[hamdani yusuf (OP)]

By defining consciousness as capacity to pursue goals,

So a homing missile is as conscious as a homing pigeon?

Can they reproduce? Adapt to their environment? Build nest? Compete for resources?

Quite a few humans can't, And AFAIK no computer can.

Computer software has done those things in virtual environment. When they have access to physical environmentenvironment and resources, there's no convincing reason that they will never be able to do those things.

[alancalverd]

Computer software has done those things in virtual environment.

Which is a roundabout way of saying that they haven't done them. I have flown to Mars and bombed the Mohne dam in a virtual environment. You don't get medals for not actually doing something.

[mxplxxx]

The goal seems to be an infinitely intelligent, infinitely aware  being.

[hamdani yusuf (OP)]

The goal seems to be an infinitely intelligent, infinitely aware  being.

It's an instrumental goal towards the universal terminal goal, and part of consciousness. Although the distinctions are being blur as they get closer to infinity, because the infinitely intelligent being must be aware of the universal terminal goal itself.
 
The universal terminal goal must apply in the past, present, as well as the future, as long as there is at least one conscious entity in the universe.

Being intelligent and all-knowing increases the probability of conscious systems to survive and thrive in the long term.

[hamdani yusuf (OP)]

Computer software has done those things in virtual environment.

Which is a roundabout way of saying that they haven't done them. I have flown to Mars and bombed the Mohne dam in a virtual environment. You don't get medals for not actually doing something.

Not yet. Computation is just one component of consciousness. That's why I said I prefer the holistic approach for consciousness.
Combining AI and robotics like what's being done by Tesla and other tech companies can make the difference in not so distant future.

[hamdani yusuf (OP)]

I planned to make a video about natural consciousness, and how functional components of consciousness can emerge from natural processes.

Finally, here you are.

This video describes how complex systems like consciousness can emerge naturally.

I put a link in my Youtube video above to this one below as a reference.

The Evolution of Intelligence

Unlock the essence of intelligence by exploring the layers of learning. This video follows the progression of evolutionary, experiential, and abstract learning, forming the bedrock of artificial intelligence. It provides insight into various learning paradigms including unsupervised learning, supervised learning, reinforcement learning, association learning, and the ingenuity of genetic algorithms. As part of the narrative, the essence of language and its role in advancing intelligence is explored. This is Part 2 of my enlightening AI/Deep Learning series, serving as a bridge to understanding modern AI frameworks like ChatGPT and GPT models. Embark on this intellectual journey to grasp how the lineage of learning has sculpted today's AI landscape

[hamdani yusuf (OP)]

There is no terminal goal. 

There are two kinds of goals: terminal and instrumental. A terminal goal is an end in itself. While an instrumental goal is a condition pursued to achieve the terminal goal. By definition, instrumental goal requires terminal goal to exist.
Denying the existence of terminal goal means also denying instrumental goal, which means denying the existence of goal.

[hamdani yusuf (OP)]

We need to research Full Autonomy RIGHT NOW - A call to action for OpenAI and others - Reduce X-Risk

6:07
people have pointed out open source models are only a few months behind closed Source models
and the first thing that happens is they're all jailbroken.
 so you know yeah putting putting guard rails on a commercially available API.
 great that is not a permanent or long-term solution outside of a commercial deployment and
so when you you know we have to assume that in the future there are going to be
super intelligent open source models that are fully jailbroken and uncensored
that's just a fact of life that we're going to have to deal with.
 which means we should be researching how to create self-detecting or self-directing
self-correcting and self-improving models now so that we know how to do
that and then also there is the uncertainty of corrigibility
so this is something that I do agree with some of the doomers which is good luck
controlling something that is a million times more intelligent than you.
yes right now they're just inert machines yes right now they rely on you know
billion dollar data centers and you know super expensive gpus and a tremendous
amount of power so we still have the power switch.
 we should not assume that we will have the power switch forever into the future.
 and so the combination of The Duality of intelligence and the uncertainty of corrigibility means that
like right now while we have control is when we need to be researching full
autonomy because in the long run what's going to protect us from an evil AI
that's going haywire or maybe even not an evil AI but just something that is malfunctioning.
 we need something that is benevolent that is good that is stable
to help you know kind of level the playing field

[alancalverd]

The man with his hand on the "off" switch is always in control. That's called authority.
The man who made the decision to deploy AI (or any other device) bears full responsibility for the outcome.
Authority can be delegated, responsibility cannot.
It doesn't matter what havoc the machine causes: identifiable humans are liable to compensate the victims.
Followers of the Post Office Horizon scandal will be familiar with the scenario.

 

[hamdani yusuf (OP)]

The man with his hand on the "off" switch is always in control. That's called authority.
The man who made the decision to deploy AI (or any other device) bears full responsibility for the outcome.
Authority can be delegated, responsibility cannot.
It doesn't matter what havoc the machine causes: identifiable humans are liable to compensate the victims.
Followers of the Post Office Horizon scandal will be familiar with the scenario.

 

Future AGI systems are expected to have distributed agency and redundancy, to make them more resistant from being inadvertently turned off. But it would also make it harder to intentionally turn them off, especially when someone else have the ability to turn them back on, or turn their redundancy system on to take over the control.

we should not assume that we will have the power switch forever into the future.

[alancalverd]

Never mind the AGI system. The human who introduced it remains liable for whatever it does, and can be switched off in the usual way.

[hamdani yusuf (OP)]

Never mind the AGI system. The human who introduced it remains liable for whatever it does, and can be switched off in the usual way.

It won't stop the rogue AGI, would it?

[alancalverd]

The threat of criminal prosecution or direct reprisal might dissuade anyone from switching it on in the first place.

So far, every actual or potential application seems to have been expensive and pointless (MacRobots), dangerous and pointless (self-driving cars) , or a means of diluting truth with indiscriminately recycled internet garbage (ChatGPT etc).

[hamdani yusuf (OP)]

The threat of criminal prosecution or direct reprisal might dissuade anyone from switching it on in the first place.

So far, every actual or potential application seems to have been expensive and pointless (MacRobots), dangerous and pointless (self-driving cars) , or a means of diluting truth with indiscriminately recycled internet garbage (ChatGPT etc).

Some countries can decide to regressively ban AI outright. But other countries with more sensible AI regulation will outcompete them economically. Which means the ban would miss their goal.

Apollo program was once thought to be expensive, dangerous, and pointless by some people. Nevertheless, the show still went on as it was, bringing numerous scientific and technological advancements.
Humans won't be happy to work at fast food restaurant while being paid less than living wage. Future improved robots have no problems with that. Some current self driving cars are already better than average humans in most situation. And they learn about new driving experience from millions of other cars, unlike humans who only learn from their own, and perhaps some of their mentors. It makes their learning rate would be much higher than humans, and soon they will exceed the skill of the best human drivers. LLM is just one of many AI models. Weaknesses in one model will be covered by other models, which can be integrated into the AGI. I posted progress towards AGI in my other thread.
Once in a while, you might encounter the AI to make mistakes, especially in new things beyond their training data. In this case, the model will soon be improved to prevent them from repeating the mistake. Don't forget that humans also make mistakes, and it's harder to make them stop repeating their mistakes.
 
Make no mistake, the progress is exponential. It may look slow in the past, but it will arrive rather sooner than most people expect.

[alancalverd]

No point in trying to ban something that helps governments to suppress opposition, or "helps the economy to expand" at the cost of a few lives. The best we can hope for is to prosecute those responsible for its implementation.

[hamdani yusuf (OP)]

This video describes how complex systems like consciousness can emerge naturally.

The next video I plan is about extended consciousness. It externalize some functions of consciousness to something outside of the conscious systems. They can extend the function of input interface, data processing and storing, output interface, or combination of them.
Sight glasses, telescope, microscope, are some common examples for input interface. Lever, pulley, and wheels are some common examples for output interface. Calculator and books are examples for data processing and storage.
It may not be a popular view, but descendants can be considered as a form of extended consciousness.
Soldiers at the outposts, spies, and secret agents are extended consciousness of military commanders.

[hamdani yusuf (OP)]

The distinction between extended consciousness and inherent consciousness might be blurry in some cases. Sight glasses maybe firmly classified as an extended consciousness. But optical implants might be debatable, especially when the implant was grown up from the host's own cells.

[hamdani yusuf (OP)]

It may not be a popular view, but descendants can be considered as a form of extended consciousness.

The argument can be extended further to include grand children and great great grand children, vertically in the lineage. But it can also be expanded horizontally, such as twins and other siblings. Combinations of both vertical and horizontal lineages will include nephews, cousins, grand nephews, grand cousins, etc. With enough expansions, our extended consciousness could include the whole species, genus, or even higher taxonomic category.

[alancalverd]

If you keep redefining consciousness without actually defining it, you can convince yourself of anything. That slippery slope leads to politics and economics.

[hamdani yusuf (OP)]

If you keep redefining consciousness without actually defining it, you can convince yourself of anything. That slippery slope leads to politics and economics.

I have defined consciousness as capacity to pursue goals, and never change since then. My latest posts regarding the extended consciousness are describing the boundaries of conscious entities. How functional components of consciousness can be classified as inherent within a conscious system, or rather an extension from outside.

What's your definition of consciousness?