From: Mark Nuzzolilo (email@example.com)
Date: Mon Jun 16 2008 - 21:47:56 MDT
What is technically defined as external input? If this can simply be
fulfilled by using perceptual data to gain facts and data about the
environment, then this data can serve as fuel for improvement, and data is
cheap. Or does external input have to satisfy its own complexity
requirements? (excuse my ignorance if this is basic academic knowledge,
maybe somebody could point me to specific literature if needed).
On Mon, Jun 16, 2008 at 8:16 PM, Krekoski Ross <firstname.lastname@example.org>
> You would run into a complexity ceiling if it was machines improving
> machines without external input.
> On Tue, Jun 17, 2008 at 1:19 AM, Mark Nuzzolilo <email@example.com> wrote:
>> I'll take a swing at this.
>> Let's start with the assumption that a machine cannot output a machine of
>> greater algorithmic complexity.
>> Now for a thought experiment put humans in that same category. A single
>> human would not be able to produce something "greater" than itself. The
>> details of this are unimportant. The point is that when you take a larger
>> group of humans, the complexity increases and you can now produce a machine
>> potentially greater than a single human. This machine could then improve
>> the intelligence or ability of single humans at a time, and then those
>> humans could then create a greater machine.
>> This is obviously not a "typical" RSI scenario but if my reasoning is
>> correct here (correct me if I am wrong), then in theory RSI would be
>> possible even by taking this concept and abstracting it to specific (and
>> properly designed) AGI components rather than specific components of a group
>> of humans (the humans themselves).
>> Mark Nuzzolilo
>> On Sun, Jun 15, 2008 at 1:18 PM, Matt Mahoney <firstname.lastname@example.org>
>>> Is there a model of recursive self improvement? A model would be a
>>> simulated environment in which agents improve themselves in terms of
>>> intelligence or some appropriate measure. This would not include genetic
>>> algorithms, i.e. agents make random changes to themselves or copies,
>>> followed by selection by an external fitness function not of the agent's
>>> choosing. It would also not include simulations where agents receiving
>>> external information on how to improve themselves. They have to figure it
>>> out for themselves.
>>> The premise of the singularity is that humans will soon reach the point
>>> where we can enhance our own intelligence or make machines that are more
>>> intelligent than us. For example, we could genetically engineer humans for
>>> bigger brains, faster neurons, more synapses, etc. Alternatively, we could
>>> upload to computers, then upgrade them with more memory, more and faster
>>> processors, more I/O, more efficient software, etc. Or we could simply build
>>> intelligent machines or robots that would do the same.
>>> Arguments in favor of RSI:
>>> - Humans can improve themselves by going to school, practicing skills,
>>> reading, etc. (arguably not RSI).
>>> - Moore's Law predicts computers will have as much computing power as
>>> human brains in a few decades, or sooner if we figure out more efficient
>>> algorithms for AI.
>>> - Increasing machine intelligence should be a straightforward hardware
>>> - Evolution produced human brains capable of learning 10^9 bits of
>>> knowledge (stored using 10^15 synapses) with only 10^7 bits of genetic
>>> information. Therefore we are not cognitively limited from understanding our
>>> own code.
>>> Arguments against RSI:
>>> - A Turing machine cannot output a machine of greater algorithmic
>>> - If an agent could reliably produce or test a more intelligent agent, it
>>> would already be that smart.
>>> - We do not know how to test for IQs above 200.
>>> - There are currently no non-evolutionary models of RSI in humans,
>>> animals, machines, or software (AFAIK, that is my question).
>>> If RSI is possible, then we should be able to model simple environments
>>> with agents (with less than human intelligence) that could self improve (up
>>> to the computational limits of the model) without relying on an external
>>> intelligence test or fitness function. The agents must figure out for
>>> themselves how to improve their intelligence. How could this be done? We
>>> already have genetic algorithms in simulated environments that are much
>>> simpler than biology. Perhaps agents could modify their own code in some
>>> simplified or abstract language of the designer's choosing. If no such model
>>> exists, then why should we believe that humans are on the threshold of RSI?
>>> -- Matt Mahoney, email@example.com
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:03 MDT