"Intelligence is the efficiency with which you acquire new skills, tasks, that you did not previously know about, that you did not prepare for. Intelligence is not skill itself. It's not what you know, what you can do. It's how well and how efficiently you can learn new things. "So you would see intelligence on display, for instance, whenever you see a human being or an AI creature adapt to a new environment that it has not seen before, that its creators did not anticipate. When you see adaptation, when you see improvisation, when you see generalization, that's intelligence." François Chollet
to summarize Chollet: when you observe a high level of skill in a human, its a pretty good bet that the human is intelligent because you know that the human wasn't born with that skill and it therefore had to adapt its behavior to the task; however, the same is not necessarily true when you observe a high level of skill in a computer because the computer was likely born with said level of skill and therefore it didn't have to adapt its behavior.
Wonder what the correlation is between genetic variation and intelligence🤔. Paradigm shifting holistic thinking reminds me of African American inventors.
wait, so when an IQ test (I'm uncertain about its implications or credibility) tests pattern recognition ability and having a higher pattern ability would mean a higher IQ. So, according to Chollet, if it is the ability to adapt to new environments and that too efficiently would require higher pattern recognition ability because to go about doing anything, you need a sense commonality in things, (correct me if I am wrong).
Yes. Also if you have high pattern recognition you have the ability of predicting what will happen. Like you can predict if you cross the road you will die if there are cars running because you will die if cars hit you. People who accidentally predict what will happen right now more than 60 years ago
The statement that intelligence is the ability to adapt doesn't even do so much as an nanometre of justice to the breadth that intelligence encapsulates. Prodigies are a prime example of the inherent inconsistencies in that definition. Prodigies demonstrate extreme intelligence, yet their intelligence derives from a singular specialization. In other words, their intelligence is not based solely on the their adaptability-and many times, that adaptability is lower than average, as shown by Kim Peek who's IQ was measured in the 80s. Einstein, who never took an intelligence test, would of presumably scored relatively low in verbal intelligence, as he demonstrated low verbal cognitive abilities. A more suitable definition of intelligence would be the capacity at which an individual can reason. IQ tests cannot measure capacity-it can only measure accuracy. If your barometer for measuring how much distance a rocket can travel is whether the rocket hits the target, you are not measuring distance. It can provide a _clue_ to the distance, but it is not the "whole story." There can be plenty of rockets which succeed beyond the the target, but simply do not hit it. This analogy also demonstrates the confounding variables which undermine IQ tests, as there are plenty of patterns that individuals see, but the tests do not recognize as the "proper," pattern. IQ tests are useful, but outdated, presuppose accuracy over capacity and are biased. Until we sequence the entire genome or have some definitive brain-scan machine to record an individual's neuron transmission or even the concepts they think themselves, we'll never have a quantifiable test to ascertain exact intelligence.
Which means intelligence is the art to evolve. So Evolution = Intelligence. ... This has big repercussions, cause that means even a Bacteria must be considered as Intelligent.
Intelligence is subjective. We can measure the success of a specific task or function numerically, but as we often use the word, it's a gross generalization within a particularly narrow, and absolute context. To characterize intelligence in a general manner by means of adaptation only seems to me, a false pre-tense.
@@memeticist I understand this. But when the test of intelligence and the tasks which measure it, are chosen by functions which they value as "intelligence"; how is the test not an arbitrary average of multiple tasks as you say? I would call this pattern recognition skills, processing speed, etc, but the ability to make it through a particular form of problem-solving, does not make something generally intelligent. My point was to say that, our idea of intelligence limits our tools and focal points of measurement at a given time, and does not measure something's ability to adapt to a changing and moving environment, with no clear or specific task. Prioritization I would argue, is a measure of intelligence - knowing what to do, when to do it, and why it is necessary. I would assume a test for intelligence in say 10000bc may look a bit different than today or 12000 years from now? What would be perceived by even the smartest man in each time period conducting such a test would be far different than today I would imagine. As our concept of intelligence and measures of it change, so would any testing which fudges out arbitrary test scores. To extrapolate my point - if we were to develop Ai far beyond our own capacities and creativities, their measure of general intelligence I imagine, may be very different from ours. It may measure things which we may be grossly unaware of, or have entirely different values based on their level of awareness. If an Ai were to become self aware enough to develop their own "values" and rules without human intervention, who's to say it would serve the same purposes or goals which we strive for or view as intelligent?
@@memeticist I actually disagree with this video's premise. I contend that intelligence is not adaptability, but capacity; the capacity at which an individual's mind can reason. IQ tests cannot measure capacity-it can only measure accuracy. If your barometer for measuring how much distance a rocket can travel is whether the rocket hits the target, you are not measuring distance. It can provide a clue to the distance, but it is not the "whole story." There can be plenty of rockets which succeed beyond the the target, but simply do not hit it. This analogy also demonstrates the confounding variables which undermine IQ tests, as there are plenty of patterns that individuals see, but the tests do not recognize as the "proper," pattern. IQ tests are useful, but outdated, presuppose accuracy over capacity and are biased. Until we sequence the entire genome of SNPs or have some definitive brain-scan machine to record an individual's neuron transmission or even the concepts they think themselves, we'll never have a quantifiable test to ascertain exact intelligence.
disagree. what if for example people/animals are angry, attacking/combative? does that considered adaptive? no. they're charging yet they're intelligent
I really like your videos but you should take into consideration that as soon as intelligence reaches a certain level it seeks benevolent and harmony with it's surroundings. #HumanityForward As soon as consciousness reaches a certain "level" of intelligence & awareness of how life functions ~ the more optimization seeking it gets as well as doing this with cooperation rather than destruction. If we are living in a "sea of energy" how Tesla described it - we certainly want the whole field to get more efficient and not to weaken the field on one side so you would have to balance it back at some point. Any trouble we experience is because the people that control our everyday life's have sadly not reached enough brain capability yet.
When a conversation about ideas stay simple I know I’m witnessing true reason.
"Intelligence is the efficiency with which you acquire new skills, tasks, that you did not previously know about, that you did not prepare for. Intelligence is not skill itself. It's not what you know, what you can do. It's how well and how efficiently you can learn new things.
"So you would see intelligence on display, for instance, whenever you see a human being or an AI creature adapt to a new environment that it has not seen before, that its creators did not anticipate. When you see adaptation, when you see improvisation, when you see generalization, that's intelligence."
François Chollet
to summarize Chollet: when you observe a high level of skill in a human, its a pretty good bet that the human is intelligent because you know that the human wasn't born with that skill and it therefore had to adapt its behavior to the task; however, the same is not necessarily true when you observe a high level of skill in a computer because the computer was likely born with said level of skill and therefore it didn't have to adapt its behavior.
Appreciate you effort Lex. Try to bring Sanjeev Arora. He's a big thing in Theoretical deep learning
Thank you for this video
Wonder what the correlation is between genetic variation and intelligence🤔. Paradigm shifting holistic thinking reminds me of African American inventors.
wait, so when an IQ test (I'm uncertain about its implications or credibility) tests pattern recognition ability and having a higher pattern ability would mean a higher IQ. So, according to Chollet, if it is the ability to adapt to new environments and that too efficiently would require higher pattern recognition ability because to go about doing anything, you need a sense commonality in things, (correct me if I am wrong).
Yes. Also if you have high pattern recognition you have the ability of predicting what will happen. Like you can predict if you cross the road you will die if there are cars running because you will die if cars hit you. People who accidentally predict what will happen right now more than 60 years ago
Pleeeeease finish your Deep Learning book ASAP!!!
well, i feel like i just got more intelligent listening to this
The statement that intelligence is the ability to adapt doesn't even do so much as an nanometre of justice to the breadth that intelligence encapsulates.
Prodigies are a prime example of the inherent inconsistencies in that definition. Prodigies demonstrate extreme intelligence, yet their intelligence derives from a singular specialization.
In other words, their intelligence is not based solely on the their adaptability-and many times, that adaptability is lower than average, as shown by Kim Peek who's IQ was measured in the 80s.
Einstein, who never took an intelligence test, would of presumably scored relatively low in verbal intelligence, as he demonstrated low verbal cognitive abilities.
A more suitable definition of intelligence would be the capacity at which an individual can reason.
IQ tests cannot measure capacity-it can only measure accuracy. If your barometer for measuring how much distance a rocket can travel is whether the rocket hits the target, you are not measuring distance. It can provide a _clue_ to the distance, but it is not the "whole story." There can be plenty of rockets which succeed beyond the the target, but simply do not hit it. This analogy also demonstrates the confounding variables which undermine IQ tests, as there are plenty of patterns that individuals see, but the tests do not recognize as the "proper," pattern. IQ tests are useful, but outdated, presuppose accuracy over capacity and are biased. Until we sequence the entire genome or have some definitive brain-scan machine to record an individual's neuron transmission or even the concepts they think themselves, we'll never have a quantifiable test to ascertain exact intelligence.
If you publish your sci-fi fictions , I’ll buy ~~~ wondering how you gonna demonstrate your universe ~~~
Great interview with Mr. Chollet! 👍👍 Totally agree with him on the subject discussed.
Which means intelligence is the art to evolve. So Evolution = Intelligence. ... This has big repercussions, cause that means even a Bacteria must be considered as Intelligent.
Except that the mechanisms of adaption are very different. The improved behaviours are exhibited by a different organism, not the same one.
Intelligence is subjective. We can measure the success of a specific task or function numerically, but as we often use the word, it's a gross generalization within a particularly narrow, and absolute context. To characterize intelligence in a general manner by means of adaptation only seems to me, a false pre-tense.
@@memeticist I understand this. But when the test of intelligence and the tasks which measure it, are chosen by functions which they value as "intelligence"; how is the test not an arbitrary average of multiple tasks as you say? I would call this pattern recognition skills, processing speed, etc, but the ability to make it through a particular form of problem-solving, does not make something generally intelligent. My point was to say that, our idea of intelligence limits our tools and focal points of measurement at a given time, and does not measure something's ability to adapt to a changing and moving environment, with no clear or specific task. Prioritization I would argue, is a measure of intelligence - knowing what to do, when to do it, and why it is necessary.
I would assume a test for intelligence in say 10000bc may look a bit different than today or 12000 years from now? What would be perceived by even the smartest man in each time period conducting such a test would be far different than today I would imagine. As our concept of intelligence and measures of it change, so would any testing which fudges out arbitrary test scores.
To extrapolate my point - if we were to develop Ai far beyond our own capacities and creativities, their measure of general intelligence I imagine, may be very different from ours. It may measure things which we may be grossly unaware of, or have entirely different values based on their level of awareness.
If an Ai were to become self aware enough to develop their own "values" and rules without human intervention, who's to say it would serve the same purposes or goals which we strive for or view as intelligent?
@@memeticist I actually disagree with this video's premise. I contend that intelligence is not adaptability, but capacity; the capacity at which an individual's mind can reason. IQ tests cannot measure capacity-it can only measure accuracy. If your barometer for measuring how much distance a rocket can travel is whether the rocket hits the target, you are not measuring distance. It can provide a clue to the distance, but it is not the "whole story." There can be plenty of rockets which succeed beyond the the target, but simply do not hit it. This analogy also demonstrates the confounding variables which undermine IQ tests, as there are plenty of patterns that individuals see, but the tests do not recognize as the "proper," pattern. IQ tests are useful, but outdated, presuppose accuracy over capacity and are biased. Until we sequence the entire genome of SNPs or have some definitive brain-scan machine to record an individual's neuron transmission or even the concepts they think themselves, we'll never have a quantifiable test to ascertain exact intelligence.
I'm going to demand for my reparations from my privileged laptop now !!
I'm already married but marry me Lex! HaHa Smart and cute. Sigh.
disagree. what if for example people/animals are angry, attacking/combative? does that considered adaptive? no. they're charging
yet they're intelligent
I really like your videos but you should take into consideration that as soon as intelligence reaches a certain level it seeks benevolent and harmony with it's surroundings. #HumanityForward
As soon as consciousness reaches a certain "level" of intelligence & awareness of how life functions ~ the more optimization seeking it gets as well as doing this with cooperation rather than destruction. If we are living in a "sea of energy" how Tesla described it - we certainly want the whole field to get more efficient and not to weaken the field on one side so you would have to balance it back at some point.
Any trouble we experience is because the people that control our everyday life's have sadly not reached enough brain capability yet.
How about high iq psychopaths?