How about we stop trying the analogy clothing on and just tell it like it is? AI is unlike any other technology to date. Just like predicting the weather, we don't know what it will be like in 20 months. Everything is a guesstimate.

This is the correct take. We all have that "Come to Jesus" moment eventually, where something blows our minds so profoundly that we believe anything is possible in the immediate future. I respect that, it's a great take to have and promotes a lot of discussion, but now more than ever we need concretes and definitives instead of hype machines and their adjacent counterparts.

Too much is on the line here regardless of what ultimately ends up being true or just hype.

Probably the point is to think whether the horse or chess engine analogy is a good one. The premise being there will come a certain point when technology reaches a level that makes the alternative obselete suddenly. I don't have good reasons to think that AI will not be able to automate simple jobs with an acceptable error rate eventually, once that happens whole categories of jobs will evaporate. Probably dealing with more people type job, making excel models, transactions based, same thing day in day out, those teams may be gone and only a person or two to do a final review

Against that you have the Moore's law like predictions that AI would be getting to around human levels around now from Moravec and the like that have proved fairly spot on. I think you may find it's more like the AI chess ranking graph than the weather.

I think that you're on to something here, though I agree more with your first sentence than the second.

AI is not identical to, as the article compares, mechanical power.

But your weather-forecasting comment suggests a possible similarity (though not the one you go to): for all the millions-fold increase in compute power, and the increased density and specificity of meterological measurements, our accurate weather-forecasting window has only extended by a factor or so of two (roughly five days to ten). That is, there are applications for which vastly more information-processing capacity provides fairly modest returns.

And there are also those in which it's transformative. I'd put reusable rockets in that category, where we can now put sufficiently-reliable compute (and a whole bunch of rocket-related hardware) on a boost-phase rocket such that it can successfully soft-land.

For some years I've been thinking of the notion of technology not as some general principle ("efficiency" is the classic economics formulation), but as a set of specific mechanisms each of which has specific capabilities and limitations.[1] I've held pretty constant with nine of these:

1. Fuels. Applying more (or more useful) energy to a process.

2. Energy transmission and transformation.

3. Materials. Specific properties, abundance, costs, effects, limitations.

4. Process knowledge --- how to do things. What's generally described as "technical knowledge", here considered as a specific mechanism of technology.

5. Structural or causal knowledge --- why things work. What's generally described as "scientific knowledge".

6. Networks. Interactions between nodes via links, physical or virtual, over which matter, energy, information, or some mix flow. Transport, comms, power, information.

7. Systems. Constructs including sensing, processing, action, and feedback. Ranging from conceptual to mechanical to human and social.

8. Information. Sensing, perceiving, processing, storing, retrieving, and transmitting. Ranging from our natural senses to augmented ones, from symbolic systems (language, maths) to algorithms.

9. Hygiene. Sinks and unintended consequences, affecting the function and vitality of systems, and their mitigations or limits.

AI / AGI falls into the 8th category: information, specifically information processing. And as such, getting back to my original point, we can compare it with other information-related technological innovations: speech, writing, maths, boolean logic, switches (valves, transistors, etc.), information storage/retrieval, etc. And, yes, human thought processes. We do have some priors we can look at here, and they might help guide us in what a true AGI might be able to accomplish, and what its limitations may be.

It's often noted (including in this thread) that AGI would not presently be able to persist without copious human assistance, in that it's predicated on a vast technological infrastructure only a small portion of which it would be capable of substituting for. It's quite likely that AGI would be both competitive with and complementary to much human activity. In the horse analogy, it's worth noting that the first stage of mechanised transport development, with steam shipping and rail technology, horses were strongly complementary in that they fulfilled the last-mile delivery role which steamships and locomotives couldn't furnish. Horse drayage populations actually boomed during this period. It was development of ICE-powered lorries which finally out-competed the horse-drawn cart for intra-urban delivery. AGI-as-augmenting-humans is an already highly-utilised model, and will likely persist for some time. Experiments in AGI replacing humans will no doubt occur, some successful, others not. I'd suggest that my 9th category, hygiene, and specifically failure modes of AGI, will likely prove highly interesting.

Mechanised transport also relies heavily on fuels and/or energy storage. The past 200 or so years were predicated on nonrenewable fossil fuels, first coal then oil, and there were several points in that timeline where continued availability of cheap fuels was seriously in question. We're now reaching the point where even given abundant supply, the relatively-clean byproducts of use are proving, at scales of current use, incompatible with climatic stability, possibly extending to incompatible with advanced technological civilisation or even advanced life on Earth (again, category 9).

AGI relies on IC chip manufacture (the province of vanishingly few companies), on copious amounts of electricity, scarce physical resources, and various legal regimes concerning use of intellectual works, property, profit, and more (categories 1, 2, 3, and 7, at a minimum). Whether or not a world with pervasive AGI proves to be a stable or unstable point is another open question.

________________________________

Notes:

1. A sampling of prior HN discussions may be found with this search: <https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...>.