Post-Scarcity Economics

[MSimon]

I think a lot of the discussion reflects the Luddite fallacy:

If weaving is automated what will we do for work?

Become a computer programmer perhaps? Design a weaving machine that can recreate a pattern from a photograph? Too far away to see.

======

Take the water example: Why do Chicagoans buy water instead of dipping buckets in the lake?

1. Time is money
2. Quality issues
3. Quantity requirements
4. Place requirements: you want indoor plumbing? It is going to cost ya.

Anyway. Humans will find a way to profit. What it will be is not entirely clear at this point.

[Luzr]

Personally, I think interfaces capable of transfering data between electronics and brains are much more achievable than an artificial consciousness.

Definitely. Working implementation is keyboard, mouse and LCD display.

[djolds1]

I think a lot of the discussion reflects the Luddite fallacy:

If weaving is automated what will we do for work?

Possibly. One webfriend of mine is convinced that post-scarcity economics are utopianism dating back to the zero-labor dreams of Joachim of Fiore's Age of the Holy Spirit. It is a question of foresight I've been trying to answer for the last week or so - what is a probable line of adaptation to these potentials?

Take the water example: Why do Chicagoans buy water instead of dipping buckets in the lake?

1. Time is money
2. Quality issues
3. Quantity requirements
4. Place requirements: you want indoor plumbing? It is going to cost ya.

All of which (except perhaps time?) are addressed at minimal to zilch cost by moderately mature fabbing.

Perhaps charge for the raw material feedstock and energy to drive the fabbers (assuming each one doesn't have its own polywell)? But still, the economic incentives of that system are more like the Soviet Union than what we regard as "optimal" economies.

Humans will find a way to profit. What it will be is not entirely clear at this point.

Yup. That is the crux of the matter. What is the probable vector along which adaptation will occur?

[alexjrgreen]

Yes, I guess there is sort of consensus that the real AI must "self-emerge" or "grow".

The "prediction of future states" that Hawkins discusses is a property of memristance: Memristor minds: The future of artificial intelligence.

It may be that this ability to predict is what a network scales up to achieve higher levels of intelligence.

MRAM has a better shot and is currently in production.

Igor Aleksander was using randomly connected memory chips to recognize Margaret Thatcher in the 1980s.

That's given us football crowd face recognition, but not intelligence.

[alexjrgreen]

The Hero engine was a velocity engine (analogous to a turbine) while the engine that actually got things going was a torque engine. To make turbines work they need to operate at constant high velocity. Which says that you need gearing to do useful work.

The Greeks had gears, too.

[MirariNefas]

Along that line of thought, what comes first, the mind machine interface, or the strong AI?

Depends on how you mean. "Psychic" controllers for game stations came out this year. Interfaces of that type will explode over the coming few years. If you mean "uploading," AI is MUCH easier. Sapient AI is easier than uploading because we don't need to know HOW a brain matures. Put the silicon or protoplasm for the brain in place, pump in stimuli, and "decant" an adult in 4-25 years. How the 100 trillion neural or quasi-neural connections "line up" is unimportant, so long as the "product" (typically called a child) works. For uploading, you need to understand each and every connection in that 100 trillion strong network. Which connections have reinforced, which atrophied.

Not meaning uploading right now. I just mean the capability of a human brain to shunt some calculations over to a chip, and to access data stored in electronic hardware or on a network. The idea is just brain plasticity - we all have a little of it, children have more, some of it may be genetically/chemically inducible. So you may be able to tell a brain to start decoding new signals that arrive through a special sort of chip/cell package, and hopefully you could build software to figure out what to do with signals that come back.

Personally, I think interfaces capable of transfering data between electronics and brains are much more achievable than an artificial consciousness. If so, we'll have the capacity to improve ourselves well before we could create our replacements. Moreover, this will remove the incentive to build our replacements, because you could just hire a human to interface up and do the same job.

Possibly. AI would have the advantage of much faster processing speed; the neurons "click" orders of magnitude faster.

An issue, yes, but if the brain interfaces with a bunch of these hyperfast chips, perhaps we could train it to send difficult calculations over that way. If it happens automatically that's best, but if it happens like a calculator that works too. You come across some advanced physics problems with too much data to take in quickly with your eyes, so you consciously download the data through the network and activate a program to sort it. Then your chip bottomlines it with a program a lot like our cortex, and sends the answer down to the gray matter. The gray matter does some thinking about implications, sends some more instructions back to the chip, calculations ensue.

Not as fast as a full AI, but it may happen sooner, and may be sufficient for all human purposes.

[TallDave]

Some more sophisticated software in specialized apps rewrites its own code in response to various situations

And just how do you verify such a system?

As I understand, in some of these systems you aren't verifying the code, you're verifying the results they give. Expert and predictive systems are sometimes intended to produce possible solutions, sometimes where it's hard to do things in real space.

This is one reason why I don't trust GCMs; the results aren't the sort of thing you can assign a lot of certainty to.

[TallDave]

An issue, yes, but if the brain interfaces with a bunch of these hyperfast chips, perhaps we could train it to send difficult calculations over that way. If it happens automatically that's best, but if it happens like a calculator that works too. You come across some advanced physics problems with too much data to take in quickly with your eyes, so you consciously download the data through the network and activate a program to sort it. Then your chip bottomlines it with a program a lot like our cortex, and sends the answer down to the gray matter. The gray matter does some thinking about implications, sends some more instructions back to the chip, calculations ensue

I think one of the first, most useful things you could do is increase the short-term memory register. Not being able to think about more than a dozen or so things at once is very limiting.

[TallDave]

Anyway. Humans will find a way to profit. What it will be is not entirely clear at this point.

No doubt many surprises are in store. It would certainly be hard to predict the economy of 2009 in 1959.

[djolds1]

Not meaning uploading right now. I just mean the capability of a human brain to shunt some calculations over to a chip, and to access data stored in electronic hardware or on a network. The idea is just brain plasticity - we all have a little of it, children have more, some of it may be genetically/chemically inducible. So you may be able to tell a brain to start decoding new signals that arrive through a special sort of chip/cell package, and hopefully you could build software to figure out what to do with signals that come back.

That's one possibility. IIRC some of the blind have been able to "see" via electrode arrays taped to their tongues. Linking that to the new "psychic" controllers opens a lot of possibilities.

And just how do you verify such a system?

As I understand, in some of these systems you aren't verifying the code, you're verifying the results they give. Expert and predictive systems are sometimes intended to produce possible solutions, sometimes where it's hard to do things in real space.

Genetic algorithm systems? No way in Hades we can verify the logic trains that lead to the output code, and the output code is often very "ugly" in human terms. Not parsimonious.

I think one of the first, most useful things you could do is increase the short-term memory register. Not being able to think about more than a dozen or so things at once is very limiting.

Already being pursued via drugs ("nootropics"), and I'd bet good money there are some genetic therapy methods in the works like the myostatin blocking method for muscle growth.

[TallDave]

I've a tried some nootropics and use a few regularly. The benefits are hard to quantify, but I have found dimethylaminoethanol to be helpful, and melatonin has some enjoyable effects. Omega-3s are a no-brainer.

[IntLibber]

The "prediction of future states" that Hawkins discusses is a property of memristance: Memristor minds: The future of artificial intelligence.

It may be that this ability to predict is what a network scales up to achieve higher levels of intelligence.

MRAM has a better shot and is currently in production.

Igor Aleksander was using randomly connected memory chips to recognize Margaret Thatcher in the 1980s.

That's given us football crowd face recognition, but not intelligence.

As researchers have known for some time, human intelligence is not one unitary skill or circuit. It is a host of intelligences, capabilities, that feed each other data: face recognition, speech recognition, musical recognition, creating music and speech and images, mathematical processing, smell recognition, touch recognition, and a very complex distributed neural network method of processing all that data, distilling it, storing it, and associating it all with many relationships (smells to feels to tastes to sights and sounds, etc).

To date I don't know of any AI effort that has sought to combine all these capabilities in one machine and program them all to work together, with the same database, building relationships and hardening them by repetition, and enabling a chat bot to use this database as source of information, with the ability to create output on its own initiative.

[Betruger]

Possibly. One webfriend of mine is convinced that post-scarcity economics are utopianism dating back to the zero-labor dreams of Joachim of Fiore's Age of the Holy Spirit. It is a question of foresight I've been trying to answer for the last week or so - what is a probable line of adaptation to these potentials?

Actually reaching for usage of full capacity of the human brain instead of living in mediocre, stagnant state of bondage with the rest of society; where only a tiny portion is actually spent moderately free (roughly sandwiched between early ignorance and late impotence)?
At a truly post-scarcity stage, I still think originality of design (whether products, lifestyles, or whatever) will be a major appeal to put a price on. What needs to happen first is for people to settle down to the idea that we can live much longer lives. There'll never be a shortage of work in discovering new things. I see it the opposite way: we can't get to infinite knowledge and power soon enough. And like De Grey says, everyday we delay it is a massive number of people who eventually die because they will miss that boat.

[Betruger]

An issue, yes, but if the brain interfaces with a bunch of these hyperfast chips, perhaps we could train it to send difficult calculations over that way. If it happens automatically that's best, but if it happens like a calculator that works too. You come across some advanced physics problems with too much data to take in quickly with your eyes, so you consciously download the data through the network and activate a program to sort it. Then your chip bottomlines it with a program a lot like our cortex, and sends the answer down to the gray matter. The gray matter does some thinking about implications, sends some more instructions back to the chip, calculations ensue.

Not as fast as a full AI, but it may happen sooner, and may be sufficient for all human purposes.

Sounds a bit like something I read off some Google tech report.. "Why try to make an elegant, exactly accurate model if you can model something that walks and talks just like the real thing, even if it's very complex?"

[alexjrgreen]

Igor Aleksander was using randomly connected memory chips to recognize Margaret Thatcher in the 1980s.

That's given us football crowd face recognition, but not intelligence.

As researchers have known for some time, human intelligence is not one unitary skill or circuit. It is a host of intelligences, capabilities, that feed each other data: face recognition, speech recognition, musical recognition, creating music and speech and images, mathematical processing, smell recognition, touch recognition, and a very complex distributed neural network method of processing all that data, distilling it, storing it, and associating it all with many relationships (smells to feels to tastes to sights and sounds, etc).

To date I don't know of any AI effort that has sought to combine all these capabilities in one machine and program them all to work together, with the same database, building relationships and hardening them by repetition, and enabling a chat bot to use this database as source of information, with the ability to create output on its own initiative.

You're right, of course, but you're also missing the point.

There's a fundamental aspect of intelligence, that even single-celled organisms display, that scales up with numbers and networking. It may be a result of memristance or even perhaps borderline quantum effects, but it's missing from the research we've done so far.

The ability to scale up intelligence allows all the different techniques to be integrated.