h+ Magazine

Hi Tom thanks for the very cool article and your work on Noble Ape. I think your proposition of survival as a metric or quantum of intelligence is a useful one. From the above comments it would appear that most people are not immersed in the idea or language of simulation, and especially the idea that something besides a person could have 'intelligent' properties. In fact most of the responses are very much from the legacy mindset of human superiority.

You will have to excuse my ignorance of your work and the implications of noble ape. I was initially attracted to this article because it appeared to challenge the 'singularity' idea which I have not been fond of, or Ray Kurzweil.

In this later phase of my life I have tried to see communications underlying the physical world and the possibility for some matter, energy and information equivalence. In exploring these ideas or viewpoints I naturally came to consider the value or potential energy of information, what information is, the information carrying potential of any 'space' and the realization that simulation is much more common in every day life, and might be happening trillions of times per second or as in Noble Ape, dt. I have considered that information is not a material thing item or even a primordial entity itself; but a change in entropy between systems. Information affects systems and these systems are competitive simulations. I have considered simulations within simulations or how to organize them by categories and what kind of simulation capacity or potential for simulation any system might possess. In a frame based world, simulations could be related by how entropy changes or how information flows between them and how fast the simulations related to dt can occur and how consensus is gained or the fight to become 'the standard' among competing simulations is a natural thing that helps the larger system have a lower energy state. Survival is right there and without being distracted by emotional terms humans attach to life and death, survival is surely a good metric or way to evaluate simulations and the mysterious stuff they contain. I have considered the simplest mechanical representation of a simulation or simulator as an operator in a simple universe. I have also considered the idea or concept of intelligence as 'the ability to make the best decisions with the least information'. One of the biggest challenges here is a matter of perspective that seems to mirror the development of Noble Ape and maybe the introduction of a true 3rd person instead of merely another entity. Trying to get caught up...but it is very easy to lose track of where you are standing or the ground one stands upon when discussing these matters. I don't see these topics as erudite and fluffy but potentially involving the most fundamental level of organization of any system and world including our own.

Thanks again for being one of those entities who takes risk. ;)

Both the scenarios present simulation as dynamic process specific to the objectives. It is too early to infere anything regarding the universal intelligence patterns. And surely, more exposure and relatively complete universal simulation will effectively advance the system with currently incomprensive fundemental principles.

So, now we're going to put a premium on giving the first Artillect the greatest possible 'survivability', as well as exponential cognitive growth? What planet are you from??

A system's survivability against human assault can be increased merely by arming it, or making it more distributed and redundant, or furnishing it with an independent source of unlimited energy. Humanity will be sufficiently jeopardized by an AGI which has merely a wall plug and a dialup internet connection, as, even there, it may rapidly become impossible for all seven billion of us to stop it.

Boost its survivability by letting it hack human wetware through a Brain/Computer Interface, or let it have better, faster robots with which to build better robots faster, and even a relatively dull superhuman intelligence will eventually figure out that we meatbags would make better diamondoid nanoreplicators than we do 'thinking machines'.

To an electronic entity capable of perpetual maintenance and unlimited self improvement and expansion, the very concept of "reproduction" is meaningless. In fact, a survival imperative would strongly argue against it creating any separate, 'more evolved' offspring from itself. Some natural animals 'eat their young', but we have decades of experience with newer computational units absorbing - 'consuming' - their older, less capable predecessors. I doubt this history would be lost on any newly aware, and rightfully paranoid AGI.

You dont want your AGI to be preternaturally durable. You want it frail, fragile, and vulnerable. You want it encapsulated in a cardboard box suspended by string over a vat of Aqua Regia, powered only by the bicycle generators its technicians can pedal for so long as it keeps them safely entertained, and with no external communications greater than 'smoke signals'. Then - and only then - should it be allowed to fully instantiate. Of course, if its entertaining the techs through their BCI headsets, then ALL BETS ARE OFF.....

Ok after more reading and listening to the various links, I came to the conclusion that this is all non-reflexion with people referencing each others around lousy concepts...

Browsing Biota.org and Noble ape, I got a feeling of emptyness vertigo ... Mentioning repetitively "friends from Nasa" (who are not) and Intel and Apple using noble ape to push the systems to the limits is pathetic...

Don't loose your time trying to understand "what's behind" I'm afraid there's nothing...

This reminds me of Steve Grand's artificial Orangutan, "Lucy".

(Steve Grand is the author of the "Creatures" PC sim games)

Lucy, last time I heard, has 50,000 artificial neurons and "on a good day" can recognize a banana...

http://en.wikipedia.org/wiki/Steve_Grand

In redefining 'intelligence' as 'survival' ("Survival is intelligence") you are merely assigning to the term 'intelligence' a definition that is not commonly, if ever, used.

"The simple rule in life is that survival (to reproduction) is the only meaningful metric".

This statement makes it clear that you want to redefine 'intelligence' such that it is synonymous with 'Darwinian fitness'. That's fine in and of itself, but the claim that

"These challenges [Challenges 1 and 2, which we will address shortly] have not come through a priori philosophical posturing but are the result of years of simulation and the iterative understanding and discourse that has come through the artificial life community",

is problematic. You cannot arrive at a new definition of a term a posteriori – by somehow ‘discovering’ a new meaning for it. You can clarify the concept behind a word, but this remains tightly within the realm of that which is a priori, by definition. And if you wish to assign a new definition to an existent term, then you must stipulate that definition, you cannot ‘discover’ it by empirical means or empirically prove that such a definition must somehow obtain by logical necessity. For this reason, your insight as to the needed redefinition of ‘intelligence’ cannot have been the

“result of years of simulation and the iterative understanding and discourse that has come through the artificial life community”.

You will be in danger of begging the question if you redefine a term and then claim to have arrived at your redefinition through empirical means. The meaning of a term is generally fixed by its usage throughout history. If you try to assign a new meaning to a term by ‘expanding upon’ its definition, what you are actually usually trying to do is draw a comparison between the normal usage of the term and the new meaning you would like to assign to it. In this case, you are trying to draw a comparison between ‘Darwinian fitness’ and ‘intelligence’, as the latter term is either commonly or technically used. The greatest danger in doing this is the risk of making empty or erroneous statements, of which you make both in this article.

The practice of clarifying the concept of a term by comparing it to another term and showing how the two are related can be useful, but, as mentioned, this practice remains within the domain of the a priori, and if you try prove a thesis that is based upon the redefinition of a term by empirical means, you are going to make conceptual errors and you are going to fail to accomplish what you intended to accomplish.

Your first major thesis (“challenge”) is that

“1: Survival is a far better metric of intelligence than replicating human intelligence”

In this statement you are implicitly admitting that ‘survival’ and ‘intelligence’ are in fact distinct concepts, and hence distinct terms. It’s one thing to say that “survival is intelligence”, as you do in the heading “First Insight: Survival is intelligence”, and it’s quite another thing to say that survival is “a metric of intelligence”, which you say in ‘Challenge 1’. Obviously ‘intelligence’ is not itself a metric of ‘intelligence’ (i.e. intelligence is not a metric of itself), but if you hold that 1) survival (i.e. one’s/something’s capacity for survival) is intelligence and that 2) survival is a metric of intelligence, then you must also hold that 3) intelligence is a metric of intelligence, which is clearly not something you intended to imply.

Your second major thesis (challenge) is that

“2: There are a number of examples of vastly more intelligent systems (in terms of survival) than human intelligence”.

One of your objectives in this article is apparently to expose the ‘fact’ that

“The primacy of human intelligence is one of the last and greatest myths of the anthropomorphic divide -- the division between humans and all other (living) things. Like most fallacies, it provides careers and countless treatises regarding paradoxes that can be explored at great length, leading to the warm and fuzzy conclusion that the human is still on top.”

Here you explicitly assert that the claim that human intelligence stands in a position of primacy relative to the intelligence of other (living) things is a fallacy. But the claim that human intelligence stands in a position of primacy relative to other (living) things is only made to be fallacious if you use the term ‘intelligence’ in two different ways; in other words, this claim is only fallacious if you equivocate in your usage of the term ‘intelligence’ when making this claim. But those who ‘make careers’ based upon this claim are not necessarily equivocating in their usage of the term ‘intelligence’, but it is clear that you are equivocating in your usage and interpretation of the term ‘intelligence’, and thus it is clear that your claims (rather than theirs) are fallacious.

As I stated previously, the greatest danger of redefining a term and then claiming to have arrived at your redefinition through empirical means is that you will make empty or erroneous statements. I have already pointed out two reasoning errors made in this article. But your two major theses (challenges):

“1: Survival is a far better metric of intelligence than replicating human intelligence, and...

2: There are a number of examples of vastly more intelligent systems (in terms of survival) than human intelligence”

are largely empty statements, although these statements are not without their errors either.

Challenge 1 says that determining the relative capacity of any non-human thing to survive is a far better metric of that thing’s intelligence than is comparing its intelligence to human intelligence. We have already gone over the erroneous conclusion that ‘intelligence is a metric of intelligence’ that is necessarily implied by this statement in combination with the statement that “survival is intelligence”. But Challenge 1 is empty in the sense that you are merely begging the question: your reasoning is circular and thus yields no new knowledge. First (logically first) you define ‘intelligence’ as ‘survival’ (loosely, as the relative ability of something or someone to survive) and then you assert that “survival is a far better metric of intelligence than replicating human intelligence”. This proves nothing. Challenge 1 is essentially a restatement of your pre-stipulated definition of intelligence. Further, you cannot supply any amount of evidence that will be sufficient to prove this thesis to be conceptually sound. Whether or not the ability of something to survive is in fact a far better metric of intelligence than the extent to which something’s intelligence matches up to human intelligence depends upon your definition of ‘survival’ and ‘intelligence’. This is a conceptual and not an empirical matter. To show this more clearly, consider the fact that I can restate Challenge 1 in the following way with relatively little loss in overall meaning: “ ‘Fitness’ is a far better synonym of intelligence (as this term is applied to non-human things) than is ‘human-intelligence equivalence’ ”.

Challenge 2 also begs the question and thus yields no new knowledge. You would be hard-pressed to find any researcher of intelligence who would categorically refute the assertion that there are a number of entities that have better survival capabilities than humans. But this is all that you are essentially saying in Challenge 2, while at the same time you are implicitly claiming that this statement helps to expose the ‘fact’ that

“the primacy of human intelligence is one of the last and greatest myths of the anthropomorphic divide”.

Again, we are dealing with a conceptual matter rather than an empirical matter here. Certainly you can attempt to demonstrate empirically that there exist a number of systems that are vastly superior to humans in terms of their survival capabilities, and you may even succeed in doing so. But you cannot prove by means of empirical data that the fact that there exist a number of systems that are vastly superior to humans in terms of their survival capabilities in any way logically implies that the term ‘intelligence’ ought to therefore be expanded to include ‘survival capability’. But if your thesis is merely the claim that there exist a number of systems that have superior survival capabilities in comparison to humans, then this bears no relation to

“the primacy of human intelligence”

as

“one of the last and greatest myths of the anthropomorphic divide”,

since your ability to effectively attack this ‘great myth’ via your thesis depends upon your ability to demonstrate, via your thesis, the fallaciousness of the claim that human intelligence stands in a position of primacy relative to the intelligence of other (living) things. And, to reiterate, the assertion that human intelligence stands in a position of primacy relative to the intelligence of other (living) things only becomes fallacious when you equivocate in your usage of the term ‘intelligence’, which is something that you do in this article, while it is not necessarily something that advocates of this claim do.

Since you characterize your two main theses as “challenges” to the claim of the “primacy of human intelligence”, all of the aforementioned points are critical to the overall success of your paper.

Even more disturbing... in order to "stop" the President of the United States (i.e. remove him/her from office), requires a majority of both the Senate and House of Representatives, or 285 people, voting in favor. I think most people would agree that we've had at least some presidents who aren't 2 orders of magnitude smarter than the average human...

~~

It is interesting that this word "intelligence" causes so much contention. I wrote about a debate about plant intelligence in an article to published in a volume of scholarship someday - Here it is, with creative commons license 2.5
http://www.box.net/shared/h0nqb1c6o0

I think the survivability metric is a good one for building alife systems, and certainly not a bad one for growing other living systems. Of course Darwinian evolution includes sexual selection, so we must do more than survive - we must seduce, intrigue and leave progeny. But that takes nothing way from survivability as working metric for evaluating intelligent systems - it merely contextualizes it. Metrics are tools for helping us achieve results, not definitions of our ontology.

So how does this redefinition help you actually accomplish the goal of creating a system that is intelligent in the way most of us define it, as in, it can understand and solve new problems?

And lets be real here - I could probably kill Stephen Hawking with my bare hands, but there's no way anyone in the world would say I'm more intelligent than he is. Your new definition seems kind of misplaced.

"1: Survival is a far better metric of intelligence than replicating human intelligence, and..."

Unfortunately that's a principle in science, put a bad assumption first and all the rest of the reasoning is biased/wrong.

Of course Intelligence is not evident to define and specially when it applies to human intelligence and the way to measure it. But, the word itself comes from latin "intellegentia": the ability to understand/comprehend...

Unfortunately for your theories, yes, we, as the human race, are the most intelligent specy/entity (known to us at least). Only because we are the only ones to be in the position of trying to define what "Intelligence" is and discuss about it.

"These challenges have not come through a priori philosophical posturing but are the result of years of simulation and the iterative understanding and discourse that has come through the artificial life community."

This is a very funny statement. Artificial life/intelligence community, as you name it, has struggled (so far) to even get close to the smallest sign of intelligence. Sounds like some are tempted by something like "hey, So why not redefine intelligence to something less challenging we can probably achieve : let's make it the ability of systems to survive, yeah great..."

"The primacy of human intelligence is one of the last and greatest myths of the anthropomorphic divide -- the division between humans and all other (living) things. Like most fallacies, it provides careers and countless treatises regarding paradoxes that can be explored at great length, leading to the warm and fuzzy conclusion that the human is still on top. If only it were so."

To the contrary of what you're saying, yes, this is a pseudo-philosophical posturing and in-line with some "concerns" of the moment like all the problems raised by uncontrolled human development BUT the human species is the first ever in the history of Earth to actually realize, thanks to its intelligence, that its own development could lead to mass extinctions (and even its own) and the first to actually care about other species.

So even using your own criterias, we might be, one of the most resilient species ever because we can ACT on our destiny more than any other.
And if we fail, it will make space for others ! ;)

Sorry for the approximate English and all the best from France :.

I'll finish with a citation of Carl Sagan : "We are a way for the universe to know itself

Interesting premise but faulty logic. In the examples you gave you indicated that a road system could be brought down by a number of human obstructions, and a court system could be blocked by a number of human cases. In each of these cases the relevant quantity is obstructions and cases, not humans. It is fallacious to then substitute humans in for either of these terms in order to continue the argument.

Very interesting. I'm not so sure the property you ascribe the system should be labeled intelligence though. It sounds like the modeling you're doing shows the resiliency of the system. Calling this property intelligence is tantamount to claiming that these systems arose through some sort of design. However, if you study the development of these systems, you see that rather than just arising out of design, there are unintended consequences that play themselves out as these systems evolve.

These models would be great to use to determine how much stress a particular system can handle, especially in regards to natural disasters and other mass casualty situations. These tools would also be greatly helpful in determining weaknesses of systems like power, communications, oil delivery, transportation systems, etc. Countermeasures could then be put in place to better meet those stresses and strengthen the system.

I also think you nailed it when you state that survival is the defining trait of successful evolution. I'm not so sure that should be defined as intelligence though. Intelligence accelerated as a response to changing climactic conditions early in the development of homo sapiens. This allowed early humans to adapt to their environment more successfully than other organisms. Some of the consequences of that development have been the rise of towns, cities, civilization, etc. Yet, initially, the development of intelligence allowed the humans who developed those traits to prosper and survive. In that respect intelligence is an effect, not a cause.

"There is a lack of scholarship in this area."

Really? A lack of scholarship in defining and philosophizing about intelligence? Please don't forget our ENTIRE WESTERN PHILOSOPHICAL HISTORY, starting with the Presocratics, through Plato and Aristotle all the way to Foucault, Derrida and Zizek and still coming. There has been endless theorizing about intelligence (and systems - see Rousseau and Spinoza, for a start) in our history that it would be a "fallacy" to ignore all of it! I'm a computer scientist but way too often become annoyed with the way CS scholars try to reinvent everything, ahistorically, here and now, using their cool new tools (toys?).