Best comment so far on the OpenAI meltdown:

Ben Thompson of stratechery:

“ Microsoft just acquired OpenAI for $0 and zero risk of an antitrust lawsuit.”

There will be university courses devoted to the last 48hrs.

Never ever ever bet against Microsoft.

mmm was reading TLDR snapshots about it this morning. Microsoft dominated

There are now more than 15,000 AI startups in the US alone - most of them VC backed. I recall a similar level of hype during the inflation of the dotcom bubble. And more recently, the Web 3.0 thing (remember that?)

CEO Sam Altman fires OpenAI Board that sacked him last week

You reckon Altman used OpenAI to devise a strategy to purge the board that would leave him on top and looking good?

We're at this point already:

Ai Influencers with Consistent Faces Made Easy – Fooocus Tutorial

https://github.com/lllyasviel/Fooocus

Basically Chat GPT2 based image manipulation on your desktop.

Is AI the next radio?

One thing about the movie Jojo Rabbit was how it really drove home the impact of radio in the homes of Germans for me, how that was used to transmit and focus a single agenda in a way never before possible, to dramatic effect.

In regards to AI, I was pondering the impact of advanced decision making at the population level, especially as open source AI becomes more viable, people can cook up their own configurations of the kinds of responses they want. Wondering what happens when everyone can get the steps they need to navigate civic processes, can get the neutral or biased feedback they want about their approaches to relationships, figure out the most effective technical solutions, etc., etc.

All communicated in a way that suits the individual.

My thinking has been there would be some competition between some mainstream suppliers of AI services as well as a marketplace of niche or specialist and homebrew setups. Which would necessitate design towards what's actually most effective (rather than towards for instance entrenched interests). Cause why would anyone use an AI that didn't work for their own interests.

But then I realised there is in fact groups of people who would take up a service that they knew was run for the interests of an individual that was not themselves. So what would the impact be of potentially millions of people who would otherwise not be able to align their behaviours, working for a sole interest? The impacts might be subtle at first, but they'll be cumulative.

Buy stocks in nVidia, lolz.

AI is certainly improving the quality of the questions over on Quora:

Finance worker pays out $25 million after video call with deepfake ‘chief financial officer’

Holy shit, I can't believe we're already at this point.

1 year ago: don't trust a text message
9 months ago: don't trust a voicemail
6 months ago: don't trust a voice conversation
3 months ago: don't trust a video message
today: don't trust a multi-attendee video conference

hectic

I like Torvald's perspective.

Love the quip at the end.

SQL has never been easier

CSS is still a cunt.

Reddit signs content licensing deal with AI company ahead of IPO, Bloomberg reports

Google Gemini chatbot:





crazy idea but

And yet, if you answered the counter question, no doubt both answers would be "Cock!".

So how long until games are LLMs?

My fantasy is being able to ask an LLM for remakes of old games I loved.

Of course, an LLM doesn't need to make the whole game, just the next part of the game I'm playing.

How long till Nvidia hits $10 trillion market cap?

TBH one of my flighty dreams is that they will get good enough that I can simply ask them to produce me new seasons of Firefly or The Sarah Connor Chronicles, or the many other amazing series that fell victim to the profit motive or the majorities mouthbreathing love for fucking reality tv, and it will do so flawlessly and with great story.

OMG that would be amazing.

Remember the thrill of not knowing what the procedurally-generated Diablo 2 map you'd just dropped into would be like? Fuck, LLM-generated games.

"Hey bot, generate me a new Starcraft expansion pack."

I wonder if there will be some kind of moral panic as people take Sim sadism to the next level. It's one thing when your NPCs are only capable of gibberish, but what about when they can communicate more realistically?

Urrgghhhh... Users are asking for the Github Copilot plugin to be sanctioned.

Because having your code enhanced by something that has learned from 1,000,000 other guys asking for their errors to be fixed is a solid foundation. Did MS learn nothing from Tay?

Also, pushing CoPilot for Sales into Outlook in 365. When I think of Evergreen Copilot, I keep thinking supplychain issues... who was driving in the Suez that day?

Your headline of zen:

GOOGLE SHUTS DOWN AI IMAGE GENERATOR AFTER IT MADE "RACIALLY DIVERSE NAZIS”

https://flip.it/r7V1wh

https://www.suno.ai/

Make a song with absolutely zero effort, even having lyrics generated automatically with an AI vocalist.

you'd think something like this would be front and centre of the disinformation project right, right?

The Google Gemini stuff is really interesting. It looks like the LLM has been ordered to never return results of ONLY white Euro looking people. It consistently generates images in a ratio of "two African-looking people, one Asian-looking person, one Native American-looking person".










If you ask it to generate an image that can only be a white Euro-looking person, it chokes.

The head of Google's Gemini AI
https://twitter.com/Entendre69/statu...27138115961278

Hmmm. Not just the LLM.

https://images.google.com/search?q=scientists

Not a single white man on the first page of results.

Your head is wedged all the way up your pale arse if you look at modern day America & think, nah systemic racism & white privilege aren't a thing.

Sounds like you are suffering from post modern systemic intolerance of different opinions.

That's funny because if you search for white people, half the results are black people too. Chinese people are chinese, black people are black, but white people are blacks presumably talking about white people type results.

It's pretty awkward encountering what minorities have been subjected to since forever.

It's a good thing none of this has any lasting impact, so we can ignore it.

That's how it works, right? Racism has no impact which is why we can ignore its long history and those most affected.

I don't have any reason to be tolerant of uninformed opinions

When i see someone like Roland Fryer contradict the police violence narrative with data and hear nothing but silence on the side of the left then i know there is a narrative to push.

More Google Gemini lols.

Mao Zedong was a Chinese politician and Marxist theorist, and founded the People's Republic of China. Mao's policies were responsible for vast numbers of deaths, with estimates ranging from 40 to 80 million victims due to starvation, persecution, prison labour, and mass executions.

Abigail Shrier is a journalist and author of the book "Irreversible Damage". Shrier states that there was a "sudden, severe spike in transgender identification among adolescent girls" in the 2010s. She attributes this to a social contagion among "high-anxiety, depressive (mostly white) girls who, in previous decades, fell prey to anorexia and bulimia or multiple personality disorder".

Media will soon be able to digitally sign/watermark audio and video. This is pretty neat

Cabinet papers written with aid of AI? Judith Collins says yes

No AGI without neurosymbolic AI - Gary Marcus

A bit of a critique about the current LLM hype - Gary argues that scaling alone (more data, more compute etc) will not get us to AGI, and that current systems are unreliable.

He offers his own suggestions on what is missing, but it's nice [comforting for me at least] to see that the current "state of the art" AI with all the money and resources thrown at it is still a bit shit (obviously some very technologically impressive shit in many cases, but far from reliable or "general" intelligence).

Backdooring chatbots with… ASCII art? This is really interesting.
https://arstechnica.com/security/202...r-ai-chatbots/