Ancestors

Toot

Written by ch1n3du on 2024-08-11 at 08:38

Found a cool article on building tiny self replicating factories with today's technology https://www.lesswrong.com/posts/rQxrSRPJGmksGjK8z/it-s-time-for-a-self-reproducing-machine

=> More informations about this toot | More toots from ch1n3du@hachyderm.io

Descendants

Written by fumnanya on 2024-08-11 at 09:59

@ch1n3du hmm given that you’re probably going to need to supply a lotttt of prefab parts, is a box that can assemble itself from those parts all that interesting?

=> More informations about this toot | More toots from fumnanya@hachyderm.io

Written by ch1n3du on 2024-08-11 at 13:15

@fumnanya The "AutoFac" described in the article is designed to manufacture and assemble most of it's components (excluding computer chips).

The main fact that it can replicate potentially exponentially

and manufacture goods autonomously would drop the price of goods that

it can manufacture to the cost of the raw materials, which is pretty close

to post-scarcity of the flavor in @pluralistic 's Walkaway which I'm currently reading.

=> More informations about this toot | More toots from ch1n3du@hachyderm.io

Written by ch1n3du on 2024-08-11 at 13:15

The author mainly envisions it being used for mass manufacturing in

one central location but I think the best approach for this sort of

tech would be small-scale, lean manufacturing like that described in @KevinCarson1's

"The Homebrew Industrial Revolution" (HIR for short).

=> More informations about this toot | More toots from ch1n3du@hachyderm.io

Written by ch1n3du on 2024-08-11 at 13:15

In HIR, Carson argues that small-scale manufacturing has been more efficient and robust than the current mass-manufacturing approach for a long while now but is actively held back vested interests, if his hunch is correct then the creation of a truly useful self-replicating system would be the final nail in the coffin.

The large fixed-costs of storing and transporting stock are unjustifiable

when factories can dynamically scale to match demand (self-replication) and are general-purpose.

=> More informations about this toot | More toots from ch1n3du@hachyderm.io

Written by ch1n3du on 2024-08-11 at 13:15

Costs aside having this sort of cheap, general-purpose manufacturing capacity would probably lead to an explosion of open-source hardware innovation, as the rate of iteration of hardware would approach that of software. The current software doesn't seem anywhere near ready for this, the best thinking I've seen on this is Nadya Peek's "Making Machines that Make":

=> More informations about this toot | More toots from ch1n3du@hachyderm.io

Written by ch1n3du on 2024-08-11 at 13:15

On the feasibility side of things, I think that the plan in the article underestimates how quickly not being able to create computer chips would be a problem, I don't blame the author, fabricating high-end chips is mind-bendingly hard but they're a lot of ways to workaround it.

Most manufacturing tools use simple micro-controllers and those don't need the compactness of current chips, early systems could use older generation fabrication techniques...

=> More informations about this toot | More toots from ch1n3du@hachyderm.io

Written by ch1n3du on 2024-08-11 at 13:15

the costs savings from that could be huge as the cost of chip fabrication has been increasing exponentially alongside Moore's law (https://www.construction-physics.com/p/how-to-build-a-20-billion-semiconductor).

=> View attached media

=> More informations about this toot | More toots from ch1n3du@hachyderm.io

Written by ch1n3du on 2024-08-11 at 13:15

Hobbyists have had some success doing this at very low cost:

The hobbyist in the first video founded a startup that's trying to build way cheaper fabs using electron beam lithography: https://atomicsemi.com/ and he has claimed they've been able to get down to aroundd 190nm resulution.

=> More informations about this toot | More toots from ch1n3du@hachyderm.io

Written by ch1n3du on 2024-08-11 at 13:15

Also I think the author is too tied to the notion of a single machine replicating an exact copy of itself but it might be more effective to have an evolutionary approach where AutoFacs can build other AutoFacs of increasing complexity and having a bit more specialization. I first encountered the idea of having an evolutionary replicative approach in this NASA paper: https://arxiv.org/abs/1612.03238. and it makes sense when you think about it because that's how life evolved.

=> More informations about this toot | More toots from ch1n3du@hachyderm.io

Written by ch1n3du on 2024-08-11 at 13:15

For more links on interesting stuff related to this I'd recommend:

Stuff from:

=> More informations about this toot | More toots from ch1n3du@hachyderm.io

Written by ch1n3du on 2024-08-11 at 15:46

@beka_valentine Talk on the idea of "digital hardware" https://youtu.be/EA-wcFtUBE4?si=ts9SP4GZj8Jtwwxn

=> More informations about this toot | More toots from ch1n3du@hachyderm.io

Written by noarn chornsky on 2024-08-11 at 17:13

@ch1n3du im not sure if you meant to @ me on this one :)

=> More informations about this toot | More toots from beka_valentine@kolektiva.social

Written by ch1n3du on 2024-08-11 at 17:14

@beka_valentine sorry 😅

=> More informations about this toot | More toots from ch1n3du@hachyderm.io

Proxy Information
Original URL
gemini://mastogem.picasoft.net/thread/112942482908549919
Status Code
Success (20)
Meta
text/gemini
Capsule Response Time
320.560094 milliseconds
Gemini-to-HTML Time
3.255192 milliseconds

This content has been proxied by September (3851b).