I agree keyboard remapping is huge, and I use AHK, but I’m annoyed every time I mess with it—the service crashes sometimes, the config syntax sucks.
What I’m trying to say is that it’s much harder to do AI alignment research while models are still small, so TAI timelines somewhat dictate the progress of AI alignment research. If I wanted my 5 year plan to have the best chance at success, I would have “test this on a dog-intelligence-level AI” in my plan, even if I thought that probably wouldn’t arrive by 2036, because that would make AI alignment research much easier.
I think that if we saw the working AI alignment solution used in 2050 in a paper written in 2026, we wouldn’t be confident it would work. That’s because there are a lot of uncertainties about how hard the AI alignment problem is in the first place, how ML behaves when it’s scaled up, ect. I think most plans for AI safety need to go like “we make the theory now, then we keep working on it as ML scales up and adapt accordingly”.
Here is an argument I’ve heard for why we shouldn’t try to solve AI alignment super early:
If you aren’t one of the top few AI safety researchers in the world, then you are far more likely to solve AI alignment if you spend some years to develop your skills first. Therefore most people in AI alignment should forsake some early timelines (like anything before 2040) and optimize for their impact once they’re a senior researcher.
This would be false if either less experienced AI safety researchers were able to contribute to completing AI alignment in 5 years, or if they can develop skills nearly as well working on a 5 year alignment plan as they could just optimizing for learning. I think both of these are somewhat true, which weakens the argument for me.
If consequences are completely ignored, I lean towards the torture, but if consequences are considered I would choose no torture out of hope it accelerates moral progress (at least if they had never seen someone who “aught to be tortured” get away, the first one might spark change. which might be good?). In the speck case, I choose torture.
I’ve noticed that a lot of blogs in the rationalist diaspora lack favicons. This often makes it difficult to navigate tabs for different blogs, PDFs, ect. A favicon takes <15 minutes to make, saves your readers time, and improves your appearance. A win-win-win.
Maybe people think it’s cool to have no favicon, or they want to stay on the down-low, or are ultra-minimalist, but, like, seriously?
We could dub this “Long Tail Externalities”—the idea that most of the impact comes from a few indirect effects, and sometimes the more indirect the bigger—for instance, most policies might impact the future mainly through AI safety.
I’ve been reading / light working in virtual reality for ~2 hours/day for a few months. My setup is a bit different than Ozzie’s, so I thought I’d share that here.
I use a Windows gaming laptop with the Oculus Quest 2, connected via Virtual Desktop instead of Immersed. Virtual Desktop has far lower latency, and slightly higher resolution than Immersed, but only shows you one screen at a time. The latency is low enough to be practically unnoticable (smaller than latency differences between similar computers, https://danluu.com/input-lag/) . A few months ago, it was Windows only, but since added Mac support. I don’t think the GPU on my laptop is important for my setup.
The main reason I use Virtual Desktop is to get out of my chair, either to lying down or standing up. When I hava a stomachache, headache, or soreness, I get a significant productivity benefit from being able to work lying down. By default my VR productivity is a bit lower than normal, but more enjoyable.
The biggest downside I’ve found is that VR definitely gives me myopia, due to vergence problems. Whenever I take off the headset, my distance vision gets blurrer for a few hours until my eyes adjust to focusing different distances. I think this gradually leads to long term myopia if not offset by vision exercises or something, but will be sovled in future headsets. I’ve tried placing my virtual screen at the focal distance of my headset, 1.3 meters, but that didn’t help.
Virtual Desktop takes many seconds to connect, and sometimes needs to be restarted, which is a significant flow-disruptor.
Overall, I’ll continue to use VR the same as I do now, and hopefully up my VR time/other screen time when better headsets come out.
I definitely agree with Pinky on the nanobots. Biology is already highly optimized nanobots, and it has a lot of limitations. I don’t expect self-replicating nanobots that far exceed life using ordinary chemistry/physics type stuff.
“Is Soares perhaps suffering from the common misconception that words can only have a single definition?”
Personally I like words with single definitions. For example, I’d rather the mathematical objects Set, Group, Category, ECT had their own made up names than borrow words used for other things.
These problems sound significantly worse in Sweden than in the US. For example, roof replacements cost 8-30K out of 30K median income here. The problems you talked about are real and fundamental, but I think low hanging problems with the market in Sweden are responsible for most of the cost.
Yes journalism does some useful research, but on the average newspaper frontpage you see 0 remotely useful research. The amount of money spent by newspapers isn’t at all in proportion to the amount of research being done
I slightly disagree that “producing most of what there is to read requires money”. Yes most writing is made for profit, but the very best writing I is often made because the author needs to say it, not because they expect compensation. Definitely need to look at evidence on this, though.
If bad writing is helped more than good writing by ad revenue, then it’s less clear that advertising is good.
Bow and stick
Hand spin stick
Flint and steel
Flint and magnesium
Break light bulb, use filament
Stick fork in wall socket, use spark + cotton or whatever
Strip power cord, make spark in more convenient location
Light candle from another candle
Light from stove
Light from oven
Put fire starting stuff + 2 or more pieces of metal in the microwave
Put candle on tree or roof, hope lightning strikes
Magnifying glass + sun
Spark plug from car engine
Light from exhaust of car (may need to tune car first to make exhaust fiery)
bring candle to the surface of the sun
give the candle to someone else and hope they eventually light it for their own purposes
Leave candle in public space, hope someone takes it and lights it
Light off of wild fire
Put candle in box with air and throw into deep space, hope it eventually quantum tunnels into a lit state
Light it off of dark thing in hot sun
Light with electricity generated from treadmill / stationary bicycle
Light with electric only cigarrete lighter
Break natural gas pipe, hope something ignites gas + candle eventually
Light off of electric heater
Light with exhaust from firearm
Stuff candle in barrel of firearm
Drag wick between two pieces of paper, maybe lights from friction
Cut some of candle off of wick to light more easily with above method
Write proof that simulation exists that somehow depends on candle not spontaneously lighting, then simulation controllers will light candle for you to protect simulation secret
felt like i had to because everyone else uses simulation
Put candle inside LHC
replace duterium pellet in ignition fusion experiment with candle
light candle with plasma from tokemak
Break candle into lots of short candles to try more stupid candle lighting methods on it
Replace wick of candle with match, then strike match. Does this count as a lit candle?
Put candle under desk lamp. Candle is now ‘lit’
Become very excited about prospect of lighting candle, thusly rendering it ‘lit’
Touch candle to antimatter candle
Disassemble candle, take one fiber out of wick with a little wax around it to make smaller candle. This new wick is smaller, lights easier with flint+steel or other
Light cadle with cap gun
Light with wheellock pistol
Light with explosive detonator
Light candle from across the room with infrared laser, looks like magic
Throw candle out of space station, eventually burns in atmosphere
Light with rocket engine
It turns out the candle is just a picture of a candle, draw yellow + red on it
Throw picture of candle in fire
Use candle as wad of shotgun
Throw small square of high friction paper at candle wick, cutting it in half and lighting it with friction in the process
Remove battery from phone, touch battery leads together to start fire
Climb power pole, make spark to light candle
Put flammable oil something in saucepan on induction stove, light candle
Use pascal’s mugging to coerce someone into lighting candle for you
Make advertisement for candles, causing people to buy candle and light it
Manufacture more candles, so there’s more chance one of them will be lit
Keep candle in secure vault prior to lighting so it can’t be destroyed first
Read literature on formal definition of fire so you can be completely sure whether the candle is actually lit
Submerge most of candle in cold water so you don’t accidentally melt all the wax off in your attempt to light candle (would be a shame to fail like that)
Conduct a thorough chemical profiling and and CAT scan of candle to make sure there are no explosives hidden inside before you light it
Teach humanoid robot to light candle for you
Make pre-lit candle dispenser with built in lighter, then load it with the candle
Light candle with controlled gas lighter in QA testing at candle factory
Write in your will that you want your candle to be lit
Fire your candle, hire a new candle and light that one
Hope cosmic rays light your candle
Fly in hydrogen balloon to increase your chance of getting lit by cosmic rays
Blow up said hydrogen baloon to light candle
Submerge your candle in liquid oxygen, wait for candle to spontaneously combust
Find bacteria that can eat your candle. Its calories will have been ‘burnt’, so therefore the candle must have been ‘lit’.
Drop candle in lava lake
Drop candle in molten metal in foundary
Touch wick to heating elements in toaster
Just put the candle in the toaster
Drill holes in candle to ‘lighten’ it
light candle with blowtorch
Light candle with fires in aftermath of nuclear explosion
Shove stick through candle and roast like marshmellow over campfire
Time travel back to when the candle was still petroleum in the ground, blow that up with nuke
Build a wooden bridge from an already lit candle to the target candle, which burns and lights target candle
Lay additional candle wick between other lit candle and target candle
Light candle with plutonium radiothermal power source
Light candle with heat from geothermal water based power plant
Rub hands together so hard they light candle
Light candle with dozens of high-wattage spotlights
Plant ultradense material inside candle wick, then light with high energy neutrino beem
Light candle with cold-candle-seeking missile
Tell people this particular candle is immune to lighting so they try to light it
Make real life fire breathing dragon, then have it light the candle
Do nothing, because after the candle’s protons decay they will enter a new Penrose cycle universe and become another candle that someone else will light
Trigger vacuum collapse. The energy released with the candle collapses is kinda like “lighting”
Throw molotov cocktail at candle
Focus sunlight with mirrors
Focus moonlight with mirrors onto steam engine, which generates electricity, which powers electric lights, which are focused with mirrors onto candle. (because moonlight is too low frequency to light candle)
Focus starlight with mirrors of ridiculous size onto candle
Burn candle in incinerator made for corpses
Light candle in pressure cooker
Boil candle in oil
Boil candle in water until water boils off and candle lights
Throw candle into accretion disk of black hole
Expose candle to radiation from pulsar
Light with jet engine afterburners
Light with one of those exothermic chemical reactions that’s not considered fire. (don’t actually rememer any)
Probability+utility theory might be recognized as important on its own, so there might not be a big difference between including it in Elements and publishing it as its own volume.
I like the idea of editing the Koran. It spread through conquest earlier in its life than the Bible, so perhaps it’s text isn’t as vital to its success as the Bible’s which had to spread organically more before it was spread by force.
There’s also the issue of great filters: if the great filter is in our recent past, then anything we change would be net negative, and we would be better off not going back far at all.
What do ants want? population growth, possibly: curiosity, much less signalling (in the way we know). Individual ants would work towards their queen’s best interest? Would they try to defect to other queens?
Digging (foundations of new houses and such) would be a bad idea.
Ants would naturally have a slow communication rate.
Ants could communicate faster than walking by flying, putting their pheramones in the air somehow, using human internet, using human intermediaries.
Ants can increase their population quickly. I would rough guess doubling every month at least for the first month
Ant main-line cognition speed would be slower than humans, but multitasking greater. Therefore they would likely execute multiple plans at once instead of making one big plan.
Ants would look for technologies to enhance their cognition even further. Could be dense, 3d habitats that allow the colony to live closer to eachother to reduce transmission time.
Increasing speed of individual ants. I don’t think ants tend to travel at even half of their max speed, so they could think faster by just burning more energy in a way humans can’t.
Ants are everywhere. Some humans would try to escape ants by fleeing to cold places with permanently frozen soil, like northern Canada. Because of their small size, ants have extreme difficulty in those conditions. Ants couldn’t go there without vehicles.
Given that ants primarily sense touch and smell, they may still lack good eyesignt and hearing, even with better communication. This means they may be hard for them to read computer screens and eavesdrop on humans. For example, I doubt a single ant could spy on a computer screen. Perhaps 100 or more ants would be required for that. Same with hearing, although it would be easy for 1000 ants to hide within earshot of human conversations.
Therefore humans may communicate more by text to hide from ant spies.
How would ants fight against poison? My guess is by just attacking until poison is used up.
Humans need to prepare for war with ants before they show signs of aggression. They could be too powerful for humans to stop.
Ant colonies would begin communicating with each other. An ant can travel at perhaps 3km/hour. Assuming this is their mode of communication, somewhere like the Amazon would take like a year to send a message across. I don’t think it’s reasonable for ants to first-contact “native” ants through a faster means, just because they wouldn’t be able to decode it.
Would ants resent humans for often carelessly killing their kind? They may not consider worker deaths very important, but human activities would definately “anger” (anthropomorphizing) the ants before most humans knew ants were smart.
Creating supersmart supercolonies. I think this is reasonable, but linear processing speed would again be slower the more space the colony took up. Because of the speed limitation, it would take a long time before we have to worry about a global coordinated attack.
How would ants and humans communicate? By sound. Human microphones and software would detect ant vibrations, and ants would either learn to understand human speach, or hear ant noises produced by translation software. Probably the latter, because humans have vastly more expertise than ants.
Individual ants live around 2 years, queens up to 30. Intelligent ants would merge colonies whose queen dies, or produce extra queens to keep colonies alive. “individual” ant colonies could outlive humans.
Would ants war with other ant species (more than they already do). I think it’s likely the first thing that’ll happen. Bigger colonies, with more intelligence, would quickly kill off smaller ant colonies.
Which ants would win the ant-wars? Depending on how ant intelligence works, either smaller ants that can get closer together, or larger ants with more complex behavior, would come out ahead. Large ants could have advantages like better eyesight, wings, that become vastly more useful with intelligence.
Ants would destroy their predators and overeat their food. Much like humans, who killed off large mammals, ants would kill off all their predators and risk over-harvasting their food. They would start both farming and stealing from humans as well.
Ant predators include spiders, other insects, reptiles, and amphibians. This would cause some sort of ecosystem collapse.
Eventual ant technologies: ant computer interface, may be like a braile display that pokes different individual ants, or ant sould based. Because ant colonies would be good at multitasking, they could absorb material faster than human, given that it’s not all highly connected.
Human would publish this, be highly questioned. Would produce simple tests individuals could make at home and give to ants to get idea through to people.
Govts would probably put up a far worse performance than COVID. Politicians & beaurocrats are too old, too social to deal with this kind of stuff.
Immediately some young right-ists would call for the eradication of ants. It would take a few weeks for right politicians to follow suit, at least pushing for “safety against ants”
People would buy tons of bogus “ant communication devices” or “ant protection devices”
There must already be ant colonies in secure scientific experiments. They would be studied. New colonies would be taken into labs.
How much do ant colonies work with each other? Given that there are already ant-mega-colonies (their main trait is they “smell like friends” of each other), those ants would work together. They would likely work together with “smells like enemies” ants as well. I don’t expect average ant colonies to sacrifice for each other immediately, but maybe later?
Can ant colonies “mind meld” with each other? Given that the colonies themselves started working together intellectually a week ago, I think it’s likely they could.
Ant cognition speed is the #1 variable to measure.
Humans might engineer ant-to-ant comms devices to prevent ant wars, make ants more predictable, and spy. All ants using bugged (get it) comms would be awesome.
Food left out in human homes would be quickly taken by ants. Ants would quickly bite through fridge seals and eat there too. Ants could use leverage to open alumanum cans and plastic boxes.
People would find ways to make ant poison at home
Humans would disrupt ant hearing and smell with ant-verbalization noise generators or some homemade chemical that confuses or dulls ants’ senses.
Mosquitos developed resistance to DDT in “about 7 years”. Ants different reproduction. Because of investment in queens, they can’t experement genetically with them. Would need to test ant males. Could test males for pesticide resistance before mating, accelerating that evolution. ant males never go outside normally, so that would improve a lot.
There are roughly as many ant colonies as humans, so some EAs would immediately help ants. For short-termists, ants would likely be #1 priority by far. First would give comms and teach.
Evangelical Christians would attempt to convert ants. I would ROFL soooo hard if that actually worked.
Once ants learn the basics of human infrastructure and society but before they invent their own technology, what’s their best bet to defeat humanity? There are on the order of a million ants per person. I’m guessing that they wouldn’t be able to overpower humans just by swarming and biting except in very rural locations. If they did, they would strike at night. They can be silent, so you may 100,000 ants in your room without waking you up. They could suffocate you with their bodies, or with some tools. I think suffocating people at night with cloth/bodies would work. Even better, they could cut wires in cars / short circuit them, and swarm power plants / relay centers. Perhaps gnaw down trees to fall on power lines. Given the time taken to get power back after storms, and especially the Puerto Rico disaster, I’m not optimistic we could ever regain the grid. Again, I think they would take a while to coordinate that well, but we would need to secure the grid FAST.
Ants would hop onto cars, trains, boats, and eventually planes to explore & communicate with each other. Ants communicating this way may be able to jump off cars in large enough numbers to communicate, especially at lights & stop signs. Otherwise get off at the destination, talk to local ants, then all move out on new cars. This would be the fastest means of ant communication. Travel times from london to any inhabited place in the world are around 1⁄2 to 1 day. Ants would require maybe 4 extra hours to crawl between people & vehicles in airports and such. Some areas don’t have enough car traffic to easily hitch a ride to, so that might cost an extra day if they aren’t able to call taxis or manipulate humans into driving places or whatever. It would be reasonable for them to get 1-2 day global communication by physical transport.
If the time machine is a single-use object (rather than a spontaneous event), I would wait as long as possible before using it. There are a couple reasons for this: there are certainly decades of “Historical Priorities Research” to be done to find the best intervention, the actual traveller would require years of preparation, and we would have access to more technology, should we choose to bring it back. That’s given that people will devote their lives to research with likely little proof that the machine is legit, but this research is already somewhat on the EA agenda anyway. During that time people could be on guard to flee to the past in the event of a true catastrophe. Even if there was a nuclear war, I don’t know whether we should go back and prevent the war, or whether our historical plans would be more valuable. In either case, they would want to be ready to leave if anything happened to Earth. I would shoot for a reaction time of about 12 hours, and a rugged computer that’s constantly synced with general knowledge to bring with them, such that the team could learn more about the catastrophe after a quick escape.
Edit the Bible. It is the information replicated the most times throughout history, and thus it’s probably the best vehicle for a cultural or intellectual agenda. Finding the right edits would not be easy, because the bible would need to retain the qualities that made it so viral in the first place.
Edits could include reducing mysogeny/anti-LGBTQ, valuing the happiness and suffering of all beings, and putting more faith in reason. Adding more reason could easily undermine the persuasive power of the bible, but something could probably be done.
The bible was written between 0 and 100 AD in Greek, so “the team” of time travellers would need to learn ancient Greek (the known parts now, all the unrecorded parts when they arrived), go back to either 1 or 2 bc and influence early manuscripts / verbal recitations, or perhaps arrive around 50AD and write the official Bible, or influence those who wrote it.
If observers are distributed with constant density in 3d space, then adding 3 extra bits to the position of a camera covers 8x the volume, and thus 8x the observers, so the scaling is the same as if you were just numbering the people