Grab a six pack! Let's talk about living in virtual reality.

Well, let's meet half way.

I could see the laser being used to look into quarks and their connection to multiple dimensions. And I know I'm being VERY cynical about the whole space travel stuff which does no one any good to focus on.

The whole thing came up in the first place because of whether or not interstellar travel was viable and gave us a reason not to live in VR.

I would really like to know what people think in big science scale, what they think of VR.

 

Honestly I'm not convinced it is or is not possible. I was simply adding to the viewpoint that we cannot simply write it off yet.

 

Fair enough. I disagree that FTL is possible, but that's just my opinion.

At any rate, I really do appreciate the reading material. Thank you.

 

Why not duplicate the universe within the VR? We could send out probes that continually scan the universe and return the necessary data to the simulation. Travel within a VR should be nearly instantaneous after all.

 

IDK, about copying reality into VR for exploration. But I can DEFINITLY see running computer simulations for gene therapy and genetic engineering. Or testing theoretical physics in simulators. Or mapping what we see in the universe to better understand how things effect it.

 

Wow, I didn't think that tangent would merge back into the topic. But look at that, it did.

@Not_Sure if I take my entire family into space, can I avoid the effects of relative distance?

( ͡° ͜ʖ ͡°)

 

That sorta has been done before. Remember Folding@Home? Those guys were running distributed protein folding simulation.

Genetic engineering simulation would require one hell of a supercomputer, far beyond modern machine capabilities, significantly more powerful than what you'd need for VR.

There is openworm project that tries to simulate one of the simplest lifeforms that have neural system (C. Elegans, 306 neurons). That one requires 5 teraflops to work in realtime, IIRC.

 

Interesting topic. Got a buddy who's a time-traveler - he sent me this video from the far future:

 

So 1TFlop is roughly equal to 62 neurons? That sounds like useful information.

 

Actually, those teraflops were required to simulate worm's body.
The simulated worm has cells, musculature, etc.

IIRC, neural system simulation took 0.5 Teraflops.

 

Okay, considering there are an estimated 100 billion neurons in the human brain, we only need 15.6 ExaFlops to simulate a human brain. Considering the fastest supercomputer can only achieve 33PetaFlops (with a theoretical limit of 54 PetaFlops), we still have a long way to go.

 

Oh, thanks. So 1 tflop is enough power to simulate 712 neurons? That's cool.

It's too bad we don't only use 20% of our brains in the way that people generally understand that. It'd be cool to be able to use the other 80% of your brain to assist your personal computer with big calculations.

 

A virtual reality simulated by the currently living populace? It has the makings of a science fiction story.

 

On the topic of FTL even Nasa have a project looking into Warping Space so the vessel can 'move' around the relativity problem.

On the flip side genetic research that could allow people to have extended healthy lives is making progress. So even none FTL or Warp drives could be an option.

Actually if you factor in relativistic effects (time is slower for faster moving objects) short none FTL flights are probably doable with young astronauts to our nearest neighbours.

Actually where VR could be amazing is in the field of remote humanoid robotics.

 

There's most likely more than one way to simulate neuron, and there's no reason to think that openworm implementation is the most efficient one. Bluebrain probably uses something else.

It was estimated (by IBM) that "computational power" of human brain is 38 petaflops. Petaflop is a thousand teraflops.

Have you seen this picture?


With the rate of technological progress using biological brain for calculation will quickly become pointless.

 

But it's a great budget machine that runs on very little power!

If the brains are used to watch people and determine who are likely to be criminals and arrest them before they've done anything, that work of fiction is called Psycho Pass.

Why are the neurons different? Don't they just relay messages to other neurons and occasionally leak signals to ones they didn't intend to?

 

Neurons are not all that well understood, and almost certainly don't resemble very much the neurons in a computer-based neural network. I sort of did a 'course' on computational neuroscience from Coursera (which is more of just an informational overview). It seems neurons have more capabilities than you would expect in a a simple computational neuron, for example IIRC they seem to have the capability to store and add information to the information that passes through them, although exactly how this functions isn't well understood.

The idea being that simulating a human brain isn't as simple as running vast arrays of simple calculators (computational neurons).

 

Yeah it sounds like neurons can do internal computations, they are complex branching cell structures and not single binary switches after all.

There are also some theories that Neurons might be using quantum effects to process information. And as you know quantum stuff can become entangled so two quantum bits can work together, so who knows some of the spookier things might be quantum effects in the brain.

There was one scientist who was culturing, feeding and training brain neurons (from an invertebrate) to learn tasks. Think about it in a few dollars of petri dish you could grow a Watson beating jeopardy player or Deep Blue beating Chess player.

Q: How many silicon atoms can you fit in 7nm (Intel are working on 7nm chips)?

A: About 0.2nm for the silicon used in chip fabrication.

So at 7nm you have 35 silicon atoms, at some point a smaller and smaller silicon based transistor will run out of atoms, and hit massive quantum effects (you no longer know where your electrons are, in the circuits).

Now factor in that the silicon chip industry is having to spend more and more to make these smaller and smaller chips. So your graph is not going to be accurate unless some amazing new manufacturing techniques spring up very soon.

Science is charting existing data points, theories are projecting their trajectories.

Or think of it this way, in the 1800's there was a massive boom in railroad construction, and you could have probably drawn a graph like that. But there were physical limits inherent to the growth of the technology. Can you see the reverse micro/macro similarities.

What if it's cheaper and easier to start growing AI's in a petri dish, than it is to reach the 38 petaflops and probably gigawatts needed to build one in silicon?

After all we are discussing this complex issue, using image processing, character recognition, language processing and cognition using a 1.3-1.4 kg wetware brain that only needs 2000 calories to power its biological 'robotic' host.

 

My intuition says that the nature of consciousness has much to do with quantum mechanics and the nature of the 'observer'. And the ability of a particle to have multiple simultaneous states seems (at a naive level) to have some sort of analogy with the ability to 'be oneself and simultaneously observe oneself', two seemingly contradictory states that somehow coexist.

Then again, intuition is probably the last thing you'd want to apply to a problem involving quantum mechanics. But I simply can't believe that the simple addition of complexity would bring consciosuness. Quantum mechanics seems to be the only area of physics that is strange enough to potentially explain such a strange phenomenon as consciousness.

 

ah, that's a shame.

 

There is the theory that consciousness is an illusion, your brains way of saying, hey I did lot's of processing on this, in the background while you were in the shower or doing some routine task, here's a bright idea, but don't thank me you thought of it!

Or an emergent property of complex systems. Bees can build hives, Ant's or Termites giant air conditioned homes in desert conditions. Maybe consciousness is an emergent property of having billions of neurons working together. Although neuroscience is making great strides in tracking brain activity with scanning technology.

There is the idea of the chinese room theory where a man in a room who does not understand chinese, but has a look up chart and rule book. Receives chinese messages looks them up goes through the rules and picks out the characters needed to reply. The theory is with a complex enough rule set you could give the impression that the man in the room can speak chinese.

What if your brain is the man in the room and your consciousness is chinese.

 

I just don't understand how consciousness could be an emergent property of complex systems, unless there is some property within a system of any complexity, that only manifests in an obvious way at a certain level of complexity. It seems to me at least that consciousness is something additional to what is within a machine, at least as we understand them in functional engineering terms.

What do you mean by illusion exactly? Even an illusion exists within reality.

That's by far the most difficult problem with understanding consciousness, there is no objective measurement system available or seemingly possible. I know I'm conscious, according to my own definition of consciousness, but all of you guys might just be complex machines that appear to be that way how do I figure out if you're really conscious as I am conscious.

That for me is the strangest thing about consciousness, that I can experience my consciousness and you can experience yours, in the same reality.

 

There's connectivity issue. C elegans has 306 neurons and IIRC has connectivity mapped. That kind of system can have 306^2 = 93636 connections(synapses) max, in reality it has about 7600 of them.

Now human brain has 86,000,000,000 neurons. and can have 10^15..10^18 connections between them.

The bigger the system is, the more connections a neuron can have.

Also, the issue is that aside from electrical signals brain functions are affected by chemistry, which is less likely to happen within tiny system of OpenWorm.

By default, I'd assume that this kind of theory is a hogwash. "Quantum effects" are popular buzzwords frequently used to justify anything.

So, according to non-communication theorem, entanglement cannot be used to transmit information.

IBM is said to be working on 1.8 nm chips at this moment. I'd expect current semiconductor technology to be eventualy phased away by something more efficient.

Unlikely. You'll need to build ai chip once, then you'll be able to create millions of it.
An organic brain growth is harder to control, maintenance of organic matter is messy business.

Now, let me ask you something. Do you want your computers and robots to poop?
Or how about having to feed that thing every day because if you stop doing that it will die and you'll have to buy a new one?

In organic brain, language processing takes few years to train, for example, and skills require few years to train them. Preparing an adult human takes about 18 years. Don't forget that.

 

And silicon chips need power and generate waste heat and noise as well as pollution from the energy provider (depending on type of power station used).

Just like silicon based neural network software tends to be trained and built to do a specific task, I would guess a similar approach would be needed for wetware systems. Less of a brain in a jar and more of a network of petri dishes.

You should also be able to recycle the waste products from wetware systems into a hydroponics type system that would grow the nutrients needed to feed the wetware. So you should be able to build a solar powered, bio-AI system in a sustainable way.

Actually most people learn language in the first few years.

The fun begins when you want to mass produce wetware systems and have to genetically encode the networks knowledge into the system so it grows into it's job. In a similar way to how the lower parts of the human or reptilian brain work.

 

Still better than having your computer plugged into sewage system and it leaking blood all over the place or getting sick, or starting to stink. Wetware is messy.

Nope. Majority of the software today is not being trained and does not use neural networks.

Who, in their right mind would invest into something like this when you can mass-produce chips instead?

And it will be greenhouse sized.

It is not a good idea. Biological materials require very specific conditions that needs to be maintained. In case of machine, you can shoot it to the moon if you want, and it will work there. It is easier to produce electricity than it is to design and grow food for biological system.

Ugh. Genetics do not store language knowledge. Most of the useful stuff is learned. And reptiles are not very useful for language processing. Or doing anything, for that matter.

 

Only if it's actually designed to survive in space.

 

Well, not needing food, oxygen, air or water helps in that case.

 

Good point I think Nasa were still using early era chips because they were less likely to glitch when hit by a cosmic ray than modern chips with smaller transistors.

The reptilian brain gives you your fight or flight reflexes, those almost hardwired responses that bypass your thinking mind and can save your life, or give you the fright response in a scary movie theatre.

And allow reptiles and arguably birds to survive and thrive in a complex world using much smaller brains than ours.

My point is that if we can figure out how to grow and train biological neural networks then how to reverse engineer that network into dna then we could grow and mass produce wetware systems.

Take our optic system for instance it would appear that it is not a learnt system where random neurons in the brain learn to see but a grown set of systems that have been evolved to work well at image recognition and are plumbed into our eyes outputs.

So can we grow neural networks with pre built features or knowledge encoded already. That is the key to mass producing wetware AI systems.

 

That sounds like pipe dream to me.

People tend to take simplest option. So, you can make a bigger chip OR you could try to establish the whole new industry that will cover genetic engineering, manufacturing and support of wetware, and automatic hydroponics.

For the second option - wetware - the industry needs to provide significant beneift before someone decides to start dumping money into it. And right now I don't exactly see it.

 

Right. I was thinking more along the lines of everything else. Radiation, temperature, etc that affects wetware too.

 

I think wetware will be a spin off from the growth in genetics and bioscience, which is very dependent upon hardware to analyse and understand it's complexities. Also once a critical threshold in understanding how genes can be used to grow neurons into a specific network structure the technology will become very cheap very fast.

Or maybe a hybrid system will emerge where bio-manufactured silicon chips, use cultured cells, viruses or bacteria to grow silicon or carbon based computer chips.

Mind you what could really boost silicon based AI is the memristor, which wasn't mentioned on your transistors per $ graph.

 

I think that possibly the advantages offered by wetware systems belong to a time when there was no advanced society with a technological support base to be able to sustain anything better. Maybe organic systems can adapt, evolve, grow, and resist wear and tear much better than silicon/metal, which offer greater potential but were not viable materials during the early stages of evolution. Therefore I would imagine seeing in the next few hundred years a move toward becoming 'cyborgs', requiring an ever more complex and precarious technological support base, but moving beyond the possibilities offered by organic materials.

This is pretty much just speculation though, I haven't really looked into it in depth.

 

vs

 

Now if they had a VR enabled/driven robot I think it would kick butt or at least not fall over as much in the Darpa challenge.

 

Would you kindly elaborate on "etc" part?
I might be missing something, but radiation and temperature are pretty much it. Oh, and the fact that hardware needs to be designed to last for 25 years or more.

size of bacteria: 8000 nm. Size of virus: 20 to 400 nm. Viruses cannot move and are not considered to be life.
manufacturing of silicon chips requires completely sterile environment.

I highly recommend to stop fantasizing and do some research about subject you're talking about . A lot of stuff you talk about seem to be relying on info from dubious sources or hearsay.

No offense.

 

That's irrelevant. Robots tripping over is very unlikely due to poor material selection.

 

Speaking of clumsy robots:

 

I was going to post that actually, but it is still irrelevant to my point

 

That's like saying a robot that builds with lego is the size of a desk, or a 3d printer is bigger than what it prints. Think about it life maneuvers and builds with molecular strands of dna, 4 dna molecules at the molecular level working around the 2.5 nanometers scale (diameter).

What if you take that process and use it to build the structure or scaffolding that you then process to build a silicon or carbon based chips. You would wash off the underlying scaffolding in a secondary process.

I tend to base my opinion on what I know and logical assumptions as well as taking in the larger picture. Science and technology has branched into many fields but each field seems to have a narrower and narrower perspective. Maybe science should have an annual unthemed get together for a couple of weeks free holiday where they can mingle and share ideas. Kind of like a science Jam without the need to really make anything, but to share ideas and cross pollinate between fields.

 

I don't really understand anymore than when I first posted here, but that is certainly a lot of different numbers.

Getting back to the topic of living in VR, I plan to do that for most of my time starting around december of this year.

 

People go to incredible lengths to almost artificially recreate something that occurs commonly in nature, to almost understand the form or mechanics of something that an animal is born knowing how to do.

Our methods are still too crude.

Animals are self replicating, self repairing machines made of cells, controlled by quantum computers that contain programs that guide the machine in survival.

Humans need to respect biology more and abandon this silly quest to create an animal from metal and plastic.

Also, pathetic... Cheetahs move faster than that robot and they often do it on an empty stomach with things trying to kill them.

 

You'll be better off with nanomachines for that, not life.
It is not like a bacteria or protozoan will grow tiny hands and stitch new DNA molecule.

I that case I would advise to learn more and know more.

 

The current level of DARPA's ability to build robots that compete with humans or animals on physical ability, let alone mental ability, has nothing to do with where it might end up. I'm glad that all the people who attempted flight didn't just give up when the wings attached to their bicycles failed to take them to the heavens. It's utterly pointless and silly to look at these videos as if they're the full potential of robotics.

Not to mention that the current problems with these robots have a lot more to do with poor control systems than materials.

 

My main point is don't be misled by the technology singularity crowd.

There has been for about 40-50 years a boom in price performance in transistors. But that has only been possible because our economy and civilisation has valued the services they provide.

The simple laws of physics and quantum dynamics mandate that silicon transistor size can only be reduced to a finite size.

Therefore I don't expect there ever will be an affordable ($1000) 38+ petaflop AI silicon computer we can stick on our desktops. This is my opinion and until you can show me a cheap enough technology with that processing power and a working AI running on that hardware then I am factually correct and true.

So to disprove my point all you need is a working AI on a $1000 computer.

However already IBM and Google are working on AI systems on big supercomputer clusters. Will they create a human level or greater AI, quite probably.

They should be able to build a 38 petaflops system as the world leader at the moment is the Tianhe-2 at 33.8 petaflops and it cost about $380 million.

It uses the Intel Xeon E5-2692 (22nm) x 32,000 and the Xeon Phi 31S1P (22nm) x 48,000 chips.

But let's assume the Phi are the workers and the Xeon are just managers so in theory if you use 64,000 Phi you should break the 38 petaflop barrier. And let's say if we can halve their transistor size we quarter the number of chips needed.

22nm 64,000 (about $500,000,000)
11nm 16,000 (about $125,000,000)
6nm 4,000 (about $32,000,000) -- This could be possible
3nm 1,000 (about $8,000,000) -- I don't think we will see working physical transistors at this scale.
2nm 250 (about $2,000,000)
1nm 64 (about $500,000)
0.5nm 16 (about $125,000)
0.25nm 4 (about $32,000)-- single atoms of silicon as transistors???
0.125nm 1 (about $8,000) -- a single 38 petaflop machine when a chip using technology that uses half a silicon atom as a transistor!

Note that you might be able to get there by stacking layers of silicon so a chip is not just a flat pizza but a 3d cube of processing power. And this could massively speed up chips as it could greatly reduce the length of the electrical circuits.

Don't just look at the graphs do a bit of independent thinking and some basic maths and question things, especially ones that look too good to be true.

Life is nanotechnology, don't you see. Dna is an amazing information storage medium, biology a self replicating nanotechnology. Just because your pet goldfish doesn't have Intel inside stamped on it doesn't mean it's not an amazing bit of biological nanotechnology.

 

The keyword in that sentence is "silicon". We're not restricted to silicon though.

 

Edit: This is your ideology, and many share it... essentially, science as religion, forefathers of science as saints or dieties, offering the world magic and wonder and dreams. However, this ideology is absolutely crushed by philosophy, and ironically it is not even consistent with the spirit of science, which was never intended to give hope or create miracles (although more recently it is marketed as such) but to, rather coldly, discover real and present truths about our world without concern for what those truths would "mean" for humanity. Science is about discovering cause and effect, it's establishing the "how" and sometimes the "why" but not about reaching for the heavens or any such nonsense. In fact, emotion and poetic sentiment is by definition not useful in science. It's not sexy. It's not funny, or cool, or even all that interesting. It is just hard ass thankless work. And real science does not lead to Transformers: Jungle Force for crissake.

Irrelevant comparison. They didn't understand flight at all, once we started to understand aerodynamics flight became plausible. That said, the design of a propelled air plane has not evolved. It's essentially the same thing as it always has been. We still can't fly backwards, forwards and upside down and the common house fly can do that.

We should understand the dynamics of walking much better than flight. Yet, walking is infinitely more complex than flight, from an engineering standpoint. But, any bipedal or quadripedal machine is going to be inferior in mobility, speed, balance and coordination to any comparable animal. And they will be incomprehensibly stupidly expensive to build and repair, and they will be damaged easily. Animal tissue resists oxidization and solar radiation and humidity fairly well, through a wide variety of complex chemical processes, but any standalone machine will be subject to carbon buildup, oxidization, degrading rubber and plastic components, deformation of parts from impact and wear...

If you want to understand the truth, you must recognize that the denizens of this planet represent exactly what it takes to survive on this rock. It took billions of years to become what we are, and innumerable species didn't make the grade.

These robots will one day be looked upon like stick figure drawings scrawled on cave walls.

Bio-engineering is the future, but it's too unethical to have my approval. And it may result in something that is more fit than we are, which is a horrible idea.

Poor humans... always playing with fire. This is why Prometheus is eternally punished.

 

How is it not a relevant analogy? We also don't understand a lot about machines. As you pointed out, a lot of things to do with aerodynamics have yet to be understood even though our flying machines have advanced to a level unimaginable since the Wright brothers flight. In 100 years, the robots we can develop will be to those DARPA contraptions what an F-22 is to a bicycle with wings.

Pure opinion. Do you realize that 99% of the current problems in robotics have absolutely nothing to do with materials? Reliable image recognition (which I have some experience in) and decent control systems, to name a couple of things, are much more of a problem. Not materials. When we can make a robot that doesn't have to think for 10 minutes just to reach over and open a door, we'll be in much more of a position to compare things at the material level.

Wrong in so many ways. What do you think damages more easily, Darpa's Big Dog or a human body? And you're comparing something that has a complex and continuously functioning maintenance system (a biological system) to something for which one has not been developed yet. If your body's ongoing maintenance and repair stopped working, you'd decay much faster than a piece of metal.

And deformation of parts from wear and tear? Why don't you make a piston engine out of human bone and see how long it lasts? Not a valid comparison at all.

If we're to have a chance of becoming more than a one-planet wonder we'll have to be adaptable to a much more varied environments. Space, for example, is pretty bad for humans at the moment.

Today's robots will. Big difference.

Your whole post reflects the notion that we've already hit the wall with robotics, when it could not be further than the truth. Our robots are clumsy and fragile now, but so were our aircraft in the beginning.

EDIT: I didn't understand the edit to your post above, or what philosophy has to to with the science of materials.

 

I skimmed your post... and since I am a bearded old bastard, I can sense you have a lot of emotion and passion. And enthusiasm, and optimism and zeal. You pose like 7 different questions, some of which appear to be rhetorical recitations of things I said with question marks added. Half of the rest is obvious strawman stuff, baiting, and just generalized craziness.

I don't have as much invested in this as you do, obviously. I'm not looking for an endless debate. I said what I have to say, and unless you can respond with something I haven't considered before, it's not worth it for me. Answer all of your own questions that you posed to me and then you will find that you protest less.

 

If we could get back to discussing robots vs humans/animals, I think anyone reading would appreciate it.

I simply pointed out that today's robots are not the 'end all' of robotics. There's no reason to believe we haven't got a long way to go. Darpa's contraptions don't represent what robotics is capable of as a field.

A robotic system that had all the features of the human body - not just skeletal or muscular but down to the DNA and cellular level, would be something more worthy of comparison.