Building an AGI/SGI (Artificial General Intelligence / Super General Intelligence) The path forward – Post 002 in a series

In general, I see the progress towards an AGI in two phases or parts; the first part would be part of a general overall total automation of the entire workforce. The second part would be the automation of the directing of that workforce. I also see that once the full automation project reaches a certain point that it would become in effect an AGI, which could then build a SGI.

From looking some of the writings and research on the subject, I think there are many problems with the approaches being taken in AGI:

1: Too much focus on pure research and not enough on applied. MY recommended approach would be to follow applied through the automation of the workforce and eventually an AGI would result. The pure research seems to be going no where or making any progress. I am reminded of a story from the book Hackers, where some teen working in one of the labs, maybe MIT, was able to coble together enough information and parts from other researchers to make a robot that could coral an object thrown in front of it into a goal area. As I recall the story, this was the 1960s or 1970s and the other people in the lab ‘went nuts’ because it was something they had not been able to do but this ‘kid’ beat them to it. While the kid may have been much better at it than them, I think part of the reason for his success was that he was focused on an end object and making something that did something; where as the other researches were mostly focused on pure research.

2: Computing power – We have more than enough computing power to do something useful towards full automation of the workforce. There are now available single board computers that cost as little as $9. I would put it to people that the problem is not of computing power, but what to do with it or how to program it. Many of the current processes that are being done for work add little to no value to producing goods and services. A fully automated workforce would be an opportunity to eliminate those items.

3: Question on what happens to the workers – This is a question that needs to be considered now, while it may be a decade or two before there is a fully automated workforce, a discussion needs to be had as to what to do with the large numbers of people that will not need to work. While I expect there would still be a need for the creativity from people, at least until the SGI come online, there is a question as to how to structure to run things. My present idea is that if we can get things to where people and governments are not in debt, and running things on a current budget, that either each person would have its own collection of machines that would work for them or there would be some kind of taxing system on a weekly basis where part of the profits from the automated workforce would be paid into a system and then everyone gets an equal share of that on a weekly or month basis, and the amounts would be such that a person would be able to maintain a reasonable lifestyle like our current middle class. I expect that under such a system that finance and banks would still exist but not as large as they are now, since if people could get enough to have a reasonable standard of living then there would be little to no need to borrow or finance current spending from future income. Many other things would need to be changed, like the constantly inflating of assets would need to be stopped by central banks since that would create problems with the value of the weekly incomes being handed out.

4: Programming methods that do not make sense – I have noted that most of the Machine Learning or AI are all centered around ‘big data’; we take a huge set of data and setup some kind of learning conditions for a machine to come up with the best spaghetti web of code to do the task. While it is clear this is better than some of the traditional programming, I do not think it is the best long term. As an example, self driving cars. We are collecting all kinds of images and photos and the car ‘learns’ from what it is told and feedback what it all means; but if you think about how you are thinking when you drive; are you searching through millions of patterns to figure out what to do? I would put it to you that no one thinks that way, and people think a lot slower than a computer, so we are going about it the wrong way. To me it seems like since we could not figure out how to program vision or driving in a car, that we instead are collecting this huge set of data and running a million IF/THEN statements against it to get it to work. While better than nothing, it is not a very good way to program. We need to figure out how things really work and then model the systems on that, not going off and making millions of IF/THEN statements.

5: Better uses for our computing power – In the books AI Superpowers: China, Silicon Valley, and the New World Order and Life After Google: The Fall of Big Data and the Rise of the Blockchain Economy the only things this big data seems to be being used for is to sell advertising and more and more things to people who already have houses full of stuff. Where is all of the world changing innovations that were supposed to be coming from all this computing power, cures for diseases, productivity improvements, automating the workforce? Instead, most programming seems to be focused on remaking software that already exists or selling advertising to sell more products to people who already have houses and businesses that are overloaded with stuff. Google, for all its computing power and wealth, is really nothing more than a seller of advertising space! What good is that in the long run, especially when everyone has a house full of stuff they mostly do not use?

These are some of the considerations and plans for fully automating the workforce.

Good Luck and Take Care

Louis J. Desy Jr., Wednesday, October 07, 2020

LouisDesyjr@gmail.com