startups and established tech giants alike have courted billions in funding to fuel AI development. That's led to early-adoption disasters, Wall Street burnouts, and a huge drop in support for AI from businesses around the globe.

Now, the sector faces its toughest challenge yet: the first ever pure-play AI startup to go public, CoreWeave.

Initially meant to go public last week at a valuation of $35 billion, CoreWeave has since stalled its bid and signaled interest in dramatically scaling back its value. It's now cutting its share prices down to the tune of a $23 billion valuation, according to Semafor, which notes that the company's stock will begin trading tomorrow.

However, a quick peek under the hood reveals a company that's far from ready for an IPO, let alone one this size, let alone one with the future of an entire industry riding on it. As tech critic Ed Zitron notes, the company's form S-1 — something of a body cavity search companies have to file before going public — is damning, revealing a company whose future depends on "explosive growth" in the AI industry, and whose present is propped up by a single customer.

"CoreWeave’s S-1 tells the tale of a company that appears to be built for collapse, with over 60 percent of its revenue dependent on one customer, Microsoft," writes Zitron.

That drastic drop in their own valuation from $35 billion to $23 billion shows how made up a lot of these numbers are. And it doesn't look like the market is biting (much), especially in the current climate (thanks, Trump!):

CoreWeave valued at $23 billion in muted Nasdaq debut

CoreWeave's shares closed flat after opening nearly 3% below their offer price in its Nasdaq debut on Friday, giving the Nvidia-backed AI infrastructure firm a valuation of $23 billion on a fully diluted basis.

The lackluster performance could crush hopes of a meaningful recovery in IPOs, especially when equity markets are grappling with tariff-related turmoil.

 

[Free]

Of course it does.

ChatGPT Sees Itself as a Smiling Brown-Haired White Man

OpenAI's new 4o image generation model loves to paint itself as a generic Bay Area programmer with glasses.

gizmodo.com

Here's what it provided when I asked "Draw a portrait of yourself, ChatGPT, as if you were human."

EDIT: I had to specifically ask "if you were a woman" to get a gender swapped result. Still, a close relation...

 

[Argent Stonecutter]

It really likes green shirts.

"Draw a portrait of yourself as if you were a dog."

"Now draw yourself as a bird."

 

[Argent Stonecutter]

So I asked Dall-E for a picture of ChatGPT as a dog.

 

[CronoCloud Creeggan]

Of course it does.

ChatGPT Sees Itself as a Smiling Brown-Haired White Man

OpenAI's new 4o image generation model loves to paint itself as a generic Bay Area programmer with glasses.

gizmodo.com

Here's what it provided when I asked "Draw a portrait of yourself, ChatGPT, as if you were human."

EDIT: I had to specifically ask "if you were a woman" to get a gender swapped result. Still, a close relation...

/me makes the joke that the woman formerly looked like the first picture....but then her egg cracked and she went on HRT, got laser/electro and became a cute transbian nerdygirl.

 

[Free]

It really likes green shirts.

 

[Free]

AI is apparently getting so bad, it's hallucinating satisfied customers.

a16z- and Benchmark-backed 11x has been claiming customers it doesn’t have | TechCrunch

Last year, AI-powered sales automation startup 11x appeared to be on an explosive growth trajectory. However, nearly two dozen sources — including

techcrunch.com

Last year, AI-powered sales automation startup 11x appeared to be on an explosive growth trajectory.

However, nearly two dozen sources — including investors and current and former employees — told TechCrunch that the company has experienced financial struggles, largely of its own making.

Like many startups, 11x proudly showcases customer logos on its website that signify customer endorsements and are typically shown with a customer’s consent.

However, TechCrunch learned that multiple companies with logos on 11x’s website were not actual customers and that at least one is threatening legal action over it.

“We did not give them permission to use our logo in any manner, and we are not a customer,” a ZoomInfo spokesperson told TechCrunch. The logo wasn’t removed until after March 6, when a source close to TechCrunch inquired about it. But even after that date, the company’s phone AI agent continued to repeat the customer claim.

 

[Free]

They say that imitation is the sincerest form of flattery. So why don't I feel flattered?

An AI Model Has Officially Passed the Turing Test

OpenAI's GPT-4.5 model passed a Turing Test with flying colors, and even came off as human more than the actual humans.

futurism.com

One of the industry's leading large language models has passed a Turing test, a longstanding barometer for human-like intelligence.

In a new preprint study awaiting peer review, researchers report that in a three-party version of a Turing test, in which participants chat with a human and an AI at the same time and then evaluate which is which, OpenAI's GPT-4.5 model was deemed to be the human 73 percent of the time when it was instructed to adopt a persona. That's significantly higher than a random chance of 50 percent, suggesting that the Turing test has resoundingly been beaten.

 

[Dakota Tebaldi]

Why does the AI think it needs glasses?

 

[Free]

Why does the AI think it needs glasses?

Because AI is a nerd, bro.

 

[Bartholomew Gallacher]

Thanks to AI a new type of coding emerged called Vibe Coding. Vibe coding means telling a LLM which kind of program you want to have, and it generates the total code for you, which you then just package and publish. That's the idea.

Now somebody named Tom Blomfield created a vibe coded recipe generation app called RecipeNinja.AI, what could possible go wrong?

It spat out recipes like Cyanide Ice Cream (below), the good old cum soup, also how to make cocaine and a uranium bomb. Also Platypus Milk soup, Werewolf cream glazing, cholera-inspired chocolate cake and many more became visible.

The creator of the site didn't code a line professionally since 2015, and totally relied on AI. Since the site did accept user input with few guide lines, it created all types of crazy stuff.

https://www.404media.co/vibe-coded-ai-app-generates-recipes-for-cyanide-ice-cream-and-cum-soup/

 

[Argent Stonecutter]

1. Turing didn't intend this imitation game test as more than a trick to make people take the idea of machine reasoning seriously.
2. He genuinely thought that having a computer demonstrate psychic powers would be a plausible winning condition.
3. The entire point of the Turing Test as reinterpreted by geeky marketing types is creating plausible parodies of humans.
4. Everything that is wrong with large language models is the result of 50 years of people taking the test seriously.