Before this is all over, MS will be charging users to extract their snapshots from a proprietary cloud-only one drive account. The recovery process will take about 3 hours, and involve scrolling through ai-authored help articles that don’t lay out clearly and methodically how to access the old snapshots. The comments on the help articles will begin with “Hello sir, can you confirm that you have followed the steps at this link?”. The link, before delivering you to an irrelevant solution, will shunt you to a landing page that forces you to log into your microsoft account before you can see the answer.

“We don’t understand. Why aren’t people simply searching for Taylor Swift”

For real!

Just kind of dawned on me while looking at the number, Reddit’s licensing deal with Google is valued at $60 million per year. That’s really not very much money at all, considering the amount of data Reddit has and continues to accumulate. And chump change for Google, no doubt. Reveals how little leverage Reddit actually has at this point. This was their flagship deal, and the best they could get was $60mil per year.

Also puts the API fiasco in a new light. “Look, we need to charge for API calls, because we need to restrict public access to data as a precondition of selling all your shit in a few months to Google, for the financial equivalent of a cup of coffee.”

Think it’s more of an allusion to lurking habits, active times, metadata, stuff not related to public posts. I’d imagine the average user has plenty of stuff they’ve browsed through that they wouldn’t want their family / co-workers, etc. to know.

Would also need to get a burner phone number w/ answering machine to take calls from 240 million grandmas, cheapskate businesses and cash-strapped public schools for any & all tech support questions until the end of time, because if there was an issue with system stability in any way whatsoever, or if the router went down or the printer stopped working, they’d assume it was the fault of ‘the guy who changed everything’.

Linux is great & everything, but this sounds like a recipe for utter disaster, not a way to make an easy buck.

Interesting. I’m curious to know more about what you think of training datasets. Seems like they could be described as a stored representation of reality that maybe checks the boxes you laid out. It’s a very different structure of representation than what we have as animals, but I’m not sure it can be brushed off as trivial. The way an AI interacts with a training dataset is mechanistic, but as you describe, human worldviews can be described in mechanistic terms as well (I do X because I believe Y).

You haven’t said it, so I might be wrong, but are you pointing to freewill and imagination as somehow tied to intelligence in some necessary way?

Thanks! I’m not clear on what you mean by a worldview simulation as a scratch pad for reasoning. What would be an example of that process at work?

For sure, defining intelligence is non trivial. What clear the bar of intelligence, and what doesn’t, is not obvious to me. So that’s why I’m engaging here, it sounds like you’ve put a lot of thought into an answer. But I’m not sure I understand your terms.

Not being combative or even disagreeing with you - purely out of curiosity, what do you think are the necessary and sufficient conditions of intelligence?

I’m glad to hear this and hopefully it fulfills my fantasy of having sex with Clippy.

Interesting.

I hope for a new paradigm in web searching. I wouldn’t even mind if a search took 5-10 minutes, if it meant a handful of quality results. I easily waste that much time or more sifting through garbage ai and ad-driven results as it stands.

All I want is a search engine that 1. doesn’t make moral judgements on the results relevant to a search, 2. filters out ai and ad farm results by default, and 3. can be toggled to effectively search web 1.0-style forums.

Microsoft OS workload on an AI-optimized chip:

(5%) consumer benefit - users can get access to Clippy+ with a Microsoft premium account subscription, that if users aren’t subscribed, they’re reminded every time they go into the settings application

(15%) anti-piracy & copyright protection

(70%) harvesting and categorizing all user activities, for indiscriminate internal use, sale to other companies, and delivery to governments

(10%) Uninstallable OEM bloatware that does the same, but with easily exploited security flaws that are never effectively patched

Not sure if you already know, but - sophisticated large language models can be run locally on basic consumer-grade hardware right now. LM Studio, which I’ve been playing with a bit, is a host for insanely powerful, free, and unrestricted LLMs.

Excellent point.

I think this type of anthropocentrism extends to chess too actually. I’m not an expert on the subject, but I’ve heard that chess AIs are finding success doing unintuitive things like pushing a and h file pawns in openings. If, 10 years ago, some chess grandmaster was doing the same thing and finding success, I imagine they would have been seen as creative, maybe even groundbreaking.

I think the average person under-rates the sophistication of AI. Maybe as a response to the AI hype. Maybe it’s because we’re scared of AI, and it’s comforting to believe that it’s operations are trivial. I see irrationality and anger cropping up in discussions of AI that I think stem from a fundamental fear of its transformative power.

NOC’s vocalizations were recorded and studied by a team of biologists from the National Marine Mammal Foundation (NMMF) led by Sam Ridgway. In 1984, Ridgway and others at the NMMF began to hear peculiar sounds coming from the whale and dolphin enclosure. They were reminiscent of two people talking in the distance, the words just beyond the limit of comprehension.[5] Later, a diver working in the enclosure came to the surface after he heard someone cry “out, out, out!”[1] After he asked his colleagues “Who told me to get out?”, they realized it had been NOC.

https://en.wikipedia.org/wiki/NOC_(whale)

$330m is not nothing. But, with a funding split between a telecom CEO, and a shipping & logistics CEO - person has to wonder what sort of direction & tuning the team might be encouraged to explore. How will they stack up against existing & proven open source non-profits with impressive releases like EleutherAI?

These open source projects are neat, in that they give the average person the opportunity to peek under the hood of an LLM that they’d never be able to run on consumer level hardware. There are some interesting things to find, especially in the dataset snapshots that Eleuther made available.

In general, kind of cool to see France being on the cutting edge of these things. And I think it’s worth saluting any project that moves to decentralize power from states and megacorps, who seal wonderful, powerful things in black boxes.

Thanks for the links, I’ll check them out. Persistence of data seems like an interesting issue.

Nobody would defend copyright if it wasn’t already in place

I don’t know about that. Say you take a few years to write a handful of poems, and it turns out people in your neighborhood really like them. You compile the poems into a book, and sell it for $5, and it sells well. Seeing this, your neighbor buys one, copies it, and starts selling it one neighborhood over for $2, and representing themself as the author. I would think most people in that situation would want to say, ‘hey, that’s not fair’. I don’t think that’s sick or rooted in greed, copyright can be a check on greed.