In complex enviroment, better inteligence gives you better chance to survive. But complexity of every real-world environment is finite. Assume that more inteligence need more computing power, more resources. Then one day self improving inteligence gets to the level when it can handle complexity of its environment optimally and improving itself will only consumes more resource but not gives additional advantage in that environment. Then artificial inteligence (because is inteligent) will stop self improvement and freezes its own inteligence at optimal level of inteligence for environment it lives in. And maybe in our environment, artificial inteligence can reach optimal state even before it overcomme human inteligence and then singularity will never happen. What do you think?

closed as off-topic by user2953, James Kingsbery, virmaior Jul 28 '15 at 4:52

This question appears to be off-topic. The users who voted to close gave this specific reason:

  • "Questions that push a personal philosophy with no question beyond "am I right" or "what do you think" are off-topic here as this is not a blog. It's ok to express unique opinions, but you must have an actual, answerable question to go with them." – Community, James Kingsbery, virmaior
If this question can be reworded to fit the rules in the help center, please edit the question.

  • Complexity is not finite in every environment. In an computational/logical environment we can define objects with infinite complexity. For example, almost all real numbers possess this property. I think you will need to restate your question, otherwise it will probably be closed for being "opinion based" and therefore outside of the site's desired format. – Nick Jul 25 '15 at 18:10
  • @Nick R: Good comment. I changed that sentence to "every real-world environment" :-) – user3123061 Jul 25 '15 at 18:23
  • I don't want to nit-pick, but even in a finite environment we can define objects of arbitrarily large finite complexity. Two electrons interact in a way that is computationally more complex than "2". There are (best current guess) 10^80 particles in the observable universe, but we may not be able to describe the observable universe with less that 10^80 bits. – Nick Jul 25 '15 at 18:31
  • @Nick R: Ok. You are right. Then, does inteligence grow lineary with available resources? If you give it two times power, will it be two times inteligent? I am not expert, but i have intuitive feeling that there will probable be some strong nonlinear relationship. Something like 4x more power to be 2times more inteligent or even worse. And because there are space-time-energy limitations there will be some optimal level of inteligence for every real environment. – user3123061 Jul 25 '15 at 18:41
  • I'm no expert either. We do not know enough about reality to say whether or not it is computable in finite terms. Any computer that is intelligent and able to reprogram itself should witness a period of exponential growth of its intelligence before converging (perhaps slowly) on something near an optimal state. The logical resources required by any computer are minimal - all computers are logically equivalent. Adding physical resources will improve relative performance between computers and that relation too may be exponential, depending on the nature of added resources (cont'd...) – Nick Jul 25 '15 at 19:04

Its optimal intelligence is limited only assuming the environment is finite in all aspects and that it has no goals other than to react to its environment as appropriately as possible. However, this is not true when you consider time in our universe. Chaos theory roughly says that disorder increases over time and small differences in the initial states in two systems build over time. This implies that there may not be a maximum intelligence due to it thinking about the increasingly more long-term effects of the current state of the system. The assumption also isn't true if it has creative desires. Creative desires would cause it to start imagining/thinking about its own theoretical universes and "things" once it gets bored with the environment it exists in and it feels there is nothing more to be gained from the environment it lives in.


The universe itself is limited, which means the possibilities of an artificial intelligence are limited. I think your question is "assuming an artificial intelligence would have a reasonable means to try to improve itself, would there be a point where improvement is either not possible, or where the AI would have good reasons to stop improving?"

An autonomous AI, even if it wasn't highly intelligent, could improve itself by buying better or more hardware. There might be limits how much improvement is possible that way. For example if the AI needs to use its intelligence to make money to pay the electricity bill and can only use what remains to improve itself, there might be a point where it grows to a size where the electricity bill gets too high.

Improving itself by improving its code is hard. Only a highly intelligent and well trained human could do this, so it would require a highly intelligent AI as well to do this. Much more intelligent than the average human. Again, improving say the intelligence of an AI that has an IQ of 180 might be so difficult that an AI with an IQ of 180 cannot do it. We don't know.

And then there might be a point where an AI has reasons not to improve itself. For example, if there is legislation that all AIs above a certain complexity will be destroyed out of fear, then the AI won't want to be covered by that legislation. Maybe if it increases its hardware, delays between components might increase and the AI might feel that as unpleasant. Or maybe improving itself might be hard work, and the AI might be too lazy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.