AI vs. Humans as Processors in an Environment

[jack action]

@BillTre , you keep putting skeptical reactions to my comments. Regarding the point I want to make:

All I want to make clear is that AI systems don't have needs to be fulfilled. They don't work for their own survival.

Do you think AI systems have needs to be fulfilled and that they work for their own survival?

It's already doing much of what you hope it will never do.

They do what humans want them to do. If they do something humans don't want them to do, it is considered a bug, and they are shut off. To see otherwise shows a remarkable abundance of imagination.

 

[BillTre]

@BillTre , you keep putting skeptical reactions to my comments. Regarding the point I want to make:

You are equating not having competition with a lot of things that are irrelevant to whether competition exists or not. This weird redefining makes me skeptical of your arguments.

You also claim that AIs don't have needs for some reason while they obviously require certain things to function. This does not inspire confidence.

 

[Swamp Thing]

I learned from past experience, something new on the horizon small companies jump in to sell add-ons or consulting.

That reminds me of the sad tale of someone who built a career and a small enterprise, based on his deep knowledge of "Terminate and Stay Resident" ( TSR) programs for MS-DOS. You can imagine what happened when Windows 3.1 was released.

 

[jedishrfu]

Yes MS destroyed many promising companies in their rise to dominance.

Zip tools when MS introduced a compressed filesystem. I don't think users responded well to this new os since disk failures become unrecoverable for text files that looked encrypted.

OS/2 when MS repackaged the code to sell Windows NT. IBM made sure MS didn't get access to the networking part of OS/2 and also sold separately as a product. When Windows NT would have a system fault the error would say OS/2 error.

Lattice C initially resold by MS until they expanded their compiler portfolio with multiple languages including C. Lattice faded away a few years later.

 

[jack action]

OK, let's go over the OP:

Humans and AI systems have the possibility to be competing systems in an environment of limited resources. Where the use of both people (as communities) and large energy and water consuming sites in the same areas (areas supplied by some economically available source), competition will ensue.

I don't get where the competition is. Do you really think that if there is a server farm somewhere that uses water for cooling (or any other industry using water for that matter, for any manufacturing reason), and that the local population has trouble finding water for themselves, that humans - including the officials and even the owners of the business - will end up rooting for the machines and accept that people have trouble finding water? Maybe with a small population under the control of a larger outside group (that would be humans competing with other humans), but never as a whole, within a society. People will burn the plant before that happens.

The only competing needs are both for humans: the need for water and the need for AI systems. If their AI systems require water, then it is not a competition; at best, it is resource management.

And this is where it is important to see that there are humans behind the AI systems. Not only do they need water, but they also run the AI systems. How could anyone possibly think that humans would deprive themselves of water to run their machines? That makes no sense.

Here, I am thinking of humans as a form of computational mechanism. They are in competition with individual AIs as computational processes.

AIs may have more of some computational ability, than the average human.

You seem to reduce humans to simple, inefficient machines. Is computing really the end goal of human survival? Are you saying that humans would somehow sacrifice themselves to ensure AI systems survive because computing is so important?

It is not clear to me how one might determine if some change would be considered as good or bad.

People not having access to a water source is definitely bad. Not having access to enough energy to heat their homes or cook their food is also bad.

 

[jedishrfu]

!!! SECOND PUBLIC SERVICE ANNOUNCEMENT !!!

We aren't here to debate issues of competition for resources between humans and AI. It's real, but it is more about the conflict between consumers vs. big business, a dynamic exacerbated by the rise of huge data centers for AI products.

Let's have a lighter discussion and identify the pain points brought on by the advent of AI, data centers, and the mega corporations who build them.

---

From my experience living in Texas:

Data centers require vast amounts of electricity and water to run their server farms. They may generate increased heat or pollution depending on the type of resource. In Texas and elsewhere, data center electrical bills are sure to raise homeowners' electric bills.

Why? Because of the tiered structure of the electrical grid, with free but intermittent resources of wind and solar, then the almost free nuclear and hydro, then not-free resources of natural gas, older gas plants, peaker turbines, and emergency generation.

To mitigate day-to-day rate increases, electric grid planners add capacity, which means building more plants for the future. Consumers pay in either case.

Homeowners concerned about their electric bills will make the hard decision about whether to add solar panels to their homes and get a payback of a few days per month.

In Texas, it was particularly bad because peak usage led to rolling blackouts, or businesses agreed to voluntarily reduce electricity consumption.

The same goes for water: in Texas, we pay for it, and the more we use, the more we pay. Baked into those rates are the total demand and demand for sewer services. Sometimes, water districts sell water to other districts because of need.

The effect on consumers is tiered pricing and the water-monitoring notion of measuring usage in the winter months, assuming that usage primarily goes into the sewer, whereas irrigation and other activities go into the ground.

---

Bottom line, consumers pay higher water and electric bills, and they pay for AI subscriptions.

 

[jack action]

In Texas and elsewhere, data center electrical bills are sure to raise homeowners' electric bills.

The AI users should pay for the resources they use, but are you saying that you expect all consumers to pay more in the future for electricity, even if they don't use AI (directly or indirectly)?

The same goes for water: in Texas, we pay for it, and the more we use, the more we pay.

But what happens if there is a water shortage? Will data centers be seen as just another customer, or will drinking water be prioritized?

 

[BillTre]

You seem to reduce humans to simple, inefficient machines. Is computing really the end goal of human survival?

Are you having trouble with abstracting an aspect of human behavior, or are you just trying to force a point?

People not having access to a water source is definitely bad. Not having access to enough energy to heat their homes or cook their food is also bad.

The question is the OP was about a more nuanced situation concerning the balance between the effects of different factors that could have good or bad effects. How to decide the goodness or badness is the issue I raised. Sorry, its not simple Are you only focusing on good vs. bad?

 

[jack action]

The question is the OP was about a more nuanced situation concerning the balance between the effects of different factors that could have good or bad effects.

Could you define those factors and effects you are thinking of instead of letting us guess?

How to decide the goodness or badness is the issue I raised.

Are you only focusing on good vs. bad?

These seem to be saying pretty much the same thing. Define what you consider goodness and badness.

Going back to the OP, in the following statements, define what you are referring to with problems and change:

One effect is that some communities have been afflicted with water and cost of power problems.

It is not clear to me how one might determine if some change would be considered as good or bad.

In the OP you begin with:

Humans and AI systems have the possibility to be competing systems in an environment of limited resources.

and end with:

Economic progress always has an impact on individual's economics (like the buggy whip manufacturers).

Discussing about limited resources and economics are not the same things. The former implies that priorities will have to be made (if drinking water is rare, you don't wash your clothes with it) and the latter implies that every choice is possible, they just don't have the same value (there is nothing wrong with manufacturing buggy whips today, the demand is just not what it used to be).

 

[PeterDonis]

Discussing about limited resources and economics are not the same things.

Actually, yes, they are. Economics is about limited resources and how we prioritize how they get used: a free market and the price mechanism it embodies is a way of doing that. The reason you stop using water for other things than drinking if it's rare, in a free market, is that its price goes up, sending a signal to economize.

That's not to say that a free market and price mechanism is the only way to allocate scarce resources. It's not. But it's the only way that scales, in general. That's why economists focus on it.

It's true that there is a temptation to add constraints to the free market: to say, for example, that water for drinking is more important than for washing clothes or other uses, so we should add a rule to the free market that forces people to respect that priority. But actually, if there is a free market for water, that's not necessary: the price signals will do it for you, because everyone already knows they need water to drink more than they need it for other things, so they will economize on their own as the price of water goes up (and the raised price will also signal to suppliers that there is more demand for water than there is supply, so they can make money by finding ways to supply more).

If that mechanism is not working, that's not a sign that free markets don't work; it's a sign to look for how the market is not free, that someone has their thumb on the scale. (The story of water rights in the US, particularly the western US, is a long drawn out case study in what happens when there are various entities at various levels involved who are all trying to put their thumb on the scale.)

 

[sbrothy]

The only way the rich get richer is by exploiting people working for them for nothing. People dying or getting sick means the rich lose workers who built their wealth. They will never let that happen, just like a farmer will prevent their flock from dying or getting sick.

Pretty cynical, if not downright nihillistic!

I feel fairly sure that you can't seperate mind and machine though. Like you can't upload a human brain (at least not in the foreseeable future). Blade Runner (main protagonist "Deckard") = "Descartes", dualism mind-body problem etc. In the end we control the energy switch. That is, unless AI get's so smart it can talk us out of pressing that switch or shutting a power plant down.

 

[sbrothy]

Your logic is unconvincing to me.

However, if the trees were planted by humans for agricultural purposes it wouldn't be competition?

It sounds to me like there are cases of competition you refuse to acknowledge for reasons unrelated to the actual competition for resources that is occurring.

In biology mitochondria could be thought of as competing for the same resources as the larger cell in which they reside. Evolution has taken care of that by putting control mechanisms in place to control these possibly lethal conflicts. This was done by selection for those rare control mechanisms which arose and allowed better survival. This is similar to the situation in Texas where:

Trees don't have agency. They work on fundamental physical principals like pressure sucking water out of the ground and the like. It's not a real competition. EDIT: IMHO.

 

[PeterDonis]

Trees don't have agency. They work on fundamental physical principals like pressure sucking water out of the ground and the like.

I'm not sure there's an actual binary distinction here. Human brains also work on fundamental physical principles. They're not magic. Whatever agency trees can be said to have is much more limited than the agency humans have; but that doesn't necessarily mean it's literally zero.

 

[sbrothy]

I'm not sure there's an actual binary distinction here. Human brains also work on fundamental physical principles. They're not magic. Whatever agency trees can be said to have is much more limited than the agency humans have; but that doesn't necessarily mean it's literally zero.

So you're of the opinion that trees make decisions about for instance where it's smartest to send their roots? Do trees have brains? I think the distinction is more profound that that. In my opnion trees have literally zero agency. I mean where should it come from? Comparing trees to humans seems downright ridiculous to me, but then again I'm neither an anthropologist or a dendrologist(?).

EDIT: It's almost like saying grass thinks.

 

[jedishrfu]

The AI users should pay for the resources they use, but are you saying that you expect all consumers to pay more in the future for electricity, even if they don't use AI (directly or indirectly)?

But what happens if there is a water shortage? Will data centers be seen as just another customer, or will drinking water be prioritized?

Yes, Jack you need to look at the big picture here. New tech bring benefits, and we pay to enjoy those benefits in subscription fees but we also pay indirectly when resources are constrained and new infrastructure is needed then we all foot the bill too.

I don't know how it will play out in Texas. I imagine there's a clear priority list with grid protection, medical facilities, first responders then consumers and businesses in the middle and special load circuits at the bottom. This list could change based on need but it's mainly used to keep the grid from collapsing.

Water constraints are likely very different as you want to preserve the flow but at the same time use less. Flint Michigan made that abundantly clear when they changed the source of the water to save money and created an epic disaster because of the corrosiveness of the local river water and then didn't correct for it until it was too late.

I know it's hard to see but we live in an economic network paying for services directly as a subscription and indirectly in our utility bills. Sometimes, we pay for things we’ll never use.

 

[PeterDonis]

So you're of the opinion that trees make decisions about for instance where it's smartest to send their roots?

Since they don't send their roots everywhere, they clearly are doing some kind of "deciding" about where to send them and where not. So yes. Of course these "decisions" are very, very simple, much simpler than decisions that humans make. But that doesn't mean the amount of decision is literally zero. It means that "decision" is not a binary concept. Even a thermostat can be said to "decide" when to turn your heat or A/C on and off based on the temperature it detects and its settings. Trying to draw a bright line here is neither necessary nor practical.

Do trees have brains?

No. But brains are not required for agency. All that's required is some ability to affect the outside world, and some kind of filtering that means only some of all the possible effects actually get realized. Trees have that. Thermostats have that.

where should it come from?

See above.

Comparing trees to humans seems downright ridiculous to me, but then again I'm neither an anthropologist or a dendrologist(?).

The concept of "agency" that I'm defending here comes from neither anthropology nor dendrology. It comes from cognitive science. One of the key lessons of cognitive science over the past few decades is the lack of bright lines that I described.

It's almost like saying grass thinks.

Having agency does not require being conscious of what you're doing, or explicitly thinking about what you're doing. All it requires is what I said above.

Of course, if you are conscious and can think, you can have much more complicated, much richer agency than a tree or a thermostat, as we humans do. But again, there's a continuum here, not a bright line. Humans don't magically have some property that no other organisms have. The difference with humans (and it is a big difference) is many, many orders of magnitude more complexity; and as Philip Anderson (IIRC) once said (he was talking about condensed matter physics, but I think it applies more generally), "more is different".

 

[sbrothy]

Since they don't send their roots everywhere, they clearly are doing some kind of "deciding" about where to send them and where not. So yes. Of course these "decisions" are very, very simple, much simpler than decisions that humans make. But that doesn't mean the amount of decision is literally zero. It means that "decision" is not a binary concept. Even a thermostat can be said to "decide" when to turn your heat or A/C on and off based on the temperature it detects and its settings. Trying to draw a bright line here is neither necessary nor practical.


No. But brains are not required for agency. All that's required is some ability to affect the outside world, and some kind of filtering that means only some of all the possible effects actually get realized. Trees have that. Thermostats have that.


See above.


The concept of "agency" that I'm defending here comes from neither anthropology nor dendrology. It comes from cognitive science. One of the key lessons of cognitive science over the past few decades is the lack of bright lines that I described.


Having agency does not require being conscious of what you're doing, or explicitly thinking about what you're doing. All it requires is what I said above.

Of course, if you are conscious and can think, you can have much more complicated, much richer agency than a tree or a thermostat, as we humans do. But again, there's a continuum here, not a bright line. Humans don't magically have some property that no other organisms have. The difference with humans (and it is a big difference) is many, many orders of magnitude more complexity; and as Philip Anderson (IIRC) once said (he was talking about condensed matter physics, but I think it applies more generally), "more is different".

OK, I'll meet you in the middle and admit that it's more of a spectrum than a binary divide, but I'm not gonna back down from this as I did with @pbuk giving a me well-deserved spanking telling me I'm in over my head. I'm gonna swallow that and see what else I can get out of him. Still, comparing the agency of trees to the agency of humans.... I just don't see it, sorry. The reason trees suck water up is because water evaporates from the top creating a vacuum that sucks water up from the roots. Calling that agency seem to me almost funny (If I wasn't discussing with people that I realize are much smarter than I am). I realize that and I don't want to provoke anyone so I'm going to hold my tongue from here on out.

EDIT: After all I know I'm an autodidact hack.

EDIT2: Wait are you saying now that a thermostat has agency?! A thing designed by a human engineer? Where does that "agency" come from you think?

Sorry couldn't resist. I'll s... up now.

 

[jack action]

Economics is about limited resources and how we prioritize how they get used: a free market and the price mechanism it embodies is a way of doing that.

There are plenty of cases where limited resources have to be prioritized for one essential need, where the free market is not a good control method.

For example, two cities on the banks of the same river, where both cities discharge their sewage into the river because it is cheap, while the river is also the source of drinking water for both cities. The downstream city has to pay more to clean the water before it can be used for drinking. If they can't clean it because what is sent into the river by the upstream city is too toxic, that is even worse. You cannot just compare the price of sewage disposal and the cost of drinking water locally without thinking on a larger scale. In such a case, the hidden price might be a future war between the cities, which a free market couldn't evaluate.

Another example would be cutting trees in the forest. At one point, you realize that those trees play a vital role as "lungs of the Earth" (among other things) by filtering pollutants and converting carbon dioxide into oxygen. No matter how much a tree is worth for lumber or firewood, one must restrain oneself from clear-cutting forests, as this apparently free role is irreplaceable.

People who depend on a well on their property understand that economics and free-market theories don't work in a drought. You don't reduce both a little bit of your water intake and shower time (higher cost of water in general): you reduce your shower time a lot while keeping your daily water intake. This basically increases the price of water for showers only. Reducing your water intake is the last resort when you can't afford showers anymore.

If people cannot afford drinking water on a large scale, industries (especially those considered a luxury) that use water for manufacturing or services will be the first to have their prices increased until they reduce their water consumption.

 

[PeterDonis]

Wait are you saying now that a thermostat has agency?

A very, very limited form of it, yes.

A thing designed by a human engineer?

The engineer isn't there telling the thermostat when to turn your heat or A/C on and off. The engineer gave the thermostat a tiny bit of agency when the engineer designed it to do those things on its own, based on the temperature it senses and the settings it has.

 

[PeterDonis]

comparing the agency of trees to the agency of humans

Which is not at all what I did. Please don't concoct straw men. I went out of my way to emphasize how different the agency of humans is from the agency of trees (and thermostats). Please read what I actually wrote.

 

[sbrothy]

I didn't consciously contruct concoct strawmen (I just had some perhaps tasteless fun, although not on your behalf although it might have looked like that), but I promised to stop my discussion with you and I will hold my word.

EDIT: Also time is 05:40 here. I really should go to bed.

 

[BillTre]

Any kind of simple cell able to respond to changes in its environment will have some simple form of agency. It is (in my mind) similar to computational agents that can make decisions to achieve certain goals depending on sensed environmental conditions.
Here is a book that goes into that general issue in depth:

By stacking biological agents to make possible more complex computations, teleology may result. Its like sets of logic gates.

 

[jack action]

To learn more about intelligent agents, Wikipedia has a nice little resume. Those are the classes of intelligent agents:

• Simple reflex agents
• Model-based reflex agents
• Goal-based agents
• Utility-based agents
• Learning agents

A thermostat would be a simple reflex agent.
ChatGPT and the Roomba vacuum are examples of goal-based agents.

 

[sbrothy]

OK you dragged me in anyway. It seems you left the definition of actual "agency out:

Actual Intelligent Agency
Definition
Actual intelligent agency, such as agentic AI, represents a more advanced form of artificial intelligence. It is characterized by:

Feature Description
Autonomy Can make independent decisions and set its own goals without constant human input.
Proactive Behavior Identifies needs or opportunities and acts on them without explicit prompts.
Complex Task Management Capable of coordinating multiple agents or systems to achieve complex, multi-step workflows.
Adaptability Can adapt to new situations and learn from experiences with minimal human guidance.
Key Differences
Decision-Making: Levels of agents operate within strict boundaries, while actual intelligent agency can make decisions based on broader contexts.
Task Complexity: Simple agents handle straightforward tasks, whereas agentic AI can manage intricate workflows involving multiple agents and systems.
Learning and Adaptation: Learning agents improve over time, but agentic AI can autonomously adapt to changing environments and requirements.
Understanding these distinctions is crucial for leveraging AI effectively in various applications.

Now my point is not really artificial intelligence but IMHO a roomba is nowhere near this level. Not to mention a tree or a thermostat. They may technically be designated as various degrees of agents (read: they are preprogrammed with very basic responses), but I'd still argue that they are nowhere near actual agency as humans, other mammals and more advanced marine animals display. It seems to me to be a very very academic distinction. To the point of mootness.

It still seems to me that you're arguing that roombas and trees display agency, but I think there's a world of difference. EDIT: And I know this is unfair. I realize that is not your point.

I already regret getting dragged into this again. I hope you can at least agree it is pretty academic.

 

[sbrothy]

"Enchanted Looms" is another good book about this subject.

 

[256bits]

Interesting discussion, all over the place connecting ideas.

teleology may result

Can one believe at the same time in evolution and its processes of randomness of natural selection, and teleology which in its final extension leads to some sort of divine intervention? Ie-the universe has a purpose.

classes of intelligent agents:

So, everything with a feedback loops becomes an agent?
Add some computer code and the thing automatically becomes smart with more intrinsic agency, even if the end result, purpose, or goal is the same.

https://en.wikipedia.org/wiki/Teleology
https://en.wikipedia.org/wiki/Teleonomy

 

[256bits]

To the point of mootness

It actually is.
Even so or has quite a few followers as it now extends more into common everyday life vocabulary with the overwhelming hype of AI. Cybernetics and all that bleed over.
I wonder how kids and graduates being told that the being human comes with 'free' agency and you can be anything you want, face the reality that one has to work at it, and yet the imagined goal may not realize.
A nihillistic society with (unrealistic) hopes and dreams dashed with an AI that has more valuable agency than you.
Are the Luddites to come back in force at some point.

 

[jack action]

but I'd still argue that they are nowhere near actual agency as humans, other mammals and more advanced marine animals display.

It still seems to me that you're arguing that roombas and trees display agency, but I think there's a world of difference.

So, everything with a feedback loops becomes an agent?

From the link I gave above, note that the first four classes of intelligent agents are pretty similar, each one having an add-on over the simpler one. The model for the fifth one is completely different, though. This might help reconcile the facts that every automated machine is an agent, and that there is also a big difference between the decision processes of a human (or even an equivalent AGI) and a thermostat.

Reflex

Goal-based

Utility-based

Learning​

 

[PeterDonis]

everything with a feedback loops becomes an agent?

The best heuristic I can give is that anything that can do something, even something very simple, on its own, without being told what to do, is an agent. A thermostat, for example, tells your heating or A/C to turn on and off, without being told when to do it. So it's an agent--an extremely simple one.

 

[256bits]

From the link I gave above

Intelligent agents can be organized hierarchically into multiple "sub-agents." These sub-agents handle lower-level functions, and together with the main agent, they form a complete system capable of executing complex tasks and achieving challenging goals.

The sub-agents designation leads to the suggestion that if the human is an agent, then every cell within the human body is an agent performing actions from stimulus from the environment.
Should the concept of agent be terminated at the single cell, or continue down into the cell internal structure including the proteins and chemicals within that react to their local environment .

It seems to have a similarity with the Schodinger's Cat problem in quantum mechanics, except the approach is top down rather than bottom up.