“Working synthetic intelligence” : a sentient being that uses only artificial components. (a brain in a life support jar does not count)
“sentient” : capable of performing all cognitive tasks a human is capable of performing.
The being might be a massive collection of artificial neural networks developed using the new methods that google and other giant corporations have recently started to get real results from. Or, it might be an emulation of a once living human brain. (this is why I’m using the word “synthetic” : a human emulation isn’t really what we mean by AI usually, but the hardware emulating a human is made in a factory)
Or, we might make a detailed synaptome mapping of a human brain, then design artificial neural network modules to duplicate each major feature, a few systems at a time, and eventually get a being that is sort of a hybrid and also sentient.
Cost to do this : probably trillions of dollars.
Benefits from doing this : whoever gets this working would probably ultimately get immense returns on their investment, since they could put the synthetic beings to work performing trillions of dollars of labor, or have them design and manufacture weapons and use those weapons to steal the resources belonging to everyone else on the planet. Either way, they get their money back.
Large Scale manned space travel : sending more than 1000 people outside of Earth’s orbit with the destination being another astronomical body (the Moon counts).
Cost to do this : trillions of dollars
Benefits : some humans get radiation exposure, bone loss from microgravity, and get to explore a hostile and merciless environment from inside the safety of pressure suits and buried capsules. They most likely find nothing but rocks, vacuum, and radiation. No return on investment.
And, as a strong-AI proponent and afficionado of the motif in both real computer science and speculative fiction, I favor AI.
But large-scale space travel is possible today, and effective AI ain’t. So space travel is, to that degree, more likely. Plus, it’d be nifty, and would have all sorts of side-stream benefits for earth-side tech, industry, manufacturing, and abstract research. So I want it to happen too!
Large scale space travel is only “possible” in that if you paid an army of engineers to design and prototype much larger rockets, and landers, and habitats, and all kinds of long life life support equipment…after developing reams of new equipment, you could do it.
It would be like knowing how, in theory, to build a wide body jetliner, but never having built one, and expecting to get one built within 20 years or so and flying passengers.
And then scale it up 10 times.
The thing is, emulating an entire brain…we could pay a large array of scientists and engineers to measure the exact timing and other properties of the various synapse types in a human brain. Current estimates say there are a few hundred permutations.
We could pay a bunch of chip designers to design custom ASIC parts intended to emulate sections of a digital brain. We’d pay other chip designers to design the high speed network switches to interface these parts.
Unlike the rocketry example above, no one will die if a small error is made. It just won’t work right away and some money is wasted. So it’s at least 10 times, probably more like 1000 times easier to get stuff built. There’s much less need for bureaucracy - either you are modeling a piece of a human brain, or you aren’t, and you don’t have to have the reams of paperwork justifying every decision and committee meetings to get people to sign off on risks.
When stuff goes wrong, nothing explodes (small electrical fire at worst but that is unlikely), you don’t have to wait for months of travel time between planets, no one’s lives are directly at stake…you just connect up some diagnostic equipment and figure out what the problem is. Skunkwork together new diagnostics equipment if you can’t figure it out.
I think the scale of the effort would be similar. We actually could build a computer capable of modeling an entire human brain today. It would “only” take about 5-10% of the total global output of refined silicon to do it.
The reason I bring this up is that if this is really true - working on large scale space travel is a waste of time. Humans will never do it. Once you build a prototype synthetic intelligence that works, you can begin working on shrinking it. We know you can fit a sentient brain into the size of a basketball and it won’t need more than 10 watts of power. (that’s how big the brain is now)
If you could make something that weighed 2000 kilograms and needed a kilowatt of power to run all the time, but was sentient, space travel would be a cinch. It doesn’t matter how long it takes to get somewhere - your sentient explorer can wait. It doesn’t matter how dangerous it is - your sentient explorer can upload the mind state changes of itself/himself via radio as a telemetry stream, so even if the hardware fails or is destroyed, he/it will be recoverable.
You can harden synthetic chips against radiation a lot easier than you can shield humans. It’s very easy to make your logic trinary redundant with majority gates, and you don’t have to worry about long term damage or cancer - you’ll just replace the hardware when the happens.
But since we have (in this metaphor) already built and flown DC3 aircraft, we know it can be done. We’ve sent people to the Moon, and established long-term space stations. There aren’t any significant scientific discoveries that need to be made.
Actually, no, we couldn’t. We don’t have the controls to operate such a sim, and we don’t have the modeling of individual neurons yet. There’s a lot about the brain that we do not know. Fundamental scientific discoveries would have to be made first.
Now, sure, we could just go with “stochastic” modeling – not emulating neurons, but “heuristic” programming. (I’m not actually sure if either of those words means anything concrete here, but they’re fun to use.) i.e., we could “fake it” – just do a super-duper “Eliza” program. Watson with bells and whistles.
I have to disagree with that.
Your arguments in favor of a robotic space program are, of course, familiar ones, and I fully agree with them. I like the idea of our current approach, sending out probes to the distant planets, where we cannot send people. We don’t even need sentient/sapient probes: we’ve accomplished miracles with robots at our current level of programming.
I think that the “problem” will solve itself. I think that within the next 500 -1000 years we will have the capability to download a human brain into an effectively immortal synthetic brain/body. At that point you can explore the universe at will without worrying about all the frailties and necessities of the human body.
Beyond this unless some form of FTL is discovered extensive space exploration is going to be a very, very long term exercise. The cost and lack of compelling need too leave the earth are the main things that will hold interstellar efforts to a minimum. It’s romantic but the necessity to do lots of manned outer space stuff is pretty minimal.