Self replicating robots of another kind

[alan2here]

Robots. The more complicated you make them the more complicated they have to reconstruct to make copies of themselves.

 

I have an idea, how about.

 

Buying a building. Putting a computer in it. Having the computer do something on-line that makes it money such as renting it's CPU power. The money it makes has to be enough to maintain the building it is in and make a prophet (although it could be quite a small prophet).

 

When enough money is saved the computer has to buy a new computer and pay the person who delivers it to install it. This if needed could be done in two chunks. Delivered with instructions of how to get in and leave it inside and then instructions to someone else about how to get in and install the computer for the next day.

 

The person who installs it needs to run it with the correct software. This could be as simple as put this USB pen in here and this one in here, turning the PC on and then taking another one out. Theoretically making a copy of the software for another install as well as starting the PC booting from the correct information. He also needs to connect the PC to the Internet or to another of the PC's in the room.

 

The Network needs to maintain a bank account on-line that it can use to pay bills, pay it's workers and pay for materials.

 

Bills:

Building maintenance

Internet connection

 

Workers:

PC\Network Installer

Cleaner or building maintenance guy

Expander

 

Materials:

PC's

Memory Pens

 

Once the network has got to a certain size then it could expand to another building. This then requires someone to be paid to check out the building (lets call this person the expander) and put the things in place to start a new unit. The new unit could have it's own bank account and an Internet link back to it's parent.

 

Each year after a certain age provided the parent has done this already the unit could have a small chance of selling all its assets including the building (death) and giving it's money to its parent or one of it's close relativities brothers\sisters\children etc... This would happen much less often than an expansion.

 

I was planning to avoid any central maintenance however someone (a person) will have to keep the list of resources (including human resources) and templates for how to send convincing emails and make payments up to date.

[Pangloss]

Well this is perfectly valid structural thinking and you should embrace your ideas and keep thinking about them. The main problem is maintenance, which mandates a certain level of human interaction, ramping up in a fairly linear fashion according to how many computers you have and what their hardware utilization rate is.

 

Web Server farms are a fairly typical scenario that you could study for existing data. They have a very high ratio of computers to service personnel, and a built-in incentive to eliminate as many human beings as possible. You could also look at Data Centers in general, perhaps, or (just for grins and giggles) you might ponder MMO clusters or supercomputer networks. (MMOs are typically quiet about their infrastructures, but that's been changing in recent years as the tech matures, and they're interesting because they have high responsiveness requirements compared with business networks.)

 

But note that the incentive for no-human clustering is already there, and yet it doesn't exist. There are obviously reasons for that, and the best thing you can do to further your idea is to study what those reasons are. Look into hardware failure rates, for example, and try to figure out a rating of man-hours per type of software/task/application being run, etc.

 

I would say that your problem area revolves around the linear escalation of human personnel to increased machinery (i.e. the more machines you add, the more human beings you need, and the number of machines to humans appears to be a fixed ratio). If you can break that chain (reduce that number or eliminate the "fixed" nature of that variable), you might be able to find something there worth investigating/developing/publishing.

 

A literature survey at the IEEE or ACM portals would be a good place to start. Look at the publication journals, build a good set of keywords, and be sure to narrow your results to, say, 2002 or later.

 

Good luck.

[alan2here]

Thank you for a very good post.

 

The booting from custom built software instead of running software from an OS was to greatly reduce software faults. Do you think the hardware failure rate would be particularly large?

 

I was thinking that if hardware seemed to have failed (could be monitored by the rest of the network) it could often just be left and not communicated with, this is also because acting naturally with any human resource would be a goal. In a commercial environment however it makes more sense to get a person to repair that PC thus scaling the people involved to the computers required. Perhaps the "death" mentioned in my post above post could be a way of resolving old age (lots of failed components, run out of space in the building). By then it could have several large children and many grandchildren. The reproduction also allows units (farms) to die (and so redistribute there wealth and covered in the above post) due to unexpected faults or real world problems such as communication problems (as effectively emails from bots are being sent to humans) or fire etc...

 

I wonder how legally the ownership of the building could work. I could imagine the government claiming buildings it found that had been purchased automatedly where there was no obvious path back to a human owner.

 

IEEE seems to involve purchasing articles. However the website implies that affiliation with certain institutions could allow you to be exempt from paying so I will check with my UNI when I return after easter.

 

I'm no great academic when it comes to creating formal reports, best practice in research, academic standards etc... I have learned it to some extent in UNI but only to a foundation degree level and that was never my strong point.

 

 

Good luck.

:¬)

[Daecon]

What about if you have a computer running an automatic computer-building factory to build computers to run automatic computer-building factories?

[Pangloss]

The booting from custom built software instead of running software from an OS was to greatly reduce software faults. Do you think the hardware failure rate would be particularly large?

 

Hardware fails, yes, but interestingly you typically have a useful number on how often the hardware fails, such as the Mean Time Between Failures rating for hard drives. I don't know what "large" is, but at this point you probably don't know either -- what constitutes "large" is going to be loosely defined as something like "what makes me add another person to the payroll", and that's a variable that you'd have to puzzle out and nail down.

 

Bear in mind, by the way, that writing software that's not running off an OS is extremely time-consuming and problematic -- not for novices. You'll have to write all your own networking code, too. I don't know what your experience level is but I thought it worth mentioning.

 

 

I wonder how legally the ownership of the building could work. I could imagine the government claiming buildings it found that had been purchased automatedly where there was no obvious path back to a human owner.

 

Well this one is simple enough: Real estate ownership is a matter of public record, and computers can't own buildings.

[alan2here]

Nobody is on the payroll. The system of how this is the case it described im my first post.

 

If computer would have to provide a real name and details to purchase buildings then I guess someone would have to own all the buildings on behalf of every unit.

 

This is all theoretical. I shouldn't think there are too many individuals capable of setting this up alone, it may need to be observed in detail to iron out the bugs.

[Mr Skeptic]

An AI intelligent enough to earn money on its own would have to be made by a human. The human(s) that created it would consider themselves the owners of that AI. How would an AI acquire its own independence?

[alan2here]

The human would have to disown it. While the human considers themselves still in control there would be to much tenancy for them to intervene.

[alan2here]

After a few new buildings had been occupied if the human lost track of one of them. This would effectively give it independence.

[bascule]

What you're describing sounds something like a cloud computing installation, such as Amazon's EC2.

 

You'll still need a full-time maintenance staff, just to handle system failures. This is especially true if you have large numbers of computers. If you have thousands of computers, you'll see multiple hardware and software failures daily which need to be addressed.

[insane_alien]

well, the AI could hire some human employees until it has developed robots to replace them.

[alan2here]

well, the AI could hire some human employees until it has developed robots to replace them.

I wasn't proposing the AI had any real intelligence. How were you proposing that it design and build robots?