Future Timeline

AGI has been taking a backseat in my mind of late but I am feeling a vibe like water disappearing before a tsunami.
I am on about 4-5 hours sleep but I want to chuck this post up anyways lest I forget.


Today I saw a rumor that OpenAI has built AGI and people saw it as credible enough to get hyped over it.


When I checked what Yuli Ban was up to publicly he strongly agrees with a guy who says AGI that will soon be ASI is less than 2 years away.
He retweeted this.
(Note within this Future Timeline post you seem to be able to click show more on the tweets to see the full tweet without a twitter account).





And then tweeted these himself







Ray Kurzweil himself someone who has stuck to his "2029" prediction though thick and thin is starting to say he thinks we may have AGI before 2029 despite there being no social desirability benefit to making the date earlier.


AGI fears and hype are starting to touch the mainstream something I believe we thought would not occur until AGI was at our doorstep?
Well now it's happening.

Prior to seeing the above I had also thought that AGI would be here in 2025 which was based off seeing mathematical predictions saying
Dr Alan D. Thompson's AGI % countdown would reach 100% in 2025.
https://lifearchitect.ai/agi/
I'm not particularly a smart person but I seem to have a knack for knowing whose predictions are most likely true.

Remember the various predictions and rumors a couple years ago about our current level AI now?
I think most people thought "that would be nice but according to "rational" or "realistic" thought I don't buy it."
Well it largely happened.
(Unless I am mistaken maybe I really am).
I have a kind of Déjà vu where I feel the same thing is happening now but with actual AGI, it doesn't feel "realistic" to say it is at our doorstep but maybe it actually is.
Only time will tell.

Real question is what OpenAI/Microsoft will actually do with said ASI agent even assuming it would be coming so soon? Do they keep it to themselves and have it incrementally design/update GP4's architecture until they feel like it's safe to unleash it on the public? Of course, if they decide not to, some other competitor eventually will unless they all do the exact same thing.

Suppose we'll know by sometime next year if we truly are that close to the event horizon. I can see tech oriented Megacorporations superseding the all too often slow and incompetent U.S. government, as they hardly understand the nature of AI, caught up in meaningless rambles about culture wars, and too busy appeasing just as non-informed bases to adequately respond to a singularity style event.

Yuli is right that when we often think of AGI, human society is already more enlightened like in Star Trek, but not always. Worst case to me is that things will fragment based on how different communities or nations respond to ASI. This itself doesn't have to necessarily be a terrible thing either, it truly depends on the nature of the entity in question, and how quickly we can adapt to a world with different rules than prior.

Also have to consider even if ASI comes this quickly, updating the infrastructure at a quickened pace won't scale at the same rate, unless as I said above, Megacorps take on the role traditionally allotted to the government, with ASI at the helm.

.

erowind wrote: ↑Wed Sep 27, 2023 9:15 pm

Cyber_Rebel wrote: ↑Tue Sep 26, 2023 8:22 pm Also have to consider even if ASI comes this quickly, updating the infrastructure at a quickened pace won't scale at the same rate, unless as I said above, Megacorps take on the role traditionally allotted to the government, with ASI at the helm.

Some info on this in pure economic terms based on what humans are capable of with their labor. China during the great leap forward directly reinvested ~50% of their GDP every year back into planned economic initiatives. Supreme Allied Command during WWII excluding the Soviets managed around 40% of GDP. The Soviets like the Chinese approached ~50% in that same time frame.

Without advanced cybernetics planned economies were capable of immense economic mobilization, with modern cybernetics we could certainly get that number above 50%. How far above I don't know. 60%? 80%? At some point you start reaching a hard limit as individual citizens have a need for some personal discrepancy in spending in all societies, but very efficient mobilization is possible. We already track most of our economy cybernetically, the same computation and complexity required to track the stock market, commodities markets, etc could be trivially redirected to a cybernetically planned economy.

This is the primary reason I'm not a complete doomer on the ecological crises despite how dire I sound a lot of the time. We don't have much time, it's true. But if our civilization was firing on all cylinders even without AI we could completely transform our economic systems, incentives, flow of goods, supply chains, etc in years, not decades, years. Humans are extremely adaptable and if we are threatened existentially we probably will at least attempt to do what is required of us.

If ASI does exist and is used for cybernetic planning GDP reinvestment in proportions higher than 90% are probably possible imo. The economy would transform rapidly, on the scale of 5 years or less.

I thihk the idea that the economy is too complex to be managed by people is severely mistaken.

The economic calculation problem is a flawed argument. Mises argued that the problem with socialist economies is that they would be incable of estimating the production costs, thus, they would not know how to invest wisely. The idea is: in a capitaist economy, you can for example compare the costs in monetary terms, for example, you have 2 roadbuilding methods, 1 costs 100 and the other costs 50, so you chose the later. According to him, in a socialist economy, you would not know the costs of the construction options, so you could not plan wisely. He dismissed the possibility of calculation in kind, because, according to him, you can't compare different types of units. For example, how can you compare 100 tons of steel with 50 tons o coal ? In market economies, you can compare things in monetary terms.

However, Mises never proved that it would be impossible tho devise methods to overcome these issues. He never offered a technical or quantifiable proof. He also never proved that the price system was rational. In the USSR, the solution used was a method called "material balances", it compared costs using natural units in a input-output table. For me, the fact that the USSR build a industrial economy in record time is proof that the method worked and economic calculation was never an issue. The deficiencies in the soviet economy were never caused by issues with economic planning, but by long term institutional deterioration, as the system became more corrupt and competent technicians started to be replaced by incompetent politicians, mainly in the Brezhnev period.

I believe the economy is actually very simple. It is far more simple than a single electronic circuit, for example. A common mistake is to believe the planners have to know everything, every gear in every machine down to the exact dimensions. But this is not true. The planners only have to know the interrelations between the sectors. Also, the central planners only take macroeconomic decisions. The microeconomic decisions are taken by lower ranking managers.

.