12 Comments

What thing could be extremely valuable to consumers, but cost-effective to supply mass-market? Experience value. If the human brain could be safely affected in such a way that a minute of subjective experience would be worth hundreds or thousands of current dollars to average consumers, but suppliable at average cost of $10 or less, that would certainly constitute a veritable revolution in aggregate economic surplus.

Expand full comment

Since the death of the last full-blooded neanderthal, we've been the smartest entity on the planet. That's going to change soon. So that will be the biggest change in ~40,000 years.

And shortly after that, those intelligent entities will probably be living on other planets and moons in this solar system. That will be the biggest change in ~4 billion years.

Expand full comment

Quite often a 2-3x productivity gain outweighs the gains from learning from a particular day's work. Especially for workers near the peak of their lifetime productivity.

Expand full comment

However, a spur loses all accumulated experience when it terminates, a single Em operating sequentially would be able to retain whatever it learned in the course of its task - it's not obvious that the spurs come out ahead necessarily.

And not to put to fine a point on it, the ethical concerns raised by spur terminations would not be in play - mightn't they come to be regarded as outweighing the benefits of a 3x (at most - if the emulation isn't at a ridiculously low level I doubt it's going to include natural causes of fatigue) speed gain?

Expand full comment

A world full of ems may be much better able to manage and control AI, and software descended from ems might win against software descended from AI in many areas. Spurs save on the cost of resting up to be ready for another workday.

Expand full comment

The main point of dispute between the Em-first scenario and the AI-first scenario appears to be that the Em-first scenario posits a two-year-long "age of Em" preceding the "age of AI", while the AI-first scenario posits that strong AI will be developed by regular humans without recourse to Ems. Given that the uncertainty in the timeframes involved is obviously much greater than two years, doesn't that render the two essentially equivalent from the perspective of long-term planning?

A side question - if Em processing speed can be arbitrarily dialed up or down limited only by available computing power, why are "spurs" needed at all? A single Em using the resources of a pair of spur Ems would run twice as fast, and so presumably would have the same productivity as the two spurs combined.

Expand full comment

It is useful in discussions such as this to recall that all change happens with limits - those of physical law. In some scenarios (e.g. time to communicate a short message) we reached the limit some time ago. In other scenarios we may still be far from the limit, but can determine what it is; and in all cases a limit (more generally, an envelope of what is technologically feasible) exists whether or not we can currently determine it. (I think that Eric Drexler emphasized this in his most recent book Radical Abundance; alas I misplaced it before reading it and it hasn't turned up yet:-)

Expand full comment

I agree that there are different possible ways to define parameters. I don't agree this means I'm "equivocating."

Expand full comment

Yeah, people tend to extrapolate recent prominent trends when they picture the future. The word "prominent" is important, because at any given time, most real trends are under the radar of most people. Usually, the prominent trends get our attention because they're already changing society. So in the early 70's everybody was expecting planes to go faster and space projects to get bigger. Why? Because we made huge, recent leaps forward with aerospace propulsion tech. In 1972 our grandparents could cross oceans at supersonic speeds, wondering how soon we'd go even faster. The fantasy about jetpacks and flying cars comes from this era - a time when propulsion was the tech that was making the biggest impression on us, thus making it the most natural object of easy extrapolation.

Today's lazy extrapolators can only think about Moore's Law, AI, pixel density, bandwidth, etc. If I told my grandfather in 1972 that 50 years later the Concorde will long be scrapped and now we only go a tad over 600 without being too bothered (and yah, we still use miles, grandpa)... that in the intervening 50 years we haven't sent another person beyond low earth orbit, without being too bothered... I don't think he'd understand why not.

So I'm trying to picture a grandchild visiting me from 50 years in the future, telling me that yeah, processors kinda got stuck at 4 GHz and 12 cores, and that some places still happily run 50 year old PCs, which are considered inefficient but adequate... that 100 megabit internet is still basically seen as plenty fast enough, that hard drives can go to 10TB but most people don't fill them up, that using a 30 year old iPhone is no weirder for them than wearing a 30 year old Rolex is for us, and that all this happened without people being generally too bothered, because their attention got transfixed by something different, something that mattered more than mere computing... well...

I wouldn't be able to imagine what such a thing would be. And that's a lot like the situation of my grandfather in 1972.

Expand full comment

This seems to equivocate on parameterization and what counts as 'change'. If you consider your example of slavery: because of population growth, there may well be a larger number of slaves, (debt, penal, legal, and otherwise) now than ever before, in which case slavery can both increase and decrease more than it has ever before; as a percentage of population, perhaps it's low and obviously one can't drop slavery by another 50% of global population, but one could just as easily use a ratio or log parameterization instead, in which case, like absolute counts, it is perfectly possible to have changes as large as before. Or consider smallpox: was the recovery of the last smallpox victim and the resulting extinction of smallpox in the wild the *largest* or the *smallest* change in smallpox ever?

This becomes much more so true if you consider ems or genetic engineering, which allow both quantity and quality changes of orders of magnitude.

Even humdrum things like adult life expectancy can still change a lot: adult life expectancy in the West has only increased something like 20 years over the past 2 centuries, so another change of equal magnitude would only push us up to ~95 age - doesn't seem impossible. For comparison, in 1817 (200 years ago) in medicine, Parkinson's disease was described for the first time, the germ theory of disease was still controversial, Snow had not stopped a cholera epidemic, hygiene was unknown, vaccination did not go beyond cowpox (just starting to be popularized after its introduction 20 years before), the stethoscope had been introduced just the previous year, etc.

Expand full comment

For a specific concrete objection, if the em era happens in 100 years' time then the degree of change inside it alone will surely be as large as from the start of the industrial era to the end. While this probably won't represent a reversal on all the particular parameters that changed the most during the industrial era, there'll be an additional (by your calculation) 98 years in which God-knows-what further eras will happen, plausibly running as fast as the em era if not faster. So it seems hard to argue that all this can happen, and nonetheless the world won't change as much in these next two centuries as it did between the start of the 19th century and today.

I also strongly suspect most people who voted on your poll expect ems or some other kind of AI to happen within the next 200 years. You hint at overall rates of change remaining constant in your post, but I think that would need to be an explicit, clearly stated assumption for people to actually consider it when answering the question.

Assuming we stay in the industrial paradigm for the next 200 years, your claim here seems right.

Expand full comment

"future change must be less than past change." My response to this point is that we will very likely create entirely new things that we value. Concepts that may seem alien from the perspective of humanism.

Expand full comment