Thought Experiment

[ernestocastillo]

There is no right or wrong answer to this, but the discussion tends to be very insightful (as it was on a recent retreat with some of my co-workers). As this question invites more educated opinions than concrete answers, I have decided to post here rather than in the Puzzles section. Here it goes:

You are able to select a team of 20 distinguished individuals (scientists, business leaders, craftsmen etc.). Assume any living person is ready and willing to join your team.

Your team is dropped on an island of abundant natural resources.
Everything you could possibly need for your project is available in its natural form.

On the island you find an unlimited supply of willing laborers.

Assume the weather conditions and political conditions of the island are stable. Also assume there is no external trade.

The question is......How long would it take the team to develop a PC (with the equivalent capabilities, size, features etc. of say an IBM Thinkpad)?

It comes down to a question of what knowledge is worth if it precedes infrastructure.

 

[Rolfe]

Hmmmm, just read 1632, by Eric Flint, and am in the middle of 1633. Much the same problem. Except that didn't have the political stability, that's what made it such fun....

Rolfe.

 

[ceptimus]

My guess: not less than 20 years.

 

[LuxFerum]

I think that the first thing to solve is how to get energy.

 

[Prester John]

Two initial thoughts.
1) in the normal world, knowledge is unlikely to precede the ability to implement it as ideas develop from infastructure and infrastructure from ideas, tey develop together at similar rates.

2) in the artificial environment, having knowledge of what is possible will lead to a relativly quick development. Most knowledge is obvious once it is known. It the step of knowing it that is difficult.


i think!

and

3) Damn the thinkpad, i'm going to Mars

 

[Keneke]

An interesting concept; it often comes to play in armageddon survivalist stories such as The Stand.

 

[El Greco]

The answer to this, presupposes knowledge I don't have. What materials and resources are needed in order to manufacture all computer parts (microchips, plastic, circuits, metallic parts, magnetic media, leds, springs etc) ? What kind of materials are needed in order to make the machines that are necessary for manufacturing all these things ?

Just thinking about it, it seems to me that 20 people cannot guarantee that we have all the knowledge we need. They will most likely have to reinvent stuff. But if we assume that all the knowledge we need is available, then I guess they could make a computer in less than a year. Maybe even less than a month, if we think that they can have millions of people working 24/7 under very specific orders.

 

[MRC_Hans]

How long to reinvent a PC? That depends. Are they marooned there for good, having to make life as comfortable as possible? Then I would say never, or at least not in the first few generations. What need would they have for a PC?

Are they sent there to do that task (the ultimate nerd's reality TV, heheh?)? Then I'd say six months. It might not be your 3.5Ghz super box, though .

Hans

 

[geni]

A PC. Ok you are going to need knowlage of:
glass making
How to make ultra pure silicone
microcircutry design
eletronics
How to make that machines that create the chips
whatever system you are using as a monior display
Programing
quite a few other things

20 poeple? I think you would need more

 

[Michael Redman]

There is no right or wrong answer to this . . .

Actually, there are several wrong answers.

I don't think it would take long. Years, maybe, but not too many. If you could interview the people ahead of time, and make sure that they had the right knowledge, you could develop a plan from start to finish rather quickly. I suppose it would depend on your ability to find 20 people that would have between them all the knowledge you needed. If you could, I would think it would go rather quickly. A matter of months. If you had holes in your knowledge base, and had to work things out, that could really slow the process down.

Hmmm. . . the more I think, the longer it's going to take. Just to get started, you're going to need to build:

Mines
foundries
oil extraction
refinery
chemical plants
power plant

Sounds like a good reality TV concept.

 

[Dragonrock]

Computers of today are designed by humans but built by machines. The actual etching of the chips, laying solder runs, and attaching components is done by computer. It would require building several generations of compters, each more complex generation being built with the power of the previous. It would be very frustrating to build very complex machines who's sole purpose was to operate for a few months to build a couple of slightly more complex machines that would operate for a couple of months...and on and on with each iteration taking longer than the previous.

I believe that in the end we are talking about 10+ years but that's just a guess.

 

[Martin]

Approximately? Next Thurdsay. You pay up front. Cash.

 

[Andonyx]

Your team is dropped on an island of abundant natural resources.
Everything you could possibly need for your project is available in its natural form.

This statement is too ambiguous.

Because the amount of resources would change the faster you wanted to get it done.

Just for energy and diamonds needed to etch alone, if we had unlimited time, maybe 4 workers, and 3 palm trees would be enough. One of the workers would die and eventually turn into coal, and diamond pieces. Over thousands of years.

But if we wanted the energy and so forth now, then we would need an island already preganant with coal and preferrably diamonds already, and an endless supply of trees.

 

[Martin]

On the island you find an unlimited supply of willing laborers

Woohoo! Infinite monkeys! It'll be done in mere hours. Now why didn't I think of that before?

 

[El Greco]

Are there going to be any women on the island ? Because in that case I have to modify my amswer...

 

[roger]

Then I'd say six months.

Really? I don't think it is feasible to _design _ a chipset for a thinkpad in 6 months, given access to modern facilities. And ever tried to write microcode? Takes a while. Ever tried to design an arithematic unit? They are not trivial.

I would bound the task as longer than 6 months, and shorter than 100 years (given that when they are dropped on the island, they will be technologically somewhat behind what the US was in 1900, and a thinkpad was realized in that time).

This link estimates the cost of upgrading a CPU core at 10 man years. That's to improve an existing CPU design, given modern facilities. The skill set required will use up at least 3-4 people from your team.

Energy production
glass making
chemical refinery
metal refinery
paper production - got to document what you are doing
mines
tool shops
schools - to train the natives
curriculum developers
programmers
chip designers
etc.

I really don't think 20 people would contain enough expertise amongst themselves to just walk in and do the job. A lot of stuff is going to have to be rediscovered.

Think about it - making a copycat CPU in today's marketplace is an extremely expensive and risky undertaking, and you already have all the manufacturing facilities, highly trained workers, etc.

 

[El Greco]

Really? I don't think it is feasible to _design _ a chipset for a thinkpad in 6 months, given access to modern facilities. And ever tried to write microcode? Takes a while. Ever tried to design an arithematic unit? They are not trivial.

I have considered the above to be already acquired and readily accessible knowledge. We will not have to design them. We have people who know how to do it almost mechanically.

 

[Michael Redman]

Think about it - making a copycat CPU in today's marketplace is an extremely expensive and risky undertaking, and you already have all the manufacturing facilities, highly trained workers, etc.

In the real world, you can't use someone's already proven design, because it belongs to them. But you don't have to innovate or "upgrade" in our scenario, just build something that already exists, and which at least some of your experts should be very familiar with already. There shouldn't be any real R&D, except where holes in the knowledge of your experts exist.

 

[roger]

I have considered the above to be already acquired and readily accessible knowledge. We will not have to design them. We have people who know how to do it almost mechanically.

No such thing. In the real world it takes a long time to make a CPU, with copious amounts of reference material, experts for hire, etc.

Just ONE task needed to make this CPU - circuit line layout. Getting millions of transistors on a chip is not a task suitable for the human mind - it must be done by a program. However, the problem is NP-complete, essentially unsolvable. Thus, programs use a wide variety of heuristics and testing to come up with a resonable layout. Writing just this program is not a mechanical process. This one program represents man years of work, assuming access to a fully functional computer and compiler.

And without this program, a modern CPU is not getting built, period.

 

[roger]

In the real world, you can't use someone's already proven design, because it belongs to them. But you don't have to innovate or "upgrade" in our scenario, just build something that already exists, and which at least some of your experts should be very familiar with already. There shouldn't be any real R&D, except where holes in the knowledge of your experts exist.

Certainly the task is eased if we have a complete set of specifications for the chipset - but that is not part of the initial question, if I read it correctly. But that is probably not what you are saying.

Modern chips are too complex for a mind to encompass. You _are_ pretty much starting from scratch, in any real sense of the world. You will just steal the instruction set and architecture from some existing design, sure, but implementing those, in the real world, given all of our resources, takes many, many man years. They don't get whacked together in the real world, and it will take much longer on our island.

You need _serious_ computer resources to build a modern CPU. It'll be like another poster posited - build a very weak computer to allow you to build a better one, etc, etc, etc. Each time you will need to rewrite/upgrade an operating system, assembler, compiler, utilities, etc.

My point was that the upgrade cycle for an existing CPU is mcuh simpler than what we are proposing, and I think that making a copy cat chip in the real world is much easier than what this thread proposes.