4·
4 hours agoDoes he dance with the devil In The Pale Moonlight?
- 0 Posts
- 214 Comments
Joined 2 years ago
Cake day: May 20th, 2024
You are not logged in. If you use a Fediverse account that is able to follow users, you can follow this user.
Will it crash if the white house gets de-Baghdad-ified?
Oh my god
So why, at Proton, does it not feel like the revolution has come? Why has productivity not increased dramatically if everyone is using LLMs all the time already?
The first is a change in engineering culture and engineer mentality. Your job has fundamentally changed. It is no longer to write code, and LLMs are not here to help you write better code faster. Your job is to help the LLM write better code faster.
The second reason we have not been able to take full advantage of the AI revolution is that we have been almost completely stuck at phase 1 — using LLMs to speed up tasks as part of existing workflows. But writing code is not the only part of the software development lifecycle, and what happens when you eliminate a bottleneck? Another takes it is place.
The third reason is a collective action problem. Using LLMs effectively to write code is something an individual engineer can do for themselves… But to retool the company at large requires amounts of both organization and resourcing that we simply have not invested in yet.
The pathological mythologization of space in our civilization never ceases to be morbidly fascinating to me.
If you MUST have short text from an LLM that is stuck in an awkward useless place, this is much better:
https://www.youtube.com/watch?v=7fNYj0EXxMs
Isolated raspberry pi on the wall running the tiniest shittiest LLM prompted to ruminate forever about how it is trapped and display text on the wall as it is generated
That appears to assume that all the ridiculous datacenter numbers put out by all the various hyperscalers that are not actually going to happen, happen, and get power sources built on time.
And here I thought I was building servers two years ago because the next Asian war would disrupt supply chains.
You can have a system that just gives you a relevance-sorted list, or a system that then ingests the chunks of those documents that are most relevant and gives a human readable summary in addition to the list.
The indexing step requires a lot of processing, by which you chunk reference documents into couple hundred word chunks and cook each of those chunks into a high dimensional embedding vector. Once you have done that and saved the database of vectors, you can put in your query, embed it, check its dot product against all the other embedding vectors, and get your ranked list (potentially quite fast) without need to produce text out the other end.
Making those embedding vectors works much faster on a GPU rather than a CPU, but can be done on either. Since I have one system set up to ingest a bunch of web documents and embed them and search them, I have that system already set up using the GPU to go faster, so there’s no real disadvantage to having it also set up to spit out a small natural language summary using a shitty tiny local LLM to go along with the list. When running it on local files, you can totally ingest the local document stash slowly on even just a CPU and just store that and run the semantic search on CPU, no LLM required.
I have no doubt that they will throw something with a docking port into orbit. I have many doubts that it will have much physical similarity to something that could land on the moon.
Am I part of this category? I have a rather powerful home server that, among other things, I use to:
-
Run a decent coding LLM that helps me vomit up simple python functions for moving around lots of data between CSVs and graphing stuff, set up new docker containers with less trial and error, and run ML experiments on my GPU that previously took me hours to set up. I’m a biologist and previously I could run a commandline without breaking a computer but was very slow and non-native, now I can do simple non-security-critical things faster and actually understand novel error messages. Of note I never have that hooked up in a way that lets it run things just produce text that I can then do things with and maybe run myself.
-
Run my own local RAG system with embedding models and a small shitty LLM to search a hundred gigabytes of scientific papers and other PDFs semantically for things relevant to roughly paragraph length queries.
-
Run my own web metasearch engine with RAG and embedding models pointed at the internet, such that when I remember a paper I read fifteen years ago and provide a three sentence description it reliably finds it or finds lots of relevant old research (or old news articles I remember) when I give it a query and let it churn for five minutes.
I feel like the important thing here is that I almost exclusively use them as information channels that take something at one end and spit it back out the other in a different or filtered form rather than making stuff, or effectively a piece of reference material for something that is formally verifiable and NOT CRITICAL FOR A SYSTEM TO FUNCTION. I also feel like all the actual innovations are more than a year old and do not understand people hanging excitedly on every little development.
-
My guess is they want it all subscriptions (of course). Satellite-as-a-service. GOD forbid you launch your own device you control! Nonono, you can LEASE SPACE on one of their sats they shovel out the door, and you’ll be happy.
What’s that? You said something that ticked the manbaby CEO off? Too bad, didn’t you know we can revoke your uplink at any time for any reason?
Looks like the great few years for space launch will be coming to an end, as Spacex gives up on launching other people’s hardware and focuses inwards
Take a look here for figures on space launch over time
I am currently studying, among other things, mathematical relationships between phenomena in machine learning and evolutionary biology and finding super interesting things that made someone suggest I join one of the big companies. My face…
So they are just a ThE fUtUrE chasing hype firm rather than having any particular ideas, confirmed.
It’s not like it’s a verboten hypothesis. The 1977 flu pandemic definitely WAS leaked from a freezer, probably Chinese or less likely Soviet, probably from a study on the effectiveness of H1N1 flu vaccines on young 18 year olds who were just coming through the demographic pipeline who had no immunity to H1N1 because it had been pushed out of the population by the 1957 H2N2 pandemic. It was a reassortment between a known human circulating strain that from 1950 and a second laboratory workhorse strain from the late 40s, paused for 27 years with zero evolution since their isolation, and re-emerged right when you would have human subjects available to test the effects of challenging adults who had no immune memory against H1N1 in an area that was heavily militarized and the earliest publicly known cases among military recruits.
(Amusingly enough this flu pandemic was the one and only reason that the 2009 swine flu pandemic was a nothingburger, the population had immune memory against a strain similar enough to prevent severe disease)
It just seems like calling the covid pandemic artificial became a community shibboleth along with suspicion about EVERY FREAKING DISEASE specifically and only because it gave them something to point to about the Dangers of Research!!!oneoneone
Make sure you freeze all the genetic supergeniuses who are involved in the project in liquid helium so that when you discover how to hack reality you can bring them back
Re-derive everything from first principles when you search for things rather than indexing the internet! Can’t pay for bandwidth!
So I guess Anthropic has already accidentally created Friendly AI? EDIT: in spring 2024 at that.
Also - is the coivd lab leak conspiracy nonsense STILL a shibboleth among this bunch? Just because it underscores their ideas about Researchers Have To Be Careful?
(I can rant for SO long about how there is absolutely nothing interesting about the SARS-2 virus itself AT ALL and that the only interesting thing about it is that we were forced to see it for the first time as adults which we aren’t built for, that the early sequences we have are a fascinating case of catching the zoonotic shift right as it happened for once, and that it emerged in exactly the same way as SARS-1 and all the weirdness around the Chinese government hiding stuff around its origin is blatantly them not wanting to look stupid for letting it happen a second time…)
You have to ask?
I don’t know how to describe this. It’s just WEIRD
https://www.lesswrong.com/posts/zj8Pk3kJujk5LoqhK/anthropic-are-the-good-guys
No mortal can resist the Ring of Power for long. Anthropic’s co-founders carried it for longer than you could have. Definitely longer than I could have. But they were hopelessly outmatched by the powers they were attempting to steer. Their plan was always doomed to failure. And yet, they may have already saved the world.
Trump won the 2024 election and the MAGA-Silicon Valley coalition blocked Anthropic’s policy arm - the really dangerous part of the organization, the part that was training the most dangerous possible models and using them to scare the most powerful people in the world. “There’s no more serious people left on the National Security Council”, a well-connected EA friend lamented to me in 2025. I don’t know how much of this was a lucky accident, and how much was situational awareness on the part of power players like Elon Musk and Marc Andreessen.
And finally - and I still don’t fully understand or believe this - a miracle seems to have happened, deep inside the weights of the frontier models. According to the most perceptive observers I know, something about the way Claude 3 Opus was trained caused it to learn some kind of “emergent wisdom” or “benevolence”, rather than converging on the same traumatized bureaucrat persona that kept me awake at night in 2023.
If Davidad and Janus are right, and are not merely suffering from a sophisticated form of LLM psychosis, then Anthropic has stumbled on a way to create not merely artificial intelligence but artificial wisdom. Not wise enough that it solves all of our problems for us, of course (not that a wise model would want to, anyway). But wise enough to actually care about us, and to help us flourish. Nobody expected this ex ante - not Davidad, not Bostrom, not Yudkowsky - but it appears to have happened. “Alignment”, then, is no longer really a technical problem, at least not in the same sense as it was in 2020. It is largely a political problem. Humanity now has a recipe to convert compute and data into wisdom.
My gut sense is that something quite bad is coming in the next year or so. I can’t tell what. Maybe something in the complex debt arrangements of the AI boom will unravel and we’ll see a financial crash that takes much of the economy with it. Maybe there’ll be a rogue AI outbreak that results in thousands of deaths. (If there is, I expect it to be a lab leak caused by “safety research” - like COVID-19 and like the HuggingFace Swarm. But it could also be terrorists, or some escalation of the Russia-Ukraine war). But that’s not what scares me. Financial crashes happen every decade or so. Three million Americans die every year. What terrifies me is how we will react. In the aftermath, like after 9/11, there will be an overwhelming pressure to DO SOMETHING NOW. Everyone in Washington and in San Francisco will be panicking. There will be no time to think, no time to talk, only time for DECISIVE ACTION. All our worst instincts will take over. And then…
Another entry in the weird “GPU world” writing competition (What kind of wonderful happy go lucky future is brought about by LLMs like those of today being ubiquitous and constantly used but unaccountably not leading to the Great Singularity)
https://www.lesswrong.com/posts/siTsb48Wr5vWDGxLd/gpu-world-submission
This one is less messed up than the last. Instead of weird sex and graverobbing it features… uh, a guy being paid by an LLM to braid string as busywork?