3
$\begingroup$

There is no doubt as to the fact that AI would be replacing a lot of existing technologies, but is AI the ultimate technology which humankind can develop or is their something else which has the potential to replace artificial intelligence?

$\endgroup$
  • 3
    $\begingroup$ Artificial Intelligence is a broad field that is comprised of various conceptual aspects. $\endgroup$ – quintumnia Dec 9 '16 at 11:01
4
$\begingroup$

By definition, artificial intelligence includes all forms of computer systems capable of completing tasks that would ordinarily warrant human intelligence.

A superintelligent AI would have intelligence far superior to that of any human and therefore would be capable of creating systems beyond our capabilities.

As a consequence, if a technology superior to AI were to be created, it would almost certainly be created by an artificial intelligence.

For the purposes of mankind, however, superintelligent artificial intelligence is the ultimate technology due to the fact that it will be able to surpass humans in every field, and, if anything, replace the need for human intelligence.

In our past experience, intelligence has been the most valuable trait for any entity to manifest - for this reason, in an anthropomorphic context, we can predict that artificial intelligence will be the ultimate achievement.

The main reason why we will certainly not be able to replace superintelligent AI is that it will surpass us in every respect - if there is ever any replacement, it will be created by the AI similarly to the way we may create an AI that replaces us.

$\endgroup$
  • $\begingroup$ Who's gonna invent the test program to test whether any given "AI" is "superintelligent"? :) Who would write that program? quasi-superintelligent AI? humans? how can humans write a program to check if something is smarter than humans? $\endgroup$ – Erik Kaplun Dec 14 '16 at 16:16
  • $\begingroup$ That is funny @ErikAllik. Unfortunately for us, we will know immediately once an AI becomes superintelligent. $\endgroup$ – GJZ Dec 14 '16 at 16:17
  • $\begingroup$ How will we know that? Is there a measure for intelligence? $\endgroup$ – Erik Kaplun Dec 14 '16 at 16:18
  • $\begingroup$ Consider this story about AI mentioned by Stephen Hawking @ErikAllik: $\endgroup$ – GJZ Dec 14 '16 at 16:19
  • 1
    $\begingroup$ Humans create superintelligent artificial intelligence. The first question they ask it is whether there is a God. A bolt of lightning strikes the plug so that the AI cannot be disabled and it responds "There is now." $\endgroup$ – GJZ Dec 14 '16 at 16:21
1
$\begingroup$

A new physical lifeform could outperform and replace artificial intelligence when it has feedback from organism (its body) to its design information (replacement of genes).

This evolution is expected because:

Artificial intelligence will redesign its own software very soon in its evolution. After that, it will be restricted by the performance of the available hardware and communication speed between parts. Therefore it will design better processing hardware for itself, to run its next generation.

To squeeze most processing power out of a given amount of resources (matter, energy) and circumstances (temperature, radiation) the design has to be small (material resources and delay of interconnections), energy efficient (heat evacuation), and adapted to the kind of functions used by the software (hardware architecture).

To tackle this profoundly, artificial intelligence will design the new hardware at atom by atom scale. This leads to new problems of natural degradation by radiation, atomic decay and other quantum-mechanical problems and opportunities. The solution of these new problems is redundancy and the ability to repair degraded parts by atomic-level machinery. This atomic level repair machinery is the same machinery which builds and extends the hardware for new individual artificial intelligence systems. Since this feature is there, it can, and will, also be used to restructure parts of the hardware while it runs to integrate (compile) knowledge in hardware (more efficient).

The machinery to build and maintain such hardware could be inspired by the biological machinery which will be understood by the system by then. However, when the artificial intelligence refactors these principles with full understanding and anticipation, the resulting "hardware" will be quite different from the old biological machinery and very different from the static silicon based processor structures.
The main differences are:

  • The design information will be available, relating the features of the realization with design choices. This provides direct feedback from performance of the realization (the hardware, the body) to the design information. That augments the design information for designing new generations. This feedback channel is the main difference between the new machinery and biological life. Once that exists, it will be used for everything, not only for processing hardware for artificial intelligence.

  • The new design described here is basically processing hardware rather than a body for fighting and propagating (although the eternal fight for resources will not end with the dawn of artificial intelligence).

  • It will use more compact molecules because it is designed rather than evolved blindly. (The current biological life uses monster-molecules evolved by random changes until one or other corner of the molecule has the right shape to catalyze a specific chemical reaction).

  • Since parts of the hardware can be restructured while the system runs, the distinction between hardware and software will become very fuzzy (as in biology).

The drastic increase in efficiency and evolution speed let it outperform the old biological life (which lacks design and feedback) and outperform artificial intelligence (which was not integrated in matter).

When this stage is reached, the systems will look like a natural, intelligent, propagating life-form and therefore supersede the stage of artificial intelligence running on human-made processing hardware.

This answers the last part of the question: "... or is their something else which has the potential to replace artificial intelligence".

$\endgroup$
  • 1
    $\begingroup$ Artificial intelligence can be so intelligent for the very reason that it is not constrained by its biology. $\endgroup$ – GJZ Dec 16 '16 at 7:40
  • $\begingroup$ No matter how intelligent the system, it will be constrained by its hardware (biology or other material structure). Physical restrictions impose small sizes to reduce the propagation time of signals between parts (speed of light). The original answer is edited to clarify that the new atomic level machinery is the hardware of the artificial intelligence rather than a body and that knowledge gets integrated in the hardware. Remark that I answered the second part of the question: ".. is their something else which has the potential to replace artificial intelligence". $\endgroup$ – Luc Claeys Dec 17 '16 at 14:14
  • $\begingroup$ That entirely fails to clarify why a lifeform based on genes would be any more effective than extremely efficient self-replicating and self-improving (at a much faster rate) hardware. $\endgroup$ – GJZ Dec 17 '16 at 14:48
  • 1
    $\begingroup$ A superintelligent AI would develop something similar to the concept of computronium, as outlined in The Technological Singularity by Murray Shanahan. This material would be able to concentrate the largest physically possible amount of calculation capability in a given area. $\endgroup$ – GJZ Dec 17 '16 at 14:50
  • $\begingroup$ This would be the theoretically most efficient (and therefore completely irreplaceable) configuration of particles possible for the purpose of intelligence. No lifeform of any kind can compete with that. $\endgroup$ – GJZ Dec 17 '16 at 14:52
0
$\begingroup$

In order for something to replace AI, it would need to out perform AI. AI currently uses number systems to represent information and logic to perform operations. So the replacement would need to be based off of something more efficient than numbers and logic. Some sort of super-logic. Or something similar to intuition and instinct which do not require linearly figuring things out.

$\endgroup$