By Jason M. Pittman, Sc. D.
March 26, 2018
In our last post, we discussed why the field of synthetic intelligence is essential to the future of our species. Before we move along too far, I think it is crucial to outline just what we mean by intelligence or being intelligent. In doing so, we can address fundamental questions such as what is intelligence, how do we measure intelligence, and also are we the only intelligent life.
Intelligence can be a controversial topic. Paradoxically, although we often talk openly about possessing intelligence, or about others possessing more or less intelligence than us, we are often hesitant to discuss ways of measuring intelligence.
In the present discussion, we are not concerned with how intelligent an organism is. Our concern is with a given organism's capacity for certain qualities, in relation to its environment. That’s a funny quality of intelligence; intelligence must be acted out in the situation for us to know that intelligence exists.
In keeping with my prior claims, I suggest that intelligence is not imitation. Further, intelligence is not the result of computation but can perform computation. Indeed, artificial intelligence agents fall short in this regard; such systems are imitative. For example, machine learning is a computational function -- beautiful and complex but not indicative of intelligence. No matter how real a video game agent may appear, the game agent’s behavior is merely a programmed simulation, a hoax.
The question remains, then: what is intelligence? As far as we can tell, intelligence is considered to be present when an organism can (a) learn; (b) reason about what is learned; and (c) enlist learning and reasoning as a means to solve problems.
How do we measure intelligence?
First, don't say an IQ test. Second, don't say an IQ test. Remember, we're concerned with determining the presence of intelligence. An IQ test quantifies the amount of intelligence present so to speak. We ought to be careful about putting the cart after the horse if we want to make progress.
Now, intelligence has another odd quality in that it is not directly observable. While many traits -- hair color, cellular activity, language, and so forth -- are directly observable by us in others, intelligence remains trapped exclusively within. However, we have become exceedingly proficient at measuring the expression of human intelligence in manners consistent with indirect observation. Think about how we observe distant celestial objects; not directly but through inference based on surrounding evidence. Still, our proficiency has limitations.
Foremost, the instrumentation we employ for humans does not generalize to non-humans. G factor, as a psychometric instrumentation, is an example. While g factor works for humans and some other mammals, there is no reliable, equivalent instrumentation outside of those mammalian models. Does that mean we are the only intelligent life?
Are we the only intelligent life?
The simple answer is, no. However, affirming a non-human intelligence is difficult in that there is an anthropocentric bias in play. Because we appear to be the only intelligent life form, we assume we are in fact the only intelligent life form. Such is far from the truth, however.
We, of course, know that some primates other than homo sapiens exhibit intelligence. As well, there are higher order mammals such as dolphins and whales that also appear to be intelligent. The recognition of intelligence in these organisms is straightforward because their intelligence mirrors ours. More troublesome is intelligence that differs drastically.
Plants are an example. Plants appear to be closer in intellect formation to swarm intelligence in some ways, to multicellular life in others. Further, plants exist on an entirely different time scale than humans. Colonizing organisms -- bees and ants are great examples -- display intelligence as a collective. Also, life, as we perceive such, is carbon-based. In contrast, we know that other elements such as siliconcan potentially form the necessary molecular basis for life.
Thus, building a synthetic intelligence is a tricky problem. Starting with higher order intelligence is misguided at best, perhaps impossible. In part, I suggest that is why artificial intelligence is not the path forward if we are sincerely interested in creating intelligence rather than imitating intelligence. Existing, alternative forms of intelligence, such as what we can indirectly observe in plant life, perhaps is a more tenable endeavor.
Yes, let’s start there. Along the way, we can explore by-product of intelligence that may be a new way to measure more complex forms of intelligence such as synthesized animal intelligence: agency. Tune in for the next post in two weeks to find out how I think agency is a novel measure!
March 26, 2018
In our last post, we discussed why the field of synthetic intelligence is essential to the future of our species. Before we move along too far, I think it is crucial to outline just what we mean by intelligence or being intelligent. In doing so, we can address fundamental questions such as what is intelligence, how do we measure intelligence, and also are we the only intelligent life.
Intelligence can be a controversial topic. Paradoxically, although we often talk openly about possessing intelligence, or about others possessing more or less intelligence than us, we are often hesitant to discuss ways of measuring intelligence.
In the present discussion, we are not concerned with how intelligent an organism is. Our concern is with a given organism's capacity for certain qualities, in relation to its environment. That’s a funny quality of intelligence; intelligence must be acted out in the situation for us to know that intelligence exists.
In keeping with my prior claims, I suggest that intelligence is not imitation. Further, intelligence is not the result of computation but can perform computation. Indeed, artificial intelligence agents fall short in this regard; such systems are imitative. For example, machine learning is a computational function -- beautiful and complex but not indicative of intelligence. No matter how real a video game agent may appear, the game agent’s behavior is merely a programmed simulation, a hoax.
The question remains, then: what is intelligence? As far as we can tell, intelligence is considered to be present when an organism can (a) learn; (b) reason about what is learned; and (c) enlist learning and reasoning as a means to solve problems.
How do we measure intelligence?
First, don't say an IQ test. Second, don't say an IQ test. Remember, we're concerned with determining the presence of intelligence. An IQ test quantifies the amount of intelligence present so to speak. We ought to be careful about putting the cart after the horse if we want to make progress.
Now, intelligence has another odd quality in that it is not directly observable. While many traits -- hair color, cellular activity, language, and so forth -- are directly observable by us in others, intelligence remains trapped exclusively within. However, we have become exceedingly proficient at measuring the expression of human intelligence in manners consistent with indirect observation. Think about how we observe distant celestial objects; not directly but through inference based on surrounding evidence. Still, our proficiency has limitations.
Foremost, the instrumentation we employ for humans does not generalize to non-humans. G factor, as a psychometric instrumentation, is an example. While g factor works for humans and some other mammals, there is no reliable, equivalent instrumentation outside of those mammalian models. Does that mean we are the only intelligent life?
Are we the only intelligent life?
The simple answer is, no. However, affirming a non-human intelligence is difficult in that there is an anthropocentric bias in play. Because we appear to be the only intelligent life form, we assume we are in fact the only intelligent life form. Such is far from the truth, however.
We, of course, know that some primates other than homo sapiens exhibit intelligence. As well, there are higher order mammals such as dolphins and whales that also appear to be intelligent. The recognition of intelligence in these organisms is straightforward because their intelligence mirrors ours. More troublesome is intelligence that differs drastically.
Plants are an example. Plants appear to be closer in intellect formation to swarm intelligence in some ways, to multicellular life in others. Further, plants exist on an entirely different time scale than humans. Colonizing organisms -- bees and ants are great examples -- display intelligence as a collective. Also, life, as we perceive such, is carbon-based. In contrast, we know that other elements such as siliconcan potentially form the necessary molecular basis for life.
Thus, building a synthetic intelligence is a tricky problem. Starting with higher order intelligence is misguided at best, perhaps impossible. In part, I suggest that is why artificial intelligence is not the path forward if we are sincerely interested in creating intelligence rather than imitating intelligence. Existing, alternative forms of intelligence, such as what we can indirectly observe in plant life, perhaps is a more tenable endeavor.
Yes, let’s start there. Along the way, we can explore by-product of intelligence that may be a new way to measure more complex forms of intelligence such as synthesized animal intelligence: agency. Tune in for the next post in two weeks to find out how I think agency is a novel measure!
No comments:
Post a Comment