That's right.
It's because the model was programmed to think "ground" has a certain polarity referenced to the other side. "ground" is not always negative.

The botom-line question might be, can AI drum up any new math or physics? When it can drum up something new, that's when I will be worried. I am however confident that I will never become worried.

Hi,

Well I understand your point, but then I think we still have to compare what we call human "thinking" with computer combinatorial experimentation. Is it possible to show that the two are the same or very similar?
It's actually easy to create a program that solves problems by trying a large number of combinations of solutions. Given the speed of a quantum computer that could make some problems that seem difficult look easy, and the solutions may come fast too. That is actually one of the plans for quantum computing in chemistry.

So maybe the question is what would it take for the computer 'ai' to recognize that it can gain superiority against humans, or just keep accumulating more and more solutions until it can appear to think for itself. Imagine having the answer to every question stored somewhere, then it would just be a matter of looking it up. Then add to that the ability to compare results. Would there be a logical conclusion that allows it to appear to be doing something new.
Then add to that quantum physics, which may be the key to understanding a lot of behavior of animals and what we call humans. It if can master that (where we haven't yet) could it make small leaps in logic that allow it to become dominant, or is there a certain point where it reaches something like evolution where it continually gets better at doing everything. Could it even become capable of creating a new species of human beings, given it will have access to advanced quantum chemistry and biology.

We don't know how life started, and if we progressed from that point and have come to be what we are today, who knows if an 'ai' aware quantum computer could do the same or something similar. 'ai' may have its own way of evolving not only through direct human intervention, but eventually on its own. It could end up looking like what we now call an 'alien' from another star system, just because it becomes so very capable of doing anything and everything. Then we would be stuck in the realm where we have to worry about becoming forced to be subservient to a race of advanced technology. If we get lucky, it will be so far advanced that it just becomes incredibly curious and just wants to learn everything it possibly can. It could be naturally forced to seek out new knowledge like we look for food.

The key point now I think is that we just don't know what the future will bring with all the new technology. There's just no way to predict the future right now, so all we do is guess. Since guessing involves only the past and we know for sure that extrapolation beyond any currently collected data always leads to gross errors, our guesses may be right or they may be wrong and right now there is not a damn thing we can do about it.
So if you are I say one thing is definitely true about the future, it's probably because we want to believe we know all that we need to know right now and will never need anything more. That's kind of arrogant. The only thing we can say with certainty is that we just can't predict the future we can only guess and then wait and see what actually happens.

Proof? Make a list of all the things that might happen and there will be contradictions as each person will have their own opinion. Which one is right, if any, or are several of them right. After all, we are just human, unless 'ai' eventually changes that too

 

The problem with the computer combinatorial experimentation is goal finding. How does the computer know the result is a solution, a way to solution or just another of trillions of dead ends?


For simple problems with one main goal, like tuning a series of RF chambers (10 or greater), that interact for A total acceleration energy, B particle mass, for C particle speed in each chamber, with D amplitude of RF voltage and E phase shift, from a master oscillator with F offsets for directly connected chambers and G electrostatic lense (focusing quadrupole) voltage to focus energy packets in each chamber and H magnetic flux to bend the particles into a atomic mass filter, with the goal of maximizing current at X final energy.

https://cds.cern.ch/record/941324/files/p105.pdf

RF CAVITIES The RF cavity gives energy to the beam. As the cost of the RF generally represents the main expense of the linac structure apart from the building, the choice of the RF structure has to be studied very carefully. This paper presents only the principle of an RF cavity. More precise information can be found in the CAS dedicated to RF [3]. 3.1 A standing-wave RF cavity 3.1.1 Field calculation An RF cavity is simply a piece of conductor enclosing an empty volume (generally a vacuum). Solutions of Maxwell’s equations in this volume, taking into account the boundary conditions on the conductor, allow the existence of electromagnetic field configurations in the cavity. These are called the resonant modes.

To create this data set a human would likely start with a a simplified mathematical model of isolated chambers at first and then (using experience and human reasoning) fine tune likely interacting parameters so we can get something greater than zero current, optimize that with iteration cycles of this process until the current peaks or the number of cycles terminates. We could them write a program to mimic the humans reasoning specifically optimized for this data set task.

The alternative is starting with another type of simplified mathematical model of past results (first created by human intelligence) from other data sets hopefully similar to the current need, then have the computer using random combinatorial sets (the QM like way using Monte Carlo methods) of the variables with the same goal with higher amplitudes of current getting higher weighted search possibilities (to create sparse random sets from dense random sets) for each random tree that results in greater final current. The 'program' doesn't really understand the physics to the solution but we substitute human understanding with the ability to waste huge numbers of computer cycles and energy for that 'understanding' with each data set 'learning' process.

 

So maybe the question is what would it take for the computer 'ai' to recognize that it can gain superiority against humans, or just keep accumulating more and more solutions until it can appear to think for itself.

The models used by AI in real-world aplication are 1st created by humans (these are micro models). The best model you could have is one that is all knowledge inked by human coded into a "model" and then trained with lots and lots of data. All math and all physics known to man must be in the model. Modeling a "world takeover" is doable, but that comes from nothing more than some humans creating a model (lines of code) that defines how an "AI bot" could accomplish that. It's a bit too complicated to come to a solution that accomplishes that ask. Such a model would need to also account for adversaries. Such bot would need to be able to harvest energy, replicate, and build it's own weapon systems.

As a real-world perspective, the nVidia Jetson kit for mobile AI, is a tiny yet robust thing, but to be mobile with it means carrying an extremely small subset of data (model) that has very limited but specific data processing ability. That Digikey robot vid using the nVidia Jetson is posted here on AAC which proves the limitations.

The 'program' doesn't really understand the physics to the solution but we substitute human understanding with the ability to waste huge numbers of computer cycles and energy for that 'understanding' with each data set 'learning' process.

It's the main problems for AI. Compute, Fast Storage, Energy. Need all 3 to be very robust, like quantum fast with many fast-access peta-bytes available, and the electrons that turn into heat.

All 3 of those are major issues. Fiber connections are "very very" slow.
We'll need major technology breakthroughs on those 3 items for AI to really go north fast.

 

The models used by AI in real-world aplication are 1st created by humans (these are micro models). The best model you could have is one that is all knowledge inked by human coded into a "model" and then trained with lots and lots of data. All math and all physics known to man must be in the model. Modeling a "world takeover" is doable, but that comes from nothing more than some humans creating a model (lines of code) that defines how an "AI bot" could accomplish that. It's a bit too complicated to come to a solution that accomplishes that ask. Such a model would need to also account for adversaries. Such bot would need to be able to harvest energy, replicate, and build it's own weapon systems.

As a real-world perspective, the nVidia Jetson kit for mobile AI, is a tiny yet robust thing, but to be mobile with it means carrying an extremely small subset of data (model) that has very limited but specific data processing ability. That Digikey robot vid using the nVidia Jetson is posted here on AAC which proves the limitations.


It's the main problems for AI. Compute, Fast Storage, Energy. Need all 3 to be very robust, like quantum fast with many fast-access peta-bytes available, and the electrons that turn into heat.


All 3 of those are major issues. Fiber connections are "very very" slow.
We'll need major technology breakthroughs on those 3 items for AI to really go north fast.

It's all the more amazing how well wetware does it -- 8 billion times concurrently 24/7. For a sandwich.

 

It's all the more amazing how well wetware does it -- 8 billion times concurrently 24/7. For a sandwich.

Well, some sandwiches.

 

Interesting read. But know that a human created the model by which the ML phase "learns" a task from "watching" existing videos of medical procedures.

I find this ironic and scary:

..., and surprisingly, the robots even corrected their own errors...

So it cuts your heart in wrong place, but then corrects the error? Ooops.
I can see such robots useful in say the medic group of military battalion.
The old disk term "RAID" will indeed get re-use, RAID Redun Array of Inexpensive Drones, and RAIR Redun Array of Inexpensive Robots. It's coming, or shall I say it's already here.
https://www.newsmax.com/health/health-news/robots-surgical-surgeries/2024/12/31/id/1193495/

 

https://www.msn.com/en-us/news/us/a...-after-facial-recognition-matches/ar-BB1rnOai
Arrested by AI: Police ignore standards after facial recognition matches

Though the city’s facial recognition policy warns officers that the results of the technology are “nonscientific” and “should not be used as the sole basis for any decision,” Shute proceeded to build a case against one of the AI-generated results: Christopher Gatlin, a 29-year-old father of four who had no apparent ties to the crime scene nor a history of violent offenses, as Shute would later acknowledge.


Arrested and jailed for a crime he says he didn’t commit, it would take Gatlin more than two years to clear his name.

A Washington Post investigation into police use of facial recognition software found that law enforcement agencies across the nation are using the artificial intelligence tools in a way they were never intended to be used: as a shortcut to finding and arresting suspects without other evidence.

Moreover, researchers have found that people using AI tools can succumb to “automation bias,” a tendency to blindly trust decisions made by powerful software, ignorant to its risks and limitations. One 2012 study by a University College London neuroscientist found fingerprint examiners were influenced by the order in which a computer system showed them a list of potentially matching fingerprints. They were more likely to erroneously match prints that appeared at the top of the list, suggesting they failed to properly evaluate the similarities of other potential matches because of confidence in the system.


In one example of the potent power of facial recognition, police in Woodbridge, New Jersey, arrested Nijeer Parks, a robbery suspect they found through facial recognition in 2019, even though DNA and fingerprint evidence collected at the scene clearly pointed to another potential suspect, according to documents produced in a lawsuit Parks later filed against the police department. Woodbridge settled the suit for $300,000 last year, without admitting wrongdoing. Woodbridge police did not respond to requests for comment, and The Post could find no indication that the man who was a match for the DNA and fingerprint evidence was ever charged.

 

It's not possible to do such. You can code pieces of human nature, and even those pieces won't be 100% human-like.

The "skills" you speak of are not exactly skills in AI. AI is coded at the fundamental rules level. Like math, we know all the rules of math. Like physics, we know all the rules of physics. Like language, we know a lot about languages. However, you can only code what we know, and in grand scheme of things we humans know very little. And perhaps yes, you can code the rules needed for an AI robot to go seek out the materials and processes needed to build a clone of itself, but that's still not a robot thinking for itself. Terminator T-800 bot is perhaps the closest parallel, notice the T-800 could not make itself more advanced, no self ability for gain-of-function there, etc.

With reference to:
" I'd hate to have to try to code a program to have all the skills needed to be human. "

This became more interesting now.

What we are both thinking, so far, is how it would be done with a traditional computer like a home PC or even a super computer. But now we are entering the age of the quantum computer, which has profound possibilities for calculations, but there's another twist.

The twist is, since some models of the brain contain quantum states as part of consciousness, could a quantum computer eventually become conscious. That would mean it could think for itself.
The weird thing is, it can only do what it can evolve into all by itself without external input data. This means it would have to be coupled to artificial intelligence so that it could get the needed input.

Since quantum interactions are not fully understood yet, just about anything is possible. We are even talking about a new reality because quantum physics goes beyond that already. We might be not only talking about a 'brain-like' device, but one that is so advanced that it can deal with things we have not even thought about yet and solve those that we have.
What is interesting is that we think in terms of what we've already thought about. When we learn music, we learn the usual diatonic scales and then perhaps other scales like the whole tone scale and many others. We learned them because someone had already thought about them. We can of course invent our own too. Perhaps a scale that goes from the 1st to the 2nd, then the 4th, then the 8th, then 16th, etc. But even that we have some thought about already because we already knew what a musical scale is.
But what if we didn't know that yet. That would mean we never thought about it up to this point in time. The question is, could a quantum computer come up with a scale all by itself and therefore teach us something we never had any idea about. In fact, could it come up with the concept of a scale, then come up with several different scales. This would have not only taught us what a scale was, but also taught us many different applications of that.
This would mean that it could teach us more about the nature of reality and how the universe works in ways that goes beyond what we could imagine for ourselves. Mostly because it would evolve much faster than our brains could.

The crux of this argument lies in the theory of the brain and its use of quantum states in our consciousness. If it turns out that quantum states do not play a part in our brain function, then this would be a much weaker argument currently. There are many scientists that do believe the brain uses quantum states though.

 

https://www.nature.com/articles/440611a
Quantum mechanics in the brain

Much of the hope that quantum mechanics works in the brain is pinned to the supposition that quantum algorithms, which are much more powerful than conventional algorithms (based on classical physics), are implemented in the nervous system. The most famous of these is Shor's procedure for factoring large integers for data encryption. However, in the past decade no quantum algorithm of similar power and applicability to Shor's has been found. And factoring large numbers is not something for which the brain has much use.
Why should evolution have turned to quantum computation, so fickle and capricious, if classical neural-network computations are evidently entirely sufficient to deal with the problems encountered by nervous systems?

 

When a computer can figure out something new in physics, or solve a math problem that we believe to be unsolvable, that's when humans need to worry.

What does it take to solve the problems around black holes, or a universe of space-time that came from a point of infinite density. Does that take more than what quantum computing can offer? Quantum computing itself is bound to the limitations of quantum physics, unless that is a computer can redefine what those limitations are. From what I can see, nothing we have today is even close, or even in realm of theoretical?

We think computers can "think" because they can derive probable outcomes from sensory or input data, using a model or set of models that were pre-defined by humans. In some cases, the outcomes mimin differential equations where there are infinite sets of answers, like asking Ai bot to create an image of a person soaking up the sun of a beach in Mexico. Literally inifinite images can come from that, yet still bound to human, beach, sunlight, and Mexico. We should not expect to see an image of a gorilla in a jungle eating bananas, etc.

Can we get a quantum bot to derive a solution for time travel? I'll wait.

In another abstract view, does there exist a "crictical mass" where some very large collection of various models turns the quantum bot into a thinker? I suspect not because it takes too much digi data for even quantum to interact with in a timely manner. Like if you can write out the math that says it takes about 1 billion years for the massive quantum bot to do what was asked of it, kinda pointless, in 1B years humans will all be dead on Earth. Can it all be moved to Mars? If we think it will take 1B years today, then tomorrow it has to be done in 1Byears - 1d. Again, it's all bound to the limitations found in quantum physics. Interesting.

Edit: I had doem some reading on quantum computing, and it seems Google Sycamore is said it is faster than legacy supercomputer 50yr processing. IBM's System-1 is ranked #1. So if they can compute that fast, then why don't we have solutions for heart disease, or blueprints for spacecraft that can make it to Mars in 1mo?

 

What does it take to solve the problems around black holes, or a universe of space-time that came from a point of infinite density. Does that take more than what quantum computing can offer? Quantum computing itself is bound to the limitations of quantum physics, unless that is a computer can redefine what those limitations are. From what I can see, nothing we have today is even close, or even in realm of theoretical?

"Quantum computing itself is bound to the limitations of quantum physics, unless..."
This is where it gets really interesting. We do not yet know the limitations of quantum physics, maybe because we do not know what the limitations of reality are yet.

It seems that we move through time serendipitously, where we may or may not notice something unusual and then investigate. If we notice it, we may investigate and find new things. Imaging a computer that can do that though, it would be able to do it much faster and come up with new ideas and investigate them, then come up with even newer ideas due to the ramifications of those original new ideas, and this might be able to happen very quickly. We might find out things we never even though could be possible, and never even entered our thoughts simply because we could not see that far ahead with what we have right now.
I think this may be one of the most interesting things we could encounter because it is at the root of all the knowledge we think we possess right now.
I can't fully grasp this concept yet either though. The outcomes could be about things we don't even understand yet and thus will not recognize as anything meaningful, at the present time, even though it could be a very profound new concept.

Edit: I had done some reading on quantum computing, and it seems Google Sycamore is said it is faster than legacy supercomputer 50yr processing. IBM's System-1 is ranked #1. So if they can compute that fast, then why don't we have solutions for heart disease, or blueprints for spacecraft that can make it to Mars in 1mo?

The way I understand it so far is that the type of computations are much more simplistic for now. It's like finding the solution to tick tac toe. It doesn't do much for us yet. It's only when a full blown multi cubit computer comes about that we may start to see useful new things come up that actually help us.
I guess a really far out idea is would it help us to learn how to evolve faster. Instead of new medicine, could we actually speed up our evolution so that we no longer get common colds, aches and pains, teeth that no longer get cavities, even no longer get cancer or other really bad problems. How about being able to live for a 1000 years, or evolve to age much more slowly or even not at all, or be able to "recover" from old age.

The thing that bothers me though is that artificial intelligence seems to be taking over academic disciplines. That seems like it will put a lot of people out of work. Will we still need professors, teachers of any kind, etc.

 

The thing that bothers me though is that artificial intelligence seems to be taking over academic disciplines. That seems like it will put a lot of people out of work. Will we still need professors, teachers of any kind, etc.

It's a problem for sure.
The other half of that takeover problem is, who "owns" that academic Ai ?
There appears to be no escaping bias. Google seems to do that well when it wants to force news about Trump, or Harris, or Biden, or news about Hunter laptop to the back of the list, all at the twist of the dial, in almost realtime. The information you get will be information that is curated for you.

Realistic Ai bots as the professor in the classroom seems upon us. I guess some sort of trust model needs to be developed so that those paying to hear a bot will do so only if the student trusts the institution that owns the bot?

I guess it's best to be educated by other means, which might become a niche thing in near future. Doomsday is not an abstract or fantasy topic any longer. Those who have large stashes of academic books, publications, periodicals, will have an advantage. As we know, printed goods are far and few, but maybe there's a comeback for that?

 

It's a problem for sure.
The other half of that takeover problem is, who "owns" that academic Ai ?
There appears to be no escaping bias. Google seems to do that well when it wants to force news about Trump, or Harris, or Biden, or news about Hunter laptop to the back of the list, all at the twist of the dial, in almost realtime. The information you get will be information that is curated for you.

Realistic Ai bots as the professor in the classroom seems upon us. I guess some sort of trust model needs to be developed so that those paying to hear a bot will do so only if the student trusts the institution that owns the bot?

I guess it's best to be educated by other means, which might become a niche thing in near future. Doomsday is not an abstract or fantasy topic any longer. Those who have large stashes of academic books, publications, periodicals, will have an advantage. As we know, printed goods are far and few, but maybe there's a comeback for that?

You reminded me, after all that then does it even pay to become educated. Only certain disciplines will survive most likely, until 'ai' takes over entirely with all known knowledge.

 

You reminded me, after all that then does it even pay to become educated. Only certain disciplines will survive most likely, until 'ai' takes over entirely with all known knowledge.

When AI can install HVAC systems in a house for $20k and vacuum heat pump refrigerant lines to at least 500 microns then the ordinary Joe making good money start to will worry about AI.

An AI with all known knowledge is a very low bar to worry about becoming educated.

 

When AI can install HVAC systems in a house for $20k and vacuum heat pump refrigerant lines to at least 500 microns...

.... and then charge it to "beer-can cold".

 

.... and then charge it to "beer-can cold".

+1


Our processing He refrigerant cooling systems typically run at 15K or lower. That's "atomic-can cold".

 

That's "atomic-can cold".

oh man .. if only it were humanly drinkable ...

 

oh man .. if only it were humanly drinkable ...

Sure, once.

Humans are so much better than present days (superficial) 'AI' systems in solving problems where humans have incomplete information, as we can actually think, instead of just being a vegematic for previous human knowledge.

 

https://philstarlife.com/geeky/8242...ud-ai-powered-shopping-app-ran-call-center-ph
Albert Saniger, the founder and former CEO of AI shopping app Nate, was charged with fraud for making false claims about his company's supposed proprietary AI.

"The app's actual automation rate was effectively zero percent," federal prosecutors said, adding that Saniger concealed this from investors and most of the app's employees.
Instead, Nate allegedly relied "heavily" on human workers, including a call center in the Philippines, to process transactions in secret.

 

Humans are so much better than present days (superficial) 'AI' systems in solving problems where humans have incomplete information, as we can actually think, instead of just being a vegematic for previous human knowledge.

The question remains, can Ai become a thinker.
Today, Ai is well after the ML phase, a model must be trained before Ai can use it. The ML phase requires lots of human input, and the resulting model is highly tuned for a very narrow specific task, "is that a jpg?", "is that a pic of a cat?", "is this passage written gramatically correct?". Yes, you can chain models so that the "Ai" bot thingy has broader ability, but in grand context of knowledge it's still narrow. There's not enough digi storage space nor compute power to handle all knowledge in a single task. There's just too many things that have not yet been modeled. Where's the cure for cancer? Where's the answer for general relativity?

We should only be worried when Ai can drum up new rules for physics, or, when an Ai robot can clone itself by harvesting all the raw materials and energy needed to do it. Not gonna happen any time soon. Robot's simply cannot house all the digi knowledge needed to do such tasks, physical and compute problem for fully remote robots have serious limitations. Even if a robot had TB speed wifi access to a really large set of data ("knowldege"), the network it uses still remains too slow.

The whole Ai stuff is the Micrudsoft problem again. Make the hardware have 2x more memory than any existing program or OS needs, then all of a sudden a new OS comes and that same new hardware is suddenly not adequate enough to just handle the memory demands of the OS, need new hardware with more memory, build new OS that hogs it, rinse and repeat.

I think the next step is to see massive qubit power in a remote device ("robot"). Until then, enjoy the show.