Anders Sandberg Profile picture
Academic jack-of-all-trades.
Oxygen8 Profile picture Jason Hoelscher-Obermaier Profile picture 2 subscribed
Mar 11 8 tweets 3 min read
Two days ago I (like most people) had never heard of SVB. That we tend to discover systemically essential points of failure by them failing is deeply disturbing, especially if we want to make a more resilient world. Image In 1993 the world discovered the hard way that a speciality resin used to affix integrated circuits was mostly produced in a single chemical plant. That had an explosion destroying production and stores.…
Feb 21 5 tweets 2 min read
Saying LLMs are just glorified autocomplete forgets that most human thought, action and speech is also glorified autocomplete.
Maybe the fraction that isn't autocomplete is the truly valuable part, but I suspect it is not a major fraction of everyday useful action. I am pretty serious about this: the basal ganglia action pattern generators look like they are selected by softmax influenced by cortical stimulus fit. They get made/updated by RL rather than backprop, but very much a stimulus-response chain.…
Jan 11 7 tweets 2 min read
Forwarding this on behalf of Nick Bostrom since he’s not active on Twitter. He wants to explain his views and publicly apologize for an old email that somebody might put out to damage him. The context is a thread on a mailing list in 1995 talking about offensive communication styles where he did the classic freshman philosophy student trick of writing something deliberately offensive to make a point. It didn’t turn out well (it never does): he promptly apologized.
Jan 10 9 tweets 4 min read
This is an important point. I *also* recall being super-skeptical about the JWST (I somewhat rudely asking John C. Mather if he could really get sigmas enough to handle the 344 single point failure - I owe him and the entire team an impressed apology). However... So, clearly sometimes good planning and design can make things that are exceptionally reliable. This seem to happen mostly when (1) underlying physics and technology well understood, (2) designs optimized for reliability, (3) error correction mechanisms in production and use. Image
Dec 16, 2022 21 tweets 5 min read
#FridayPhysicsFun – “Launch him/it into the sun!” My husband was griping over breakfast about the expression: surely this is not a good way of disposing of anything?… Is it actually possible to make something that will hit the “surface” of the sun?
Nov 17, 2022 14 tweets 3 min read
There is a lot to unpack in this interview. I can also see a lot of lawyers cringing.
(A little thread on FTX, what I think EA did wrong, and systemic risk)… I have, as far as I know, not been directly affected by the whole FTX implosion (no direct funding). But it certainly made my world a bit more turbulent. One should never let a good crisis go to waste: this is a good time for EA to do some institutional reinvention.
Nov 16, 2022 16 tweets 6 min read
This looks like the first step onto something interesting. Trying to get a review of lava redirection produces a paper trying to model a layer of lava. "In other cases it may be necessary to divert the flow of lava from one location to another, such as from a volcano to a settlement." Good to know it is just diverting things *away*.
Oct 28, 2022 21 tweets 7 min read
#FridayPhysicsFun - @KarimJebari and me have a new paper that is putting the longterm into longtermism. It is about doing good for the environment a billion years hence.… The main point is that if we survive or leave automation in place, can extend the lifespan of the biosphere by perhaps a billion years or more. Ecocentric ethics places value on the Earth’s biosphere, so it should approve of this. But this thread is about the physics involved.
Sep 21, 2022 6 tweets 3 min read
The more I learn writing prompts for AI-generated art, the more I note that it is a language based on clichés. ImageImageImage This is not a criticism! Clichés, tropes, standard imagery are easy references to high-dimensional vectors in concept space. "evil stepmother", "greg rutkowski", "baroque" and "central perspective" link to very complex things, and we can invoke them easily and conventionally. Image
Sep 14, 2022 5 tweets 2 min read
If this result holds up generally, merging networks trained on different parts of data looks feasible. Huge implications - privacy through federated learning, parallelization of learning, merging of em shards! The basic idea is simple: you can permute hidden layer neurons, so there are actually far fewer internal models than it looks. Training gets to one, with linear mode connectivity. Can hence interpolate differently trained networks if one is careful.
Aug 19, 2022 5 tweets 1 min read
Is there any good study of the average lifespans of villages, towns and cities? My impression is that they do not disappear very often, despite there being a fair number of examples of ghost towns (that are often lightly inhabited).… This is linked to my bigger interest in what determines the lifespans of social structures and projects. Generally, constant risk over time seems to be the generic case for states, empires, species and companies; increasing risk only in software and individual organisms.
Aug 18, 2022 5 tweets 1 min read
Nice thread about why democratizing AI makes more sense the weaker the AI is. But this only looks at "offensive" capabilities: clearly AI can also protect. The real issue might not even be the offense-defense balance, but whether defense is reliable enough. A world where bad actors occasionally have great wins may be worse than one where they can often gain small. Some credit fraud ok, not everybody's accounts drained.
Aug 12, 2022 18 tweets 6 min read
#FridayPhysicsFun - I wrote an answer on Physics Stack Exchange about moving the Earth outwards to compensate for solar brightening.… The first consideration of how to move the planet was made by Archimedes boast: "Give me a place to stand and with a lever I will move the whole world."… Image
Aug 6, 2022 7 tweets 2 min read
#FridayPhysicsFun – Coolest fact I learned this week: under some conditions light-emitting diodes can be more than 100% efficient, and act as refrigerators. Normally energy conversion introduces losses: there is a production of entropy turning high-quality energy (e.g. mechanical motion, electricity) into disordered low-quality (e.g. heat), and turning low-quality into high-quality is less efficient.
Jul 2, 2022 4 tweets 1 min read
There is an old Swedish rhyme about the cuckoo: "södergök är dödergök, västergök är bästergök, östergök är tröstergök och norrgök är sorggök."
Roughly: "south-cuckoo is death-cuckoo, west-cuckoo is best-cuckoo, east-cuckoo is consolation-cuckoo, north-cuckoo is grief-cuckoo". If you take the rhyme seriously each cuckoo hence radiates a kind of field in cells, like a cellphone tower for superstition. It presumably declines as 1/d^2 with the sound intensity.
Jul 1, 2022 16 tweets 6 min read
#FridayPhysicsFun – I got a question today: are there pyroclastic flows from volcanoes elsewhere in the solar system? Pyroclastic flows are the result of explosive volcanic eruptions, a mix of hot gas and volcanic matter hurtling downslope at potentially more than 100 km/h, burning and suffocating everything. Basically a landslide mixed with hot, dense gas and dust.…
Jun 29, 2022 4 tweets 1 min read
I knew my youthful forays into ceremonial magic and neural networks would pay off eventually... But not simultaneously, for the same application. I think the essay gets at something deep. We are not just being pre-scientific or pre-engineering with current prompt design ("ha ha, it's black magic"), but the domain has many similarities to magic because it draws on an overlapping source.
Apr 15, 2022 17 tweets 3 min read
It is interesting that the same people who would strongly endorse "societal change without caring about quarterly revenue" in the abstract likely oppose these particular people doing it. Meandering🧵 There is often (1) an assumption about what values are embodied by non-revenue aims, and (2) the assumption that such change must be joint by a group rather than unilateral by an individual.
Apr 8, 2022 11 tweets 4 min read
This Monday, I explained in the FHI kitchen why my AGI timelines were long. Over the *past three days* they have shortend a fair bit. It is not that I think PaLM, DALL-E 2 or Socratic models in themselves are huge steps towards AGI. It is that they demonstrate rapid improvement in three entirely different domains, looks like it can be generalised to a lot of tasks, and surprise "experts" like me.
Apr 7, 2022 6 tweets 2 min read
Using LAION for designing book covers looks somewhat feasible. It might not be DALL-E II, but it knows how book covers usually looks. Image I find the "spelling" and typography fascinating. Image
Feb 20, 2022 11 tweets 2 min read
Yesterday, when an friend heard I worked on the end of the world I was asked what I thought about the imminent social collapse. I was confused: "why do you think there is one?" The reasons given were basically that people are stupid and listen to stupid people, that civic virtues are decaying, and it all depends on unsustainable activities. Which has been the case for the past few millennia. Why *now*?