5

Is anyone aware of any implemented algorithms/programs that could fall under the category of Seed AI? By this I mean "recursive self-improvement."

In the wikipedia article they talk about compilers but do other examples exist?

Maxim Gershkovich
  • 45,951
  • 44
  • 147
  • 243

2 Answers2

2

If I read your linked Wiki article correctly, this is at the moment a hypothetical conept:

Seed AI is a hypothesized type of strong artificial intelligence...

Also, the compiler example is, as they say, very limited in explaining the concept, as it is only single-step and by definition the Seed AI algorithm should be infinitely recursive.

A limited example is that program language compilers are often used to compile themselves. As compilers become more optimized, they can re-compile themselves and so be faster at compiling.

However, they cannot then produce faster code and so this can only provide a very limited one step self-improvement.

If you want to know anything more, I would suggest having a look at the organizations that are doing active research in this field.

  1. Singularity Institute
  2. AGIRI.org
  3. Texai
  4. Halliburton

UPDATE

Consolidated Robotics does the AI (under the name of Consolidated), but they were bought out by Halliburton in 2005.

In 2005, the privately held Consolidated was acquired by energy conglomerate Halliburton for an estimated price of $25 billion US. The company continues to operate under the name Consolidated, though as a wholly owned and operated division of Halliburton.

Community
  • 1
  • 1
Nico Huysamen
  • 10,217
  • 9
  • 62
  • 88
  • 1
    I understand that infinitely recursive optimizing algorithms don't exist but I was interested in one that could perhaps recurse 5 or 10 maybe 20 times. Will follow your links. PS: Halliburton, really? God help us! lol – Maxim Gershkovich May 09 '11 at 10:29
  • 1
    Citation for Halliburton? I can't find anything through Google that would suggest they're doing any kind of AI research. – Cerin May 09 '11 at 23:10
-1

One can build tiny compilers that can be augmented by additional external information. This is a kind of learning. One can start extremely small ("seed") and grow the compiler by adding knowledge until it can start acquiring it. One has to start somewhere, and assuming a seed is already big enough to grow itself without help seems self-defeating; how did it get that way?

See https://stackoverflow.com/a/35676159/120163 for a truly tiny compiler that can be bootstrapped.

Community
  • 1
  • 1
Ira Baxter
  • 93,541
  • 22
  • 172
  • 341
  • This does not provide an answer to the question. To critique or request clarification from an author, leave a comment below their post. - [From Review](/review/low-quality-posts/11722239) – Jayan Mar 22 '16 at 02:53
  • I disagree that this doesn't answer the question. Now we're into opinions. – Ira Baxter Mar 22 '16 at 03:12
  • IMHO, it did not look like an answer. It is valuable as comment. Which is what I voted for. – Jayan Mar 22 '16 at 13:55