That’s a great line of thought. Take an algorithm of “simulate a human brain”. Obviously that would break the paper’s argument, so you’d have to find why it doesn’t apply here to take the paper’s claims at face value.
That’s a great line of thought. Take an algorithm of “simulate a human brain”. Obviously that would break the paper’s argument, so you’d have to find why it doesn’t apply here to take the paper’s claims at face value.
There’s a number of major flaws with it:
IMO there’s also flaws in the argument itself, but those are more relevant
Meshuggah:
https://www.youtube.com/watch?v=m9LpMZuBEMk
Listened to them before I got into metal, came back to them later and now love them. That’s from probably one of their more accessible records, they also have more experimental stuff like this:
This is a silly argument:
[…] But even if we give the AGI-engineer every advantage, every benefit of the doubt, there is no conceivable method of achieving what big tech companies promise.’
That’s because cognition, or the ability to observe, learn and gain new insight, is incredibly hard to replicate through AI on the scale that it occurs in the human brain. ‘If you have a conversation with someone, you might recall something you said fifteen minutes before. Or a year before. Or that someone else explained to you half your life ago. Any such knowledge might be crucial to advancing the conversation you’re having. People do that seamlessly’, explains van Rooij.
‘There will never be enough computing power to create AGI using machine learning that can do the same, because we’d run out of natural resources long before we’d even get close,’ Olivia Guest adds.
That’s as shortsighted as the “I think there is a world market for maybe five computers” quote, or the worry that NYC would be buried under mountains of horse poop before cars were invented. Maybe transformers aren’t the path to AGI, but there’s no reason to think we can’t achieve it in general unless you’re religious.
EDIT: From the paper:
The remainder of this paper will be an argument in ‘two acts’. In ACT 1: Releasing the Grip, we present a formalisation of the currently dominant approach to AI-as-engineering that claims that AGI is both inevitable and around the corner. We do this by introducing a thought experiment in which a fictive AI engineer, Dr. Ingenia, tries to construct an AGI under ideal conditions. For instance, Dr. Ingenia has perfect data, sampled from the true distribution, and they also have access to any conceivable ML method—including presently popular ‘deep learning’ based on artificial neural networks (ANNs) and any possible future methods—to train an algorithm (“an AI”). We then present a formal proof that the problem that Dr. Ingenia sets out to solve is intractable (formally, NP-hard; i.e. possible in principle but provably infeasible; see Section “Ingenia Theorem”). We also unpack how and why our proof is reconcilable with the apparent success of AI-as-engineering and show that the approach is a theoretical dead-end for cognitive science. In “ACT 2: Reclaiming the AI Vertex”, we explain how the original enthusiasm for using computers to understand the mind reflected many genuine benefits of AI for cognitive science, but also a fatal mistake. We conclude with ways in which ‘AI’ can be reclaimed for theory-building in cognitive science without falling into historical and present-day traps.
That’s a silly argument. It sets up a strawman and knocks it down. Just because you create a model and prove something in it, doesn’t mean it has any relationship to the real world.
Canonical lives and dies by the BDFL model. It allowed them to do some great work early on in popularizing Linux with lots of polish. Canonical still does good work when forced to externally, like contributing upstream. The model falters when they have their own sandbox to play in, because the BDFL model means that any internal feedback like “actually this kind of sucks” just gets brushed aside. It doesn’t help that the BDFL in this case is the CEO, founder, and funder of the company and paying everyone working there. People generally don’t like to risk their job to say the emperor has no clothes and all that, it’s easier to just shrug your shoulders and let the internet do that for you.
Here are good examples of when the internal feedback failed and the whole internet had to chime in and say that the hiring process did indeed suck:
https://news.ycombinator.com/item?id=31426558
https://news.ycombinator.com/item?id=37059857
“markshuttle” in those threads is the owner/founder/CEO.
On a related note, I think libraries do need a bit of a facelift, and not just be “the place where books live”. It’s important to keep that function, but also expand to “a place where learning happens”. I know lots of libraries are doing this sort of thing, but your average person is probably still stuck in the “place where books live” mindset, as you allude. I’m talking stuff like 3D printers, makerspaces, diybio, classes about detecting internet bullshit, etc.
Threads like this, with highly upvoted comments like
americans are more propagandized than they think citizens of the DPRK are
They also use sarcasm try to push the narrative that North Korea is actually just fine, OK?
Guys you don’t understand; the West has spoken; we MUST hate North Korea, our governments have already decreed it so.
Many of them are also seemingly physically incapable of communicating without hexbear’s custom reaction images, which is a weird behavior common to many cults. Makes it harder to communicate with the outgroup.
I think LW is defederated from them (or vice versa) so you can’t post over there, but for further examples, try making an account over there and saying that maybe, just maybe, Putin did a bad thing by invading Ukraine, and they’re defending an imperialist.
I’d be careful of pushing the narrative about computers not being a good choice for regular users. I’m going to channel a bit of Stallman and say that that’s how we end up without The Right To Read
For your bullet points:
GPU issues can be hard, but that’s not really Linux’s fault. There’s a reason this image exists of Linus giving nvidia the middle finger:
That being said, it’s getting better. As of this year, nvidia has started putting some real effort into making things work with wayland.
EDIT: I’ve found nirvana with NixOS, speaking of GPU drivers. I just add a few lines to /etc/nixos/configuration.nix
and it goes off and ensures that the nvidia drivers are present. I also run lots of CUDA stuff on top of that and it all works about as seamlessly as possible.
From here:
On occasion, a writer will coin a fine neologism that spreads quickly but then changes meaning. “Factoid” was a term created by Norman Mailer in 1973 for a piece of information that becomes accepted as a fact even though it’s not actually true, or an invented fact believed to be true because it appears in print. Mailer wrote in Marilyn, “Factoids…that is, facts which have no existence before appearing in a magazine or newspaper, creations which are not so much lies as a product to manipulate emotion in the Silent Majority.” Of late, factoid has come to mean a small or trivial fact that makes it a contronym (also called a Janus word) in that it means both one thing and its opposite, such as “cleve” (to cling or to split), “sanction” (to permit or to punish) or “citation” (commendation or a summons to appear in court). So factoid has become a victim of novelist C.S. Lewis’s term “verbicide,” the willful distortion or deprecation of a word’s original meaning.
I think !shortstories@literature.cafe would be a good place for it. The community sidebar says your own stories are welcome. You might want to add that you’re specifically looking for feedback
Use Tor for everything. Search for “disposable email”, find a service that you can use in Tor. Sign up through Tor using that disposable email address for any service that you want to post to. Be aware that some services try to deny access to Tor and/or disposable email addresses. Try a different service or a different disposable email provider if you encounter that.
You should define your threat model. Longer essays can probably be deanonymized with stylometry. The above will probably work fine up to maybe the NSA taking an interest in the origins of the essay. You can probably post something to the Fediverse and reputation-wash it to a larger audience by saying “look at this link that i have no affiliation with”, but it’s more likely that someone would figure out that it’s you. You can use the Tor method to post on Reddit, but many subreddits will have automods that delete posts from new/low karma users.
Unfortunately there isn’t one easy source that I’ve found. This is based on reading the stuff you linked to, as well as discourse/matrix discussions linked to from those sources. I compare it mentally to Guido van Rossum as BDFL of Python (though not any longer). He did a much better job of communicating expectations, like here
It made some people unhappy that there was no Python 2.8, but everybody knew what was happening. The core Python team also wasn’t surprised by that announcement, unlike with stuff like Anduril or flakes for the nix devs.
There was also a failure to communicate with stuff like the PR that would switch to Meson. The PR author should have known if Eelco broadly agreed with it before opening it. If there was a process that the PR author just ignored, the PR should have been closed with “Follow this process and try again”. That process can be as simple as “See if Eelco likes it”, since he was BDFL, but the process needs to be very clear to everyone.
Exactly, thanks. “politicking” != US political issues
My take on it is that the creator of Nix was very good technically but was not a good BDFL, and that was the root of the problem. He didn’t do a good job of politicking, stepped down, and now Nix is going through a bit of interregnum. I don’t think it’s likely to fail overall though, nixpkgs is too valuable of a resource to just get abandoned. I expect the board seats will be filled by people that know how to politick, and things will continue on after that.
Lessons learned is being a BDFL is hard. IMO Eelco Dolstra failed because he had opinions about things like Anduril sponsorship and flakes, and didn’t just declare “This is the way things are going to be, take it or leave it”. People got really pissed off because there wasn’t a clear message or transparency, which resulted in lots of guessing.
I like Bluey and metal, so this shirt is perfect for me
They won’t open source snaps because they want to control the snap ecosystem to make money off of it for an IPO
That’s an interesting comment from a guy that used to work for Canonical, and then went anti-snap pretty hard, to the point that he made this:
Not sure how ollama integration works in general, but these are two good libraries for RAG:
https://github.com/facebookresearch/faiss
https://pypi.org/project/chromadb/