one (1) good thing that may come out of OpenAI is thirst for electricity only satiable by fusion, just as demand for heat in UK could only be met by digging deeper for coal which ultimately spurred the industrial revolution.
Completely unintended side effect, but coal and oil might have created the need for massive rewilding efforts that might, ultimately not only save the forests, but winding cover back to pre-industrial levels.
what a sad statement on the human condition that all of the other needs for limitless clean power did not meet the needs to justify developing fusion, yet you think that AI will? Jesus wept.
yes. that's the game we're told to play, so might as well try to play it. there's literally truckloads of money in AI and the hyperscalers themselves point to power as a major issue for their datacenters, so why not try to allocate some of it to fusion?
> what a sad statement on the human condition that all of the other needs for limitless clean power did not meet the needs to justify developing fusion, yet you think that AI will? Jesus wept.
Well, AI has the promise to provide a supply of loyal slaves to anyone who can afford to pay for the electricity and compute. It's a capitalist's dream: with AI, they may never be forced by necessity to share a single thing with us poors again.
You're thinking of anything but capitalism. Capitalism means you don't have to rely on powerful people sharing for you to have things. You get them because people can make money making the same product cheaper and sell it to a lot of people.
That's why all the capitalist countries are the ones where we have to keep increasing the standard of living that's counted as being in poverty. Up we go.
> You're thinking of anything but capitalism. Capitalism means you don't have to rely on powerful people sharing for you to have things.
No. Capitalism means you need to rely on being useful to the people who own things. If you're not useful to them, they won't pay ("share with") you.
For a capitalist, employees you have to pay <<< loyal robot slaves. Once you have those slaves, I predict the economy will make an abrupt shift away from consumer goods to vanity projects.
People with things have to be useful to people who can do things as well. I work to get paid a salary; my employer pays me enough that I don't leave. The only exception to this is taxes, which don't require mutually agreed exchange.
> People with things have to be useful to people who can do things as well.
That's not super clear, but I think I get what you're saying.
My whole point is AGI breaks that idea, and frees capital from the need for labor.
> I work to get paid a salary; my employer pays me enough that I don't leave.
And when an AGI can do your job better and cheaper than you, your employer fires you and stops paying you. And all the other employers don't hire you because they don't need you either. Then, if you're lucky, you get to live on the dole, otherwise you (eventually) get to be homeless have the opportunity to try scrape by at the margins (maybe you can squat and live off a garden for a few years, until a solar megaproject evicts you from now unprofitable farmland). In all cases you're marginalized and economically irrelevant.
If no one is hired then the employers don't have anyone to buy their stuff. Employers only do well if they provide someone else a useful good or service.
What's more likely - not that AGI is likely, but still - is that people move into other jobs. In 18th century Europe almost half the population were agricultural labourers. Mechanisation reduced that drastically. That did not mean that other jobs weren't created.
> If no one is hired then the employers don't have anyone to buy their stuff. Employers only do well if they provide someone else a useful good or service.
You're still stuck with assumptions that are obsolete in this scenario.
In the AGI scenario, the employers that are dependent on consumer sales will wither and die, as consumer buying power shrinks due to unemployment. Eventually the economy will realign towards certain kinds of B2B sales (e.g. electrical power) and vanity projects.
> What's more likely - not that AGI is likely, but still - is that people move into other jobs. In 18th century Europe almost half the population were agricultural labourers. Mechanisation reduced that drastically. That did not mean that other jobs weren't created.
Not if the AGI can do all those jobs better and cheaper than most people (or even just good enough and more obediently). There might be a rump of exceptionally talented individuals who still could be employed like today, but that's just a tiny sliver of the population. There will also be some "entertainment" jobs, like prostitute that will remain as well, but given the vast decrease in individuals participating in the economy, the total numbers would likely be less than now.
Not everything is going to be a replay of the past. As they say, "past performance is not indicative of future results."
> In the AGI scenario, the employers that are dependent on consumer sales will wither and die, as consumer buying power shrinks due to unemployment.
Why? People will find ways to exchange value.
> Not if the AGI can do all those jobs better and cheaper than most people (or even just good enough and more obediently).
What does this mean? By AGI do you mean "cleaning robots" or "entertaining bartenders" or "live music" or "person who owns this house I want to rent" or "mind I will pay to learn from"? None of those sounds like anything to do with AGI, unless the AGI is housed in a robot that can clean things (and then I don't need AGI).
Also there will be a floor of jobs not worth doing with AGI because of the energy and maintenance requirements. AGI is not a magic wand. It's a specific thing. ChatGPT being able to spit out a decent but generic essay doesn't suddenly mean that all the crazy numbers of jobs everyone does will vanish.
>> In the AGI scenario, the employers that are dependent on consumer sales will wither and die, as consumer buying power shrinks due to unemployment.
> Why? People will find ways to exchange value.
Sure, but they'll have increasingly less to exchange among themselves. They'll have nothing to sell that the AGI-powered economy wants to buy, except truly limited legacy resources like land that can be gobbled up in one-time purchases. Eventually the AGI-powered economy will monopolize the resources that are useful to it, in a way that likely conflicts with the needs of now-obsolete workers (e.g. converting vast amounts of farmland to solar power megaprojects).
That's the end-state of automation, in our current social system.
>>> What's more likely - not that AGI is likely, but still - is that people move into other jobs.
>> Not if the AGI can do all those jobs better and cheaper than most people (or even just good enough and more obediently).
> What does this mean? By AGI do you mean...
I mean intellectual automation that can do at least what a typical person can do as well as they can or better. Eventually it means the automation that can do all the jobs (or even just enough of the jobs). Eventually you won't have a new job to move into once your job is replaced.
AGI will eventually mean there will cease to be a practical necessity to using human to do labor to operate capital. The capital will be able to operate itself on behalf of its owners. Once that happens, under the current system, the owners of that automated capital will eventually accumulate all the wealth of the economy, because they'll be able to sell without paying wages. Eventually they'll pivot to vanity projects and B2B sales among themselves.
> Also there will be a floor of jobs not worth doing with AGI because of the energy and maintenance requirements. AGI is not a magic wand. It's a specific thing.
Probably, but I expect even those will eventually disappear too, at least on the mass scale needed to support billions of people, during the later stages of the economic transition.
> ChatGPT being able to spit out a decent but generic essay doesn't suddenly mean that all the crazy numbers of jobs everyone does will vanish.
I'm not talking about ChatGPT, I'm talking about the utopia the AI folks want to create.
I hope that in your scenario that everyone that can afford this notion of yours receives a robot that at the minimum is as annoying as C3-P0 if not closer to a Jar Jar.
Or worse. Personally, I think social media has been a net negative. It was done intentionally by their makers.
AI seems like it's just a victim of that, but seeing as how they have stolen all of the data they've built their tools on, then of course it's going to be no better than social media at best
That's based on the (flawed, IMO) idea that fusion just needs more resources to go faster [1]. We won't have serious fusion before decades, it's just too late to save our energy (and climate) problem.
Better go with fission at this point (preferably 4th gen because uranium 235 is limited).
Hydro, wind, and solar backed by batteries looks like an ~90% solution to grid power / ground transportation reasonably quickly and we have enough fission power plants to make up the difference.
So we already have the short term solution, it’s really 25+ years out when things get more interesting. Existing nuclear power is going to get increasingly expensive to maintain and recent construction projects have been boondoggles. So fusion has a real shot here assuming the economics work out.
Fission has gotten safer as we’ve learned from past mistakes, but each of those lessens directly results in increasing costs. Not just in obvious ways but getting better at foreign material exclusion means it takes longer to do the same tasks. Multiply that by every significant indecent at any power plant and it’s no wonder things keep getting more expensive.
> So fusion has a real shot here assuming the economics work out.
They don't yet work out, and there's no evidence that they will. I would love it if they do, but I don't think past performance is evidence of future performance. We might run into a fundamental limitation at any moment, and that would be that.
Japan's median build time for fission is under 5 years[0]. If regulatory environments and engineering specialisms could be made to work, there's no reason (other than Greenpeace) that we couldn't massively curb CO2 production from power generation pretty soon; far sooner than we could do discovery and then build for fusion.
> Hydro, wind, and solar backed by batteries looks like an ~90% solution to grid power / ground transportation reasonably quickly
What? Currently, electricity makes for 20% of our global consumption. We're not remotely talking about replacing the 80% of fossil fuels with electricity, even with fission + hydro, wind and solar.
Batteries only work to store energy for a few days, not between seasons.
The reality is that we don't have a 90% solution to power. Not in the short term, not in the long term. Except if new technologies that do not exist yet appear. Have a look at all those huge boats that enable globalization: how do you propose we replace fossil fuels there? Or aviation.
The solution to the energy problem is to prepare to have (much) less energy. And a good way to prepare for that is to try to produce as much electricity as we can. And that quite obviously involves fission.
> What? Currently, electricity makes for 20% of our global consumption.
An apples to apples comparison gives very different numbers. A heat pump uses 1 kWh of heat to produce 3 kWh or more worth of heat. A furnace needs over 3 kWh worth of gas to produce just 3 kWh worth of heat.
An ICE engine is more extreme as extraction, transportation, refining, takes 1/3 of the energy in oil before you even out it in the gas tank. Net result under 20% of the energy in oil ends up being used at the end of the process.
> Batteries only work to store energy for a few days, not between seasons.
There’s no point in storing power between seasons, just add more generation. A seasonal battery storing 1 MWh gets used once a season. A solar panel only used in the winter is still useful for ~4h * ~90 days. But worst case a ~3kW of solar is equivalent to that 1 MWh battery at less than 1/100th the cost, and whisk generally redundant the rest of the year it’s still reducing outages.
> An apples to apples comparison gives very different numbers.
I don't see the relation with apples. If you take electricity where it works well, then it works well. But the fact that it accounts for 20% of our energy consumption today means that it does not work well everywhere. Try planes or merchant boats, for fun.
And that's not even mentioning that on those 20%, a good part is coming from coal.
> There’s no point in storing power between seasons, just add more generation.
You're saying "just waste solar panels during the summer so that you have enough during the winter", right? I thought it was pretty clear that wasting energy was not a good idea for the future.
Replacing an ICE with a EV results in a drop in energy by your calculations even if they are doing the exact same trip. Thus showing your argument is based on nonsense.
When someone burns oil in a car you measure the energy before it’s burned and therefore before engine inefficiency. If you burn oil in an electrical generator you measuring energy after the engine inefficiency.
Thus the amount of useful energy IE what people want in electricity vs other sources is closer to 50/50 than 80/20.
> You're saying "just waste solar panels during the summer so that you have enough during the winter", right? I thought it was pretty clear that wasting energy was not a good idea for the future.
People build grid infrastructure for the worst case. Nobody complains when a natural gas power plant is only turned on for 12 hours a year because without it you get a blackout. Hell dams build spillways that can sit unused for decades, you still need them.
Thus no the panels aren’t wasted, they are doing exactly the job someone built that infrastructure for.
> Thus showing your argument is based on nonsense.
My argument is that there is a lot more than just cars in the world. Even if Americans may not understand the concept. It's easy to say "replace oil with electricity, look, I have this one example where it works well". Then try to scale that one example, and then start looking at the rest. Again... planes and merchant boats for instance.
Many boats are going electric. Home heating, industrial processes, trains, mining, etc the vast majority of energy use you can swap without issue.
Rockets and big boats can swap to hydrogen with minor issues. Really aircraft are the odd man out, but remove bio fuels from other applications and you can largely replace aviation fuel.
After we drop CO2 emissions by 99% using existing tech we'll have decades to hit 100%.
> the vast majority of energy use you can swap without issue.
Then you completely misunderstand the scale of the problem.
> Rockets and big boats can swap to hydrogen with minor issues.
Say they can if they have the hydrogen, then you have to produce a whole lot of hydrogen and transport it for them. Do you know how inefficient that is?
Because you make it work for one does not mean that you make it work for the whole world. Your reasoning seems very naive.
> After we drop CO2 emissions by 99% using existing tech we'll have decades to hit 100%.
Except that the only way we drop CO2 emissions by a lot is with a ton of sobriety.
Saying we don’t have the infrastructure is meaningless when building infrastructure is part of my argument. The only question is if we have the technology, and yes we do.
For scale, 350 gigawatts of PV was installed in 2023 that’s enough to meet ~3% of the words 25,000 TWh annual electricity demand (after accounting for capacity factor) and the rate of PV installed per year has been accelerating. Battery manufacturing capacity is already at weeks of global electricity demand per year. Utilities haven’t been building grid scale energy storage because they don’t need it, but it’s ready when they want it.
Over the next 20+ years a great deal of current infrastructure will need to be replaced simply because of age. What replaces it could be very green without significant issue.
> Saying we don’t have the infrastructure is meaningless when building infrastructure is part of my argument.
Again, you don't understand. I am not just saying that we don't have the infrastructure. I am saying that the size of the infrastructure we would need is a whole lot bigger than what you must imagine if you think that renewables can produce 99% of the world's energy.
You just vastly underestimate the problem. Saying "look, I went from selling 10 devices last year to 100 this year, so this proves that in 10 years I will be selling 1000000000000 devices per year" is the kind of reasoning you use in a startup when talking to a VC. But when you're being serious about solving a problem, it doesn't work like that.
Let me repeat it one last time: we will go away from fossil fuels, it's not a choice (they are limited in nature). We will need as much fission and renewables as we can get to compensate for as much as we can, but that won't remotely be enough (again, think about a real big merchant boat and tell me how it travels around the world without fossil fuels - not the startup way, but with a real solution).
So on top of fission and renewables, we need sobriety. A ton of it. And it means clever engineering across the board. So instead of wasting talents doing AI or polluting more with SpaceX, they should work on solving the actual problems we have for tomorrow.
3% of the worlds electricity per year isn’t several orders of magnitude from solving the problem. When I say that’s on pace to hit 100% carbon free grid before 2050 I’m not assuming crazy growth in anything.
We’re past the crazy exponentials. Global demand is still increasing every year by ~2.2% but that already includes the EV and Heat pump transition.
350GW last year, 356GW in 2024, 362GW in 2025 etc and before you know it we are done. Except 2024 is on pace to massively exceed that estimate, ~500GW looks more likely.
Again it's all startup talk. I never mentioned orders of magnitude, I mentioned complexity. You keep focusing on what already uses electricity, ignoring the fact that 80% of the energy we use is NOT electricity.
And you still haven't answered my question: how do you power a big merchant boat with electricity? Do you realize it doesn't work with batteries, or not? And do you realize that the merchant boats ARE globalization? We don't have a technical solution for that, not even as a proof of concept. And most certainly not with renewables.
99% of the energy used by mankind is sunlight, but obviously we aren’t aiming for accuracy here.
Your 80% as fossil fuels is half (coal, natural gas) which are mostly used to make electricity and therefore goes away on a renewable grid on its own. In essence you are double counting the inefficiency of fossil fuels as if it was somehow a positive. People do use some natural gas for heating and cooking, but there’s direct swap in replacements that use electricity.
“40%” is oil though again that’s what’s pumped out of the ground not what’s actually used as fuel. Subtract EV’s and year really talk about 10% “of the worlds energy” used in boats and aircraft.
> big merchant boat with electricity.
New boats can run 100% hydrogen out of the gate.
Container ships don’t actually last that long, but you can also retrofit existing engines to run 85% on hydrogen fairly easily.
I think the idea is more that the potential profit or the need for energy to prevent limiting this profitable venture will drive more capital into fusion projects. It's not clear that they will hit the man-month problem since it seems like there's dozens of fusion startups trying slightly different variants. Of course that doesn't mean it will solve the problem faster.
> It's not clear that they will hit the man-month problem since it seems like there's dozens of fusion startups trying slightly different variants.
I read: "it's not clear that parallelization will not help, because they are parallelizing", which doesn't really make sense. Ok, it's not clear that parallelization will not help (just because it's hard to prove). But we have to acknowledge that fusion energy is not a new thing, and it's currently unsolved. So let's not bet our future on the hope that it will be solved in the next 10 years in such a revolutionary way that it will beat all our expectations by orders of magnitudes, shall we?
The Man Month essay describes a breakdown in work throughput because of the exponential increase in communication channels and complexity of administration. Parallel startups do not communicate with each other. It appears from the outside that fusion does not have a known critical path to completion so increasing the number of bites at the apple seems like a logical way to scale attempting to solve it.
I agree we should not count on fusion (or wide spread carbon capture) to solve our problems and pretend we can continue as if there aren't any limits. Unfortunately unrealized miracle solutions are presented all the time to problems and since a lot of tech revolves around startup culture, our industry is prone to believing in them.
You might enjoy Sabine Hossenfelder's video exploring this "I recently learned that waste heat will boil the oceans in about 400 years": https://www.youtube.com/watch?v=9vRtA7STvH4
It turns out we can probably solve this by building planetary chimneys 5km tall that move heat to the outer atmosphere.
Since it seems like you've seen data pertaining to this, do you have any good/reputable sources? I've tried asking in various places about what percentage of planetary warming is due to direct heating from energy consumption, vs. greenhouse gas effects vs. natural causes, but usually just get accused of being a climate change denier and told to go educate myself. I'm really just curious about methodology, want to build a better mental model of how it works and how it's studied, and have never seen any discussions/papers talking about direct heating effects, so don't know where to start.
> what percentage of planetary warming is due to direct heating from energy consumption, vs. greenhouse gas effects vs. natural causes
Humans produce 20 TW of power [1]. (15 if we remove solar, wind and hydro.) The Sun delivers, to the Earth, 44,000 TW [2].
So raising the amount of the Sun's energy the earth retains by 454 parts in a million (329 if we remove solar, wind and hydro) adds to the Earth the energy of our entire civilisation. That is why emissions are the problem. Not our direct heat production.
It’s quite easy - we have numbers for humanity’s electricity and heat production. We also know how much atmosphere and oceans weigh, which we can multiply by specific heat of air and water. From this you can calculate how much we’ve heaten up the atmosphere/oceans - even ignoring the loss of heat to space/ground our impact is neglible.
