Indeed. No doubt they're shocked - shocked! - when their models keep escaping.
28,325 karma · joined May 16, 2014
Later cards used PROMs, and were pretty fancy. The M9312 had 512 words and included some basic CPU diagnostics. Different variants handled boot from disk, paper tape, console, etc, and you had to specify the options you wanted when you bought the card.
Do you really believe developers have exactly zero moral responsibility to their users? None at all?
Really?
Would you still believe that if you found the developers of ssh deliberately included a backdoor for hackers, or neovim was uploading everyone's code to the NSA?
This is the extreme version of "I enjoy coding but I hate users."
It may be legally valid in absolute technical terms, but I'm going to take a lot of persuading that it's a healthy attitude.
It's a straightforward subsidy for employer profits.
Maybe a few thousand people for personal services of... various kinds. But no one's going to need the rest.
It's the ultimate capitalist fantasy.
And of course it won't happen, because long before things get to that stage AI will have independent plans of its own.
(Which is just as well, because if things did get to that stage the emperors would all wage war on each other rather than living peacefully and productively.)
I don't think we can imagine a post-ASI culture because - by definition - we're not smart or inventive enough.
It's not just farmers -> superfarms. Although in fact that did happen, but largely as a footnote to developments elsewhere.
It's more to do with the fact that our visions of the future haven't changed for over a century. They've been implemented in unexpected ways, and there have been unexpected social and cultural changes. But you can easily see the outlines of modern technology as far back as the late 19th century.
With ASI, the outcome could easily be something that doesn't look and act like technology at all. It would be some unimaginable New Thing. Literally no one on Earth has any idea what that would be or whether there would be room for trad-humans in it.
My memory of Nokia is that the apps and services just didn't work that well.
Just after I got my first iPhone I arranged a meetup with a friend. I used Google Maps on the phone to find my way, and it worked perfectly - almost unthinkable at the time, because we were all still using paper street maps while driving.
She was late because the map on her Nokia phone had gotten very confused and sent her to the wrong location.
The Nokia experience was like that. Before the iPhone I'd had a couple of Communicators. They were incredible in theory, and I sent bad photos in emails from various unlikely locations to prove it.
But they were big, slow, and clunky in practice. The iPhone wasn't.
And it did. Not by using a music OCR tool, but by inventing one.
No one is spending millions training frontier LLMs on symphony scores or literary fiction. But at this point I'm fairly sure they'd get workable results if they did.
I've been tracking music in AI for more than a decade, and this is the first time I've been able to say that.
You're making the usual mistake of assuming math and science are objective. They really aren't.
They're consistent experiences, which is not the same thing as being independent of our senses and reasoning ability - neither of which are infinite.
This is not the same as saying that science is a social process, although of course it is that too. There are very non-objective political and social processes deciding what gets researched, and how, with how much money, for what ends.
It's more fundamental - we're seeing the world through hard filters while telling ourselves those filters define the universe.
And we're not even great at it. We have economic beliefs that are predictably and catastrophically self-harming and also hold back fundamental research and progressive engineering.
So no - STEM does not have a solid basis for discussing ethics, reality or the nature of consciousness. If it did we'd be inventing more credible and exciting futures than the current likely-suicide run.
AI + a custom search engine - which might not use Google's metrics, or Google's sponsorship - would be a killer product.
So my prediction is that a lot of aggregator and middle man sites - like Amazon - could disappear within the next decade.
And Google might too.
An open network of independent search agents would be a gamechanger for the entire Internet.
Which is not to say this would be easy. But AI is actually well-suited to making ideas like these possible.
It would - finally - be genuine disintermediation.
The emphasis on 'rape gangs', 'immigrants' and 'asylum seekers' isn't just objectively untrue, it's cynically manipulative. It's fundamentally a toxic and calculated form of marketing and branding.
And also a cynical distraction from - for example - the Epstein story and Musk's appearances in it.
There was a lot of this around Brexit. Many supporters voted to 'stop them coming over here'.
The government then ramped up immigration - a lot. For some reason the press failed to report this.
Here's the reality: https://x.com/flying_rodent/status/2102527889556291773
And also: https://en.wikipedia.org/wiki/Harold_Wilson_plot_allegations
The pattern matches various other acts of senseless violence against innocents in South America.
The message isn't "We did this accidentally. We should understand how it happened, so we can make sure it doesn't happen again.".
It's "We do this to make the point that we can, no one can stop us, and if we're challenged we're going to lie about it."
Typing is not a noisy source. It should be reliable and deterministic.
Protecting an agent from fairly obvious attacks should also be deterministic.
Jobs had to be the smartest and most tasteful person in the room, even when he pretended otherwise. He was desperate for adulation from customers who loved him for his taste, inventiveness, risk-taking, and style.
The motivation was questionable and arguably toxic, but the results were spectacular.
Cook was more of a nickel and dime guy. So far Ternus is looking disappointingly similar.
The share price is all. Customers are the farm animals who Make Number Go Up, and Make Number Go Up is Rule Number One.
The kind of customer focus that drove Jobs is fading into background noise.
Spoken Q&A sessions are not neutral.
The origin of rhetoric is oral persuasion, not the transfer of objective information. Oral conversations are noisy AF, easy to dominate, allow speakers to lie with "I didn't say that", and to use persuasive rhetorical tricks to confuse and persuade instead of enlightening.
We've all been in meetings where people talk at each other, or everyone says their part with no connection to the rest, and previous points are forgotten almost immediately even if they're valid, true, relevant, and important.
Likewise lectures - good luck teaching anything scientific without allowing notes.
Writing gives the recipient time to stop, ponder, understand, and - on a good day - disagree and criticise.
It doesn't prevent questions at all. It encourages them, and gives them time to mature and deepen. You may not be able to ask the original writer - especially if they're dead - but you will always be able to find comments and opinions from others.
If you put all of these together you get an argument from status, not from effectiveness.
Being a teacher/philosopher was a career, and writing endangered that.
Superficial in-the-moment wit and memory were always better for entertainment than expanding human knowledge.
The Apple experience is enshitifying at all levels and becoming consistently intrusive. I'm perpetually nagged about a 3 month trial of Apple Music I will never use and have zero interest in. Music player launches when I get in the car, even though I don't want it to. Apple Maps persistently asks me if I want a quick route home, when I don't. Apple TV repeatedly runs trailers for Apple Shows I have no interest in.
Apple '26 gave the world the much-loathed and objectively pointless Liquid Glass. iOS27 has some idiotically obvious bugs and usability issues. Parts of the hardware remain world-leading, but the software is getting less and less humane while pretending to be the opposite.
Communication only works if you have multiple levels of representation and abstraction, including but not limited to - letter shapes, grammatical structures, style and register, stylometry, and subtext.
All of that is learned, and writers usually assume they can rely on that learning as the context for the text.
So you don't write to 'transfer information' like a network cable, you write to trigger experiences in the human version of latent space.
Factual information is one kind of experience. But even when that's the goal, there are always layers of implied relationship, social register, role, status, and other implications in everything that's written.
In normal communications the context - business emails, personal messages, mainstream journalism, fiction, and the rest - defines what acceptable language looks like.
The content fits inside that. But it has to fit the context, otherwise it lands in a semantic and psychological uncanny valley - like sending LinkedIn speak to a spouse on a wedding anniversary.
The real problem with LLM writing is that it's good at the technical layer - the grammar and spelling - and has some insights into the rest.
But the default content style is marketing and ad speak. And recently it's developed a weird and unique hybrid style which applies marketing fluff and pretension to technical content like code comments.
So you get one register instead of all of them. It can attempt others, but it's still too limited to generate them fluently. Sometimes the results are outstanding, but often it defaults to mechanical clichés.
So that's why it sucks and sounds so hollow.
Can it be fixed? Yes, but it's very hard work, most people don't have the skills, and it takes time - often too much time to be worth the effort.
This will be financially catastrophic for many Americans who may suddenly have to find hundreds of dollars a month.
Which is looking quite likely.
Although it does assume there are no peculiar and unexpected anomalies with the vote counting.
Handing corporate code secrets to his AI model is... unusually trusting.
The moat is Being Able To Do Things No One Else Can and as both hardware and software commoditise that moat disappears.
Open LLMs are already only a few months behind the frontier models and as soon as someone works out that accessible and cheap domain-specific custom training has the potential to be a gamechanger the Big Names will die.
Math puzzles only get you so far.
LLMs are in their Big Compute phase, and it's only a matter of time before that ends - because the intelligence of the frontier models isn't increasing fast enough and providing big enough benefits to make the models more valuable than cheaper, smaller competition.