You have absolutely no idea what you are talking about.
No you don’t know of simple algorithms that will reproduce all of the signaling on a single neuron.
Go read up on this topic, your current understanding is about 30 years behind where neuroscience is today.
Habeed, to make sure it’s clear what we are debating, I pointed out that the described process for mind uploading was not capturing the state of synapses due to not capturing the epigenetic changes to the neuron’s DNA which are a key part of the current state of synapses (and that ultimately we may need to go to a much lower level).
You seem to be countering my point which amounts to you thinking a mind upload does not really need to capture the state of synapses.
Is this correct? Do you think we can ignore the state of synapses when trying to upload a mind?
This is the reality of the singularity. I think we’ll reach a point where the horizon is just a blur with no sense of shape or form or distance from us, but we won’t care much when it happens except for some cranks and CTs. It could all go horribly wrong, but it will be too late before we realize it.
They must have mistaken flying cars for jetpacks. We’ve only had those for half a century.
This is simply untrue. We, meaning all sources, produce nearly a 50% surplus yearly. What we lack is the political will to remove the impediments between the food source and the people who need it.
Raft, what I’m saying is that mRNAs floating about in a synapse represent the derivative of future state changes to a synapse. The visible proteins in the surfaces of the 2 adjacent cells are the **present **state. Your very statement points to some level of ignorance here.
This is indisputable, because mRNAs for proteins that have yet to be synthesized have absolutely no effect on immediate activity at this particular synapse. The way it responds to immediately adjacent (in time) action potentials is the same.
A mind upload would still need a method of modeling future state changes, but it need not do so by modeling the 3d position of every last piece of RNA, or the time it takes for a typical ribosome to synthesize the peptide, or how long folding is going to take…
All those steps can be represented by far simpler variables. And yes, the paper you mentioned just described assigning a 2-layer neural network (modern ‘neural networks’ are hugely simplified software constructs as you are aware) to predicting changes across a single neuron for good predictive results.
Maybe we should start another thread on the origin of the term… My understanding was that it’s called a singularity because there will be a point at which the computers are ahead of us. By the time you think to ask your computer to make you breakfast, it’s already being delivered because the computer anticipated your need and went ahead and did it. By the time you learn how X works, it’s irrelevant because the computer has already researched and implemented its own improvements so that it’s moved from Y and into Z.
In other words, it will look like a black hole because what goes on inside will be done so fast and so independently that we will simply be unable to understand it.I’m pretty sure that’s what people are getting at when they say there will be a singularity.
Now, I’m not saying I believe this will actually be the state of affairs; I tend to think your idea of a Horizon is much more how the future will work out.
The future state of that synapse is dependent on the epigenetic changes that have already happened to the neuron’s dna.
If you don’t capture that state when uploading (that synapse X needs to be maintained at state Y) then the future state of that synapse will not be correct.
My point was never that you must have a model that mimics DNA and epigenetics, my point was that important state information was lost using your described method. Furthermore, as neuroscientists learn more, it’s entirely possible the level of important state could be down at the atom level.
“Important state” means that if you don’t go down to that level, the thing you “uploaded” won’t functionally/computationally match the original.
A mind upload must do the following:
1 - Perfectly capture all state this is relevant
2 - Follow the exact same sequence of state
At what level this must happen is a gigantic question that nobody really knows.
One thing we do know for sure is that the method described in that paper misses important state.
Technically, I believe a 2 layer neural network can approximate any function, so what you say is not really wrong, but it’s not valuable information because you could say that about pretty much anything.
A single neuron is large network of signaling and computation that occurs at many locations on the neuron for many different reasons (e.g. image pre-processing, signal compression, etc.) all impacting current and future state. We do not have the understanding today to be able to simulate a single neuron with a neural network.
NAND can be represented as a 2-layer NN, so by chaining an arbitrary number of 2-layer NNs you could get anything a computer can do, at least.
Ah. I see where you differ. My assumptions are different.
A successful mind upload would not retain most skills or memories, but could be trained to perform any human task the original was capable of.
A near perfect mind upload would retain the majority of the declarative memories and the majority of the original personality traits of the copied person.
To me, uploading is a complete success if it allows for a transfer of sentience.
What you’re describing probably isn’t practical, it could easily be easier to just repair a preserved subject’s brain and use the original hardware instead. There’s a big difference between mimicking a processor’s official instruction set and actually duplicating every possible behavior and glitch, by analogy.
But the capabilities of a mind are completely dependent on what it has been trained to do via experience and memory.
If you don’t retain most skills or memories then you haven’t uploaded a mind at all.
This sounds like a contradiction of the above.
If you want to retain memories and personality, you need to capture all of the relevant state, which is back to the same thing about epigenetics because that is a key part of capturing that state.
But that analogy doesn’t hold when it comes to our brains. For our brains, the instruction set is chemistry and physics.
To get back to the OP. I firmly believe that we will see humans land on Mars within 50 years. It will be China that lands there first, probably around 2035.
China wants to take their place in the 21st century as a true undoubted superpower, being the first on Mars would cement that. They also have the ability to use Nuclear rockets, without having to worry about political opposition. If you bring a NERVA style nuclear rocket into the equation, all of a sudden a return mission to Mars gets a lot easier.
NERVA doesn’t solve the problem of landing a spacecraft capable of reaching orbit from Mars surface with astronauts onboard. It takes a significant amount of dV to do this, and it isn’t practical to try to land a fully fueled rocket onto mars. (this is because a rocket with enough fuel to reach orbit again would be extremely heavy, and would require a gigantic aeroshell, and would have a very high terminal velocity after the parachute phase. So you need even more rocket fuel to perform the soft landing onto Mars, which means this lander needs to be even bigger and heavier still, which means…)
The proposed solution to this is to land the ascent stage with empty tanks, land a big tank of liquid hydrogen, actively cooled the entire transit, and generate the methane and O2 from the CO2 in Mar’s atmosphere. (don’t forget to bring a nuclear reactor to power all this)
This is a whole set of problems of it’s own : you want this equipment in place before the first astronaut takes the descent vehicle down, so the equipment has to be teleoperated, and so on.
Anyways, NERVA just makes it easier to do the Mars injection burn and the earth return burn. This is totally doable with conventional rockets, however.
Sure, I’m aware that any Mars mission is going to involve many unmanned landings first, even with NERVA. As I understand however, NERVA would allow each launch to carry considerably more weight to Mars and allow shorter transit times. Both of which would make a manned Mars mission considerably more feasible.
I may have been a little optimistic in the timing, according this article “China’s last announcements set a date of between 2040 and 2060” for a manned Mars mission.