[simne]

> this would take us about 31 years[2] to make such a swarm

They assume, when have already working general AI technology and it have some limited size (volume-mass-energy consumption).

Unfortunately, we still not have GAI and even cannot predict, how large will be first practical unit.

Must admit, looks like we very close to do it, but from history of previous great technical inventions, some things takes decades to achieve production status and was repeatedly reinvented in some years after another inventor fail.

[ben_w]

General AI is unnecessary. Bacteria do not possess this trait, and yet reproduce themselves, some in as little as 30 minutes.

It is also possible to have a large system where humans are just a component, if this were necessary. The human-machine ratio is a function of how close the automation you have is to what you need.

[simne]

> Bacteria do not possess this trait, and yet reproduce themselves

If you programmer, you should know from experience or from learn, that in complex system possible just two ways to achieve reliable execution.

1. Brute force, just test as many possible scenarios as could, 99.999% is better than 99.99%, and make script for each scenario.

2. Smart, run system when tested somewhere between 70..90% and make some sort of insurance, so when happen non-tested scenario and all failing, you will pay (compensate) for harm, and make additions.

That is. Bacteria lives in comfortable environment (mostly in liquid water drop), and spent billions of slightly modified reproductions, to make solutions for all possible scenarios. You may hear, DNA of simplest bacteria are more than Million pairs, that's because of number of scenarios it successfully survive.

Space is much less comfortable environment than liquid water, it have wide range of possible parameters, I even not sure if exists some structure, which could survive in all possible space environments, so need some adaptation mechanisms, to change structure, and best is consciousness AI, which could make smart predictions of causes and reasons, and control all these machinery. And also it will have memory, to repeat moves which helps to survive when something similar happens earlier.

[ben_w]

> If you programmer, you should know from experience or from learn, that in complex system possible just two ways to achieve reliable execution.

Irrelevant. A self-replicating system does not need to be highly reliable. Look to the past, any time over 200 years ago most families were a dozen kids because most didn't reach adulthood.

> That is. Bacteria lives in comfortable environment (mostly in liquid water drop), and spent billions of slightly modified reproductions, to make solutions for all possible scenarios. You may hear, DNA of simplest bacteria are more than Million pairs, that's because of number of scenarios it successfully survive.

False. Bacterial environments are hostile because other bacteria fight them for the same resources, including predation. Many chemicals are hazardous even in small quantities. Internal chemistry requires water in liquid form, yet there's only a narrow range of temperatures where water is liquid, and worse the chemical processes change rate significantly even within that range.

Also irrelevant, we've been using simulated evolution as a form of AI for ages already. It's not new or novel. I implemented a version of this in 30 minutes over a decade ago just to prove a point. A million bases is trivial to store, so is a billion or a trillion.

> Space is much less comfortable environment than liquid water, it have wide range of possible parameters, I even not sure if exists some structure, which could survive in all possible space environments, so need some adaptation mechanisms, to change structure, and best is consciousness AI, which could make smart predictions of causes and reasons, and control all these machinery. And also it will have memory, to repeat moves which helps to survive when something similar happens earlier.

Also false.

1. Space has far fewer parameters than water.

2. One does not need to make a single machine to survive "all possible space environments" to do this, just our solar system at 0.47-0.31 AU from the sun. We already have that, we sent probes there.

3. Consciousness is not necessary for any of that. Neither is episodic memory (though that is trivial to implement). Bacteria exist and do these things well enough with mere DNA.

[simne]

> Look to the past, any time over 200 years ago most families were a dozen kids because most didn't reach adulthood.

Do you know mathematics? Calculate, how slow will become your Dyson swarm, if for example only 1/20 will survive?

BTW, you may hear about baby-boomers, and they are exactly caused by much improved medicine, now in EU survive near 100% children.

Calculated? Ok, now calculate, how much suffer probability of overall success, because limited resources does not accept to make 20 turns to achieve 1 successful?

> 2. One does not need to make a single machine to survive "all possible space environments" to do this, just our solar system at 0.47-0.31 AU from the sun. We already have that, we sent probes there.

Well, now I see you are just overweening human, but without real knowledge. Solar system is itself have wide parameters spectrum, but is is also significantly different from other stars environments.

> Bacteria exist

Bacteria have sacrificed billions lives, to gather information, to achieve current success rate.

But must admit, I will consider idea you suggest me, about send hopeless missions, to just gather info, and I'm sure you also lazy, so I'll myself calculate success rate for each sacrifice rate.

[ben_w]

> Do you know mathematics? Calculate, how slow will become your Dyson swarm, if for example only 1/20 will survive?

It means the real reproduction time is t/f, where t is the time it takes to make a single unit and f is the fraction of units which survive to further reproduction. For 1 in 20 surviving, that means the real reproduction time is 20t.

Some bacteria take 30 minutes for a single reproduction, so that with a 1/20 success rate would be an effective population doubling every 10 hours. An E. coli cell weighs 1 pg, and this is only a factor of 2^128 from the planet Mercury. These random example numbers would therefore be able to consume the entire planet in 53.32 days. At this level, almost all the time (97%) is spent on waiting for the solar panels to supply enough to get the stuff from the planet's surface to solar orbit.

> Calculated? Ok, now calculate, how much suffer probability of overall success, because limited resources does not accept to make 20 turns to achieve 1 successful?

I have no idea what point you're even trying to make here.

We know we don't need to worry about your 1/20 random example for humans because we know ourselves; only the machines need this consideration. That's a number which you made up, and your own complete fiction is what you're now trying to use for an example that I don't understand.

> Well, now I see you are just overweening human, but without real knowledge. Solar system is itself have wide parameters spectrum, but is is also significantly different from other stars environments.

Completely irrelevant. I don't even know what point you think you're making. I linked you to a specific plan to build a Dyson swarm specifically in our solar system at the orbit of Mercury. The rest of the universe is irrelevant to this part of the plan, for exactly the same reason and in exactly the same way that it is irrelevant to bacteria on Earth that the rest of the universe exists.

What you do with your Dyson swarm (including colonising the universe) only matters after you've built your Dyson swarm. Building one is fast the moment von Neumann machines can be engineered rather than grown, and give you such an incomprehensibly large industrial and resource base to work from that comparing it to what we have access to today is more extreme than asking a single pre-writing cave painter to imagine our current entire world.

> Bacteria have sacrificed billions lives, to gather information, to achieve current success rate.

So?