My question today is - what kinds of tasks/jobs would a computer that is self-programming (able to create more programs/add to its operating system) be particularly good at? This computer wouldn't have the fastest or largest processors, but rather a series of highly interconnected ones.

Hope this helps,
Pat

Actually, though, the mainframe would learn just like anything else. There would not be real programming of the core system. It might, though create programs to make the system, entertainments, and things people might use.

Or it could re-stock you fridge.

I'd also like it to monitor my wishlists and the marketplace for future purchase saving advantages.

None of that has anything to do with upgrades, though. My high end apps all have automated updating anyway.

New apps that it might seek out would be derived from perceiving my interests and loading them. I'm thinking about specialized programs like Celestia, Anim8or, Dragon, Case Catalyst, Quartz, CAD, media and vocational apps, though I could see some hobby, travel and tourism apps too.

I'd love to have a computer that tracked traffic so I could avoid it, auto and pedestrian. Gosh, I'd murder to have a schedule proposed for me based on low visitation times at malls, grocers, theme parks, movie theaters, etc.

The computers are in a university lab setting, with the intent that the geeks are going to figure out how to use them and program them (with the computers' help) for certain uses. Clearly the original designers had some applications in mind. This style of computer isn't going to be the best at ballistics projections, performing massive calculations, doing advanced math or real-time weather updates because it's not *fast* (in computing terms.) But it has a number of interconnected processors (not massively parallel processing, but massively nodal? It's meant to mimic the concept of neurons), and it can add to its operating system.

And thanks for the reminder, skadder. No, it won't create software that knows how to run on copper or fiber <imagine me rolling my eyes here, LOL>

Oh, and there aren't very many of these (very costly to build.) Limited numbers, massively interconnected, what industry or government use would make the most sense? Breaking codes? (maybe not, not fast or powerful enough) War games? (sigh, been done to death) Telecom is appealing, but single point-of-failure (one computer controls it ALL, muahahaha) is the stuff of trite action flicks. Adaptive something or another...Toby, I'll take a look at evolutionary programming, there's probably something there.

But keep the ideas coming, this is really helpful (even if it just helps me narrow down the field to what these computers are NOT.)

Running simulators would be one powerful way to do it. If you consider how simple games are and what they can do, consider if it creates a "game" for each character, dictating how it is going to operate. It learns and adjusts the character when needed. In this case, it would not be using all the necessary power that running the whole game itself might be. Each node might run an aspect of the whole game, and It just acts like a referee or dungeon master. It would not have to be very powerful for that as the other nodes would do the grunt work.

Humans might give the computer the basic design for a new vessel (of whatever kind) and the computer then runs simulations on other systems, and works out the programs to make best use to run them. Instead of people sitting and programming, it would run simulators elsewhere through their paces, then guide programming computers through their paces to make sure everything works as needed. It might correct parts designs to deal with problems simulation might show.
It might not do real programming, but more the grunt work of guiding other systems to do the real job.

The real need of this computer is not where brute force is needed, but instead where patience is needed. It could be considered obsolete and therefore not put to use by officials. It might just sit there and sift through data and watch for trends, rather than details. That would require less computing power.
One application might be to test a theory, such as the concept that communications of a certain nature increases when actions of a certain kind (war, market shifts, etc) might be developing. It would poll computers sifting through the communications periodically and record the levels.

Again, an outdated system might be used for work that is not time critical. It might be given a problem that the answer is wanted eventually, but not needed right now. Solving a mathematical theorem or testing an alternative theory for the universe. It takes data from other computers and play with the data.

Or of course, the computer network could sit there for a millennium and spit out "42"--but that's been done :-)

here is what you can do.

It does not take a very intellegent computer to emulate a person, especially if the person is not brilliant. We have had for years the thousand question programs where you try to see if the person on the other side is human or computer.

My idea, is that the government has a parliment, and a president (remember I am an american so correct the design for reality). The president is always human, can serve only two terms of four years max, and is strictly a figurehead, a face of the government.

the parliment operates through conferencing programs. they never meet personally. The main purpose of the prime minister is to guide the parliment's policy.

The idea I am thinking of, is that someone took an old computer and programmed it to emulate a human. It was run for palriment and ends up getting elected as prime minister.

What it does is search the data networks and ask questions. It never really makes a decision, but forces others to test their theories and make decisions.

The computer keeps getting elected as prime minister as it is so good at it. The rest of the parliment's people do the work that people would do. The operators of the computer does the face work the prime minister would normally do, visiting constituents and such.
Your story might be where there is a lot of pressure to have a face to face meeting of all the parliment and the computer is not going to be able to show up for obvious reasons.

Another way is that the computer might be getting glitches and may show signs it may have to be retired. It is writing programs to protect the intellect and data but is losing the battle.

Another is that everybody knows it is a computer. Some organization within the government decides they are going to take over the president computer. they have the full power of the data network while the president computer is just one installation. It is working hard to create programs to block their attempts to take over, using contenplative designs while the data network computers are using brute force to try to crack the protections. It has the advantage that it can change the program since it is concious and can be imaginative while the other computers are pure brute force, oversized calculators.

Since it can program computers and 'likes' to improve itself, it could expand into other computers on the LAN and program those computers (telnet or SSH to the computer, using the passcodes it learned from its own users (same users/diff. computers/same PWs) and plant new or changed SW. It may program the other computers for its own self-improvement and/or because that is what it is evolving to do: improve computers; make them self-modifying, too, i.e., REPRODUCE. Bwa ha ha. It could be infected by a virus, learn from it, and make viruses to infect other computers on other LANs through pen drives, USB drives, Internet....

Most importantly, it should create a defensive TaleSpinner detector.