> Still. Moving now to North Europe because maybe 30 years from now you'll have to move? Why not wait it out and see if you have to move?
The older you are when you move, the harder it is. New language, new culture, new laws, disconnection from your old friends who took a different path. And less time to contribute to a state pension — a skilled 30-year-old can be a welcome addition to a workforce, a 60-year-old might get their state pension before citizenship — which may make them less inclined to accept you. Especially if you're moving because there's a huge global disaster and they need more workers rather than pensioners.
That said, the economics may radically change over the next 30 years; while I think the current humanoid robot workers are a bit gimmicky at best, and that they'll only be ready for autonomous use about 5-10 years after no-steering-wheel-included self-driving cars[0], the general trend of automation they represent is significant and began long before they were able to walk on legs.
> The whole thing is one huge speculation. There's no telling if Finland will be secure since its quite weak militarily. And we we don't really know what's gonna happen - we have models with very differing scenarios.
Indeed; very little is predictable even 10 years out in geopolitics. The UK was never going to leave the EU in 2012, but gone it was a decade later. The USSR was indomitable in 1984 and gone in 1994. 30 years? That's the gap between the height of the British Empire and the WW2 home guard starting with arm bands instead of uniforms, and being armed with a mix of privately owned guns and various improvised weapons.
For the environment, what happens depends on how people react to the models that exist. We may well completely eliminate CO2 emissions on that time scale — the technology is already known and in deployment for how to do this for electricity and land transport, it's being demonstrated for iron and other metal oxide refinement, but there are plenty of things to work on before we say we can manage this without also removing CO2 from the air.
But on the other hand, we may squander the remaining time, just like we squandered the last 30 years.
[0] More specifically, when the energy cost of the computing power needed for sufficient quality real-time vision reduces to ~10W; the computers Tesla currently use are supposedly 100W, but as Tesla's autopilot is not yet rated as a full-replacement for human drivers I must assume that a sufficient AI would also be more power-hungry. As 100W is very little power, Tesla (and others) might brute force this problem by having a bigger computer that uses e.g. 1kW rather than keeping the power use to 100W and waiting for a better computer. 1kW is my guess for the maximum that anyone would be willing to use in a car-based AI.
How dead Moore's Law is or isn't depends on what you're asking, I think it's still working for Joules-per-operation which is what matters here, and every factor of 10 improvement needed is about 5 years at Moore's Law rates.
> That said, the economics may radically change over the next 30 years;
I think you're optimistic, I think we'll see mass layoffs 5-10 years from now for white collar work, and 5-10 years after that for expensive manual labor (like doctors etc). I think the gap between "solving" white collar work (e.g something similar to AGI) and solving humanoid autonomous work will be narrow, but that's just my hunch.
Also, even without AI, sufficiently cheap telepresence robots can do to for manual labour what Mechanical Turk and remote work (and possibly GenAI) is doing for desk jobs — "good enough" can be surprisingly poor quality if it comes with a small enough invoice.
However, my main point is that things are hard to predict.
The older you are when you move, the harder it is. New language, new culture, new laws, disconnection from your old friends who took a different path. And less time to contribute to a state pension — a skilled 30-year-old can be a welcome addition to a workforce, a 60-year-old might get their state pension before citizenship — which may make them less inclined to accept you. Especially if you're moving because there's a huge global disaster and they need more workers rather than pensioners.
That said, the economics may radically change over the next 30 years; while I think the current humanoid robot workers are a bit gimmicky at best, and that they'll only be ready for autonomous use about 5-10 years after no-steering-wheel-included self-driving cars[0], the general trend of automation they represent is significant and began long before they were able to walk on legs.
> The whole thing is one huge speculation. There's no telling if Finland will be secure since its quite weak militarily. And we we don't really know what's gonna happen - we have models with very differing scenarios.
Indeed; very little is predictable even 10 years out in geopolitics. The UK was never going to leave the EU in 2012, but gone it was a decade later. The USSR was indomitable in 1984 and gone in 1994. 30 years? That's the gap between the height of the British Empire and the WW2 home guard starting with arm bands instead of uniforms, and being armed with a mix of privately owned guns and various improvised weapons.
For the environment, what happens depends on how people react to the models that exist. We may well completely eliminate CO2 emissions on that time scale — the technology is already known and in deployment for how to do this for electricity and land transport, it's being demonstrated for iron and other metal oxide refinement, but there are plenty of things to work on before we say we can manage this without also removing CO2 from the air.
But on the other hand, we may squander the remaining time, just like we squandered the last 30 years.
[0] More specifically, when the energy cost of the computing power needed for sufficient quality real-time vision reduces to ~10W; the computers Tesla currently use are supposedly 100W, but as Tesla's autopilot is not yet rated as a full-replacement for human drivers I must assume that a sufficient AI would also be more power-hungry. As 100W is very little power, Tesla (and others) might brute force this problem by having a bigger computer that uses e.g. 1kW rather than keeping the power use to 100W and waiting for a better computer. 1kW is my guess for the maximum that anyone would be willing to use in a car-based AI.
How dead Moore's Law is or isn't depends on what you're asking, I think it's still working for Joules-per-operation which is what matters here, and every factor of 10 improvement needed is about 5 years at Moore's Law rates.