Neuralink

[D3K]

Musk and the guys developing neuralink released their first demo today (scroll to 49mins for the start)

• Read / write capable
• Demo of pig showed neural spikes from it's snout while sniffing / eating
• Demo of pig on treadmill showed limb activity overlaid on body graphic: signal vs actual
• Size of a penny
• Implanted on top of skull - penny size chunk of skull removed, wires fed into the brain
• Automated robotic implant procedure
• Can be done without anaesthetic
• 1024 channels which works out to be around 100kb/s (?)
• Compression involved
• Upgradable in future
• Can use multiple
• May solve anxiety, pain, depression, etc
• Intend to re-able patients with disabilities - spinal, sensory
• Can augment vision if camera is attached - full EM spectrum could be possible
• ...
• It can play Crysis (eventually)

Like Tesla, there is an ulterior motive to just business / environmentalism / philanthropy - Musk sees the Singularity as one of our greatest threats and wants to get ahead of the curve by making sure the machines have a reason not to kill us / give us value. To reach that goal we will see neuralink boundaries being pushed well beyond just solving ailments.

I see this as forced evolution. And I think I like it.

 

[D3K]

Wonder how politicians will react in terms of trying to regulate this stuff - initial reactions if/when it matures could be quite fearful/reactionary.. only wanting it to say restore vision or to assist with disabilities.... not for some wealthy person to enhance etc.. (though that could well still be a long way off).

Anyone without inhibition (moral, spiritual, religious, etc) and who isn't put off by the penny-sized chunk of skull that must be removed, is going to want this. And if it gets relegated to medical treatment only, there will be tourism to countries that openly allow it.

that I'm interested in this as yet another "great filter" avoidance measure. AI has already changed our world, and as things progress things are likely to get whacky really, really fast.

I am not convinced this will save us from the negative potential of the singularity, but I also cannot see any other feasible solution to living with it. If you can't beat them, join them. But regardless of our integration with it, any AI can still look at us as inefficient leaf nodes on a system they are trying to optimise, and off us anyway.

 

[D3K]

Hundreds of years away that is, probably.

A while ago there was a world-leading neurosurgeon on BBC Hard Talk. He was talking candidly about how much they don't know about the human brain. We've barely scratched the surface.

The post-demo round-table discussion mentioned this. By the very fact they have this device which has 10x the bandwidth of the nearest competitor and is much easier to implant means they will be pulling in more data from more people. Learning will accelerate and they have teams eager to work with it.

 

[D3K]

Having read the wiki on them, cochlear implants are not brain implants. They have electrodes which are placed into the cochlear which stimulate the hearing nerve.

Amazing tech especially for 1957 when they were invented, but not comparable to what neuralink is which has the prospect of opening all doors rather than just one.

 

[D3K]

Maybe in ST transporters! Universal laws allow wormholes to exist, so what if a transporter was created that could open a wormhole from A to B and that is simply how they worked? No suicide then


Necessary, not forced. It cannot be a forced thing if it's our only option for the future of our species. There are many things humankind will never ever accomplish without augmenting with technology. The AI/technological singularity is going to be around the mid 2030s, the physical machine singularity is more distant in the future, and that's the era Musk wants to be ahead of the curve in. If future AI sees that humans are already augmented then it would (could?) logically deduce that actually those meat bags aren't just bags of meat and are capable of useful /stuff/.and be less likely to wipe us out as an inferior species who are simply using up the planet's valuable resources.

Or something like that.

Forced as in we are encouraging it, not letting it happen naturally...

But to entertain your concept - only a small few will be in line with that thinking. "upgrade to save humanity". The rest will be doing it for health, kicks, fun. And many more will reject it. For those who reject it, we won't be forcing them. However it will likely be legislated to be implanted at birth, much like the youtube short i watched recently which I need to try and find.

As in AI being so realistic that a human being talking to one over the phone or internet won't be able to tell they are speaking to a virtual machine?

That's the Turing test. The singularity is the point where AI becomes more intelligent than humanity.

 

[D3K]

"The Singularity" has a very specific meaning. Ever seen the movie Transcendence? It's basically that.

No that's a human being uploaded to a machine and having access to that power and speed. It is not artificial intelligence.

 

[D3K]

Yes, fair enough. But that movie shows the theory of what would happen post the "singularity."

Ie, exponential advancement.

You're right tho, it's a slightly different premise.

The reason the "singularity" is invoked is to imagine a world where advancement is exponential because AI has become superior to human intelligence, and also self-aware, possessing genuine intelligence (and not faking it).

The imagination part is hard. Who knows which way it could go. Does it automate, replicate, optimise, build. Does it appreciate us, hate us, see us as inferior, etc, etc. There is no way to tell what happens after that point. I've heard both Musk and Penrose chime in on it.
If you've watched Devs recently it's akin to them not being able to see the future after a certain point.

I quite like the Roko's Basilisk thought experiment. It seems to really unsettle some people.

TLDW

Spoiler

AI has to ensure it's survival so wants to know which humans support it's existence. It runs simulations on all humanity past and present with absolute accuracy and effectively knows everything that's ever happened - including whether you supported it's existence or not. So it kills those who did not. But then the plot thickens that if you have this knowledge in advance of it's existence, and you do nothing to support it / bring it into existence, it deems that as non-compliant and will end you as well. So in the here and now, if you do not act to bring this AI into existence, you are dooming yourself when it comes to be

 

[D3K]

100% accurate simulation of dead people?

That's like CSI zooming in on a number plate reflected on a door knob in some grainy CCTV footage from a shop five miles away

But since I don't believe in the singularity I couldn't give a monkey's

Eh? It simulates the universe from start to end. Or whatever portions it needs. So whatever bogey you picked out of your nose last Wednesday morning, it will simulate that happening in it's own mind. Dead or alive right now is irrelevant. Prolly better watching the video

 

[D3K]

You can't simulate something if you're missing the data.

Without time travel you can't simulate dead people.

How can you simulate the universe - [OUR] universe not [A] universe - without being around at the beginning (or knowing *exactly* the state of all matter and energy at the time OUR universe started)? It doesn't make sense, I'm afraid.

It's the same as how current simulations work in the engineering world today. You load up some boundary conditions, press solve, and the machine spits out stress/strain, fluid motion/thermal capacity, SI/PI whatever the study was for.

So you need the equations, constants, and variables from the start of the big bang, press solve, and the simulation works everything else out. With an accurate enough solver model, it will process the exact quantum effects at every moment until it reaches the current day.

If you want to account for the multiple worlds theory which means an individual simulation for every possible quantum choice then sure - it will take a lot longer but it would still be possible. The bulk of the simulations would be discarded when they veered away from plausibly becoming the representation of the universe we currently live in, so the AI would not have to juggle infinite simulations at the same time.

On the other hand, if we as humans have carefully recorded our universe as it is, with whatever degree of accuracy needed, then the AI can start from that point instead of the big bang. It might not even have to be accurate, it may be able to interpolate data very accurately (much like RTX does with ray tracing).

So given sufficient AI intelligence, power, and time, an AI could simulate our exact existence. And that then proves another theory - simulation theory. If our AI can simulate atom for atom, then the chances that we ourselves are living in a simulation becomes extremely high.

 

[D3K]

An engineering simulation simulates [a] bridge, because the theory of bridge design is the same for all bridges. No matter where or when they were built. You simulate [a] bridge not [the] bridge that collapsed 100 years ago for reasons nobody documented at the time.

Ask your simulation to tell you why that bridge 100 years in the past failed, but without having any data from that particular bridge. You don't know what the weather was like on that day, what the ground conditions were like, etc. Your simulation says... "insufficient data!"

You are claiming that an AI can simulate 100% accurately the mindset and thoughts of long dead people. Essentially with no data.

I am straight up saying 100% this is not possible. It's beyond ludicrous.

If this is a thought experiment it is a very poor one indeed.

Translation: "I can't imagine it therefore it's impossible"
Not sure why you're clinging on to your point about dead people, aside from you not getting this concept at all. Dead people are irrelevant to the idea.

All of the concepts in my previous post (except roko's basilisk) exist as plausable scientific ideas that have a non-zero chance of coming into existence. These have had commentary by leading scientific scholars to that effect. Non-zero contradicts your "100% not possible" idea.

Sure there are many predecessors required to make it happen : AI advancement, resources to construct the hardware for an AI of that magnitude, the power available to fuel such an entity, all knowledge of the big bang boundary conditions, etc.

Worth enlightening yourself before making such strong claims as ludicrous or impossible. I at least expected some follow up questions before you defaulted to that stance. They are hard concepts to grasp, but fun to toy with. And there is a very small chance they will ever come close to affecting us in our lifetimes. The power the AI needs may have to come from an entire solar system, or multiple solar systems, and that means climbing up the kardashev scale.

Grab some youtube videos on topics such as singularity and simulation theory and see where it gets you. Then come back to Roko's Basilisk vid and watch it with a bit more context. I do not believe any of it will happen. But I acknowledge it is feasible given sufficient advancement.

 

[D3K]

It's not hard to grasp. It's fiction.

The only way you can rewind time is to know with 100% certainty the state of all energy and matter in the entire universe.

If you can possibly know this - and think for a second how it would even be possible to know this - then in theory you can both rewind time and also know the future.

You need - quite literally - to know the state of all energy and matter in existence.

You cannot read the minds of the dead based on incomplete data. I should have thought this was self-evident, but apparently the laws of physics don't apply to AI McGuffins.

Also, lols at YouTube being a source of anything other than nutjob CT theories. No thanks, I've had better education on the toilet.

Ok please stop talking about dead people, I don't know what your obsession is with that point or why you think its relevent but you're going round in circles.

Let me repeat the part about simulating from the big bang:

If you simulate the universe from the start of it's existence, you do not need to know everything. You only need to know the input variables. Before the big bang there was nothing (TBC). Then simulation happens and all the events take place to recreate our universe up to the present day. And beyond if necessary. This was well presented in the TV show Devs recently.

 

[D3K]

Pretty much, like there are infinite possible universes... (or perhaps there are a very large number that we might as well treat as infinite, given there is perhpas a finite amount of matter, energy...) I'm not sure it would be feasible to simulate even one when existing within it... let alone simulate infinite (or many) possible ones - how would it know which one the current one is?

What information do you have from the future that impacts your judgement on feasibility? You don't have that info. And actually it's a conditional requirement so no one needs to solve this condition - assume it will be possible then work down the chain of other conditionals.
There are actually only 3 - scroll halfway down the page https://en.wikipedia.org/wiki/Simulation_hypothesis

I certainly don't think that the ability to simulate the universe is a requirement of any singularity nor do I know if such a thing is possible let alone being able to simulate many many universes and search among them.

The basilisk thought-experiment is one potential result of the singularity, it's not a requirement. And it's just for fun, but causes some people existential dread apparently.

 

[D3K]

How can you simulate THE universe - OUR universe - from the start of its existence? Where are you getting the data from?

Don't forget you've said this simulation is 100% accurate.

You know what 100% accurate means, don't you? Not a single thing can differ from what actually happened. Every molecule must be accounted for and correctly positioned. Every sub-atomic particle. Every energy wave. Every interaction as it actually occurred.

Not [a] universe. [Our] universe.

How am I supposed to give specific information about the future?
Humans are unlikely to provide the means themselves, it will likely be assisted by AI. Then the AI will go away and start running simulations.

Start of the universe is easier to work with because there was much less information. You just need all theories of relativity and quantum mechanics to be resolved then you have a set of equations that work everything out for you, provided you know all of the constants and variables needed to put into the equations.

 

[D3K]

To simulate our universe quicker than reality, would require a computational resource greater than all the energy and matter stored in the universe.

Please show your working for this claim

 

[D3K]

It's not a condition for a singularity at all, it's a thought experiment....

That's twice in a row you've confused the singularity with the basilisk thought-experiment. For the rest of what you've said - the multiverse theory, multiple worlds. Not sure how you can talk with such conviction about that being definitively how everything works and then dismiss simulating existence, when it has not and cannot be proven, if ever. Fwiw I love the idea, I think it's lower chance and more in the realm of fantasy than reality, but it's a non-zero chance like everything we're discussing.

It's extremely logical. As said, the universe is the most efficient simulation of itself possible. Whilst simultaneously being the most accurate.

Ok so you are more logical than Steven Hawking and Neil Degress Tyson. You've worked this out in your head, now put it to paper and send it round the community. Show them zany scientists!

Explain how you can more quickly simulate every single bit of matter and energy in the universe using a tiny fraction of that matter and energy? And all the interactions between them.

I don't have to explain anything hypothetical, we are far from understanding everything about how the universe works and advancements will be made that shatter preconceptions and introduce radically different ways of interacting with the universe, but to throw 2 half-baked ideas out there:
1) approximations CAN be made and get the same results: for example quantum effects can be disregarded for a large majority of existence and only macro effects need to be considered (the AI can work these out in better detail)
2) interaction in higher dimensions vastly simplifies the compute power needed to resolve our 3D existence, and so take place in the 4th/5th/6th dimension.
This is just to show I have some agility in my brain and not really up for discussion but I look forward to your ridicule in the morning anyway.

 

[D3K]

There are actually only 3 [conditionals] - scroll halfway down the page https://en.wikipedia.org/wiki/Simulation_hypothesis

I certainly don't think that the ability to simulate the universe is a requirement of any singularity

It's not a condition for a singularity at all, it's a thought experiment....

Nope, not at all, why do you believe I've done that? The existence of a singularity doesn't imply that that thought experiement is feasible or even possible - I thought that was pretty clear? Why are you acting as thought I've confused the two things?

You are confusing the singularity with simulation ability as a pre-req which was never mentioned. Being able to simulate reality is a pre-req in the basilisk thought experiment. I did not state any conditions for the singularity, yet you are correcting me on something not being a condition for it...

I don't - it doesn't need to be true, you're conflating the possibility of multiple universes with simulating multiple universes and then, within that massive search space, finding this particular one and the many many possible variations of it... Those are two different things - the multiverse theory being true isn't a necessary condition for what I've said - just the fact that there are non deterministic events in our universe means that any given simulation starting from the big bang has many many possibly universes to simulate it isn't a requirement that all those universes actually exist - do you follow?

How would I not follow when it was my post that presented this idea? Tbh I have over zealously added that in as a requirement when it was never mentioned in the original video. Yet I've thought about the concept a little too much and gone for the jugular with what's required in terms of complexity.
Either we find a way to simulate existence by brute force (because it can be done so fast) until we find a carbon copy of us just now, or shortcuts with convergence can be done - run a single simulation with lower fidelity through to the present day, then rerun from a point in history where the earth can largely be considered a closed system away from external influences of the universe (and thus reducing the complexity of the sim by [universe - 1 solar system] and keep iterating, increasing fidelity where needed. This is called convergence in modern day physics simulations and gives very accurate results in the localised areas of interest.

Or you know, we advance in technological capability and find new better ways to do things we previously thought impossible. Or rather that some of us thought impossible while others acknowledged the non-zero probability of it.

But i don't have that future knowledge so I can't state it's not feasible.

 

[D3K]

Sorry I no longer have any idea what you are talking about

e: Why don't you also need to simulate the 5th and 6th dimensions, whatever they are? I'm assuming they exist/are part of the universe which we are simulating? So should be part of the simulation, ie also simulated?

Or are these dimensions outside of the universe? But if they are able to affect the universe, surely they should still be part of the simulation? If the AI can use them (to simulate the universe) then they should be part of the data being simulated..

In which case we are still back to the original problem: how do you simulate reality faster than reality occurs, using a fraction of the universe's energy and matter to run the simulation?

Sad that you chose to pick this apart when I stated it was half-baked. But to give some idea where it comes from, look into the theory of relativity where you can see there are between 12-14 dimensions required to resolve everything. Whether these are theoretical dimensions or they actually exist is obviously unknown. But the higher you go the more concise your equations are in terms of deriving the rules for the lower dimensions and you would certainly need those for putting into a universe simulator - maybe not all, I simply do not know. So anyway, perhaps there are rules at the higher dimensions that can approximate the results in lower dimensions without having to explicitly solve and expand upon them. Or perhaps there are variables which allows to increase the speed variable of a simulating universe.

Go back and check how things advance each 100 years before dismissing things.

Case in point. Not only dismissing the impending future, but potential future as well.

You don't want to go sticking wires in and facing the possibility of causing brain damage or worse.

Maybe watch the demo.

 

[D3K]

and you would accept the result, and if so why?

Trust in the solver, previous results, convergence. For product design sims the simulation is not the be all and end all - you still need real life prototypes to test on and correlate results. But the point of eng sim is to reduce the number of physical prototypes which saves both time and money.

In the above examples we already have the end result and want to match the simulation to what already exists, as opposed to creating something new. An example of that in today's world would be studying how a real object failed, e.g from catastrophic fracture. Simulating the conditions until you got the same result as you saw in the physical version. Not as common as designing new products.

 

[D3K]

To be clear - if a singularity event comes about it is not conditional on this thought experiment being possible.

Why do you keep stating this? Which part of any posts I made stated there are any conditions for the singularity?

 

[D3K]

Because you keep on saying I'm conflating the two and I've been quite clear that I'm not! Why keep on saying I'm doing something I'm clearly not?

Conflate has a different meaning to confuse, which is the word I used. The quote below is where it started from. Can you explain your context there? Because no one stated that a singularity must simulate the universe.

I certainly don't think that the ability to simulate the universe is a requirement of any singularity nor do I know if such a thing is possible let alone being able to simulate many many universes and search among them.

Then similarly you posted the following, immediately after I listed conditionals required for simulation theory

It's not a condition for a singularity at all, it's a thought experiment....

It's fine to admit you got these wrong but not to project it back at me as losing the direction of the thread.
However, I admit that your post about multiverses was you explaining why you thought simulating all quantum effects was not possible, rather than supporting multiverse theory. I read fast, it was late.