A large language model is giving you the statistically most likely words to your prompt, weighted by prettymuch everything written online. You would describe that as “intelligent”?
Intelligence is a description of capability, not the means by which the capability is achieved. So if the output looks intelligent then the process is intelligent regardless of how it works. The difference between natural and artificial intelligence is how the intelligence is achieved - what you’re describing doesn’t match any intelligence found in nature so if it produces intelligent output then it’s artificially intelligent.
So if the output looks intelligent then the process is intelligent regardless of how it works
Most every other computer program made will also meet this definition. Hell, this definition is so loose, you can use it to describe evolution as an intelligent process.
What if I push it one step further back the chain? In so far as these programs are recognizably intelligent, it is only because conscious people put a lot of time and work into making them. Into setting up the systems that statistically weighed the models. The model is intelligent, but it’s not artificially so. It’s just an expression of plain old human intelligence, obfuscated through sci-fi terminology.
Normal computer programs only look intelligent in very narrow areas, like number crunching, which is why we don’t tend to call them intelligent. Their general intelligence is next to zero. Even if we were delusional enough to think life came from non-life and developed intelligence by random chance and natural selection, you have the same thing there where you get much more non-intelligent output than intelligent output. Monkeys on typewriters could also look intelligent some of the time, but looking at the totality of output they wouldn’t.
It’s just an expression of plain old human intelligence
All artificial things are expressions of human behaviors. That’s kind of the definition of artificial.
I think what this conversation is dancing around is that we have a colloquial definition of “Artificial Intelligence”, inherited from science fiction and broadly (albeit with much specific variation) understood as “conscious machine”…
…and then we have a set of computer programs that are - fundamentally - no different from any other program (one Turing machine can, in principle, run all the same algorithms as any other Turing machine). Yes, we can technically describe these programs with the words “artificial” and “intelligent”, but doing so is kind of disingenuous, given the cultural association that predates any use of the term in comp sci fields.
totally different conversation, but its also a fun one:
Even if we were delusional enough to think life came from non-life and developed intelligence by random chance and natural selection
yeah I won’t discount the possibility that life has other origins. At the same time, you gotta deal with Occam’s Razor: working with what we currently know about the history of the planet, life emerging from non-life requires fewer assumptions. It also cannot be discounted as a possibility.
We don’t have a mathematical definition for intelligence or artificial like we do for a Turing machine, but most useful concepts don’t have precise definitions, like human, air, porn, drugs, medicine. I don’t see how there is anything disingenuous about calling computer programs artificially intelligent because that same term describes sci-fi computer programs which display intelligence. A lot of robots in sci-fi aren’t superintelligent and they usually have robotic voices and vocabulary choices and sometimes difficulty understanding words that aren’t in the dictionary or answering slightly vague questions, things which LLMs have no trouble with.
Saying the mount Rushmore carvings was formed by erosion is a simple explanation with few assumptions that is possible if you assume erosion happens sufficiently randomly, but it’s an absurd explanation nonetheless. Occam’s razor is just a rule of thumb to apply to competing explanations which are similarly reasonable and even then it’s not telling you which is more likely to be true but which is the simplest model to work with.
I don’t see how there is anything disingenuous about calling computer programs artificially intelligent because that same term describes sci-fi computer programs which display intelligence.
In a vacuum, sure. But in the context where a small number of companies are circulating billions of dollars between each other and have a vested interest in overstating what their product is capable of? Very disingenuous. It’s the big financial bubble of the 21st century. We can have a conversation where we ignore the cultural and historical context, but frankly its not an interesting conversation.
I don’t think the Rushmore example… works? You’re trying to make an example where Occam Razor points to something we obviously know is wrong, but you’re first assuming a lack of context. Erosion is not the simplest explanation given everything we already know. I know Occam’s Razor is a rule of thumb. It doesn’t dismiss any other origin for life. But the same time, you cannot dismiss life emerging out of physics and chemistry just because you feel it is absurd.
I mean, you can dismiss it. Follow you gut and stick to what you believe. Quite a lot of good scientists have made important discoveries doing just that! But more than a gut feeling is needed to actually convince anyone else.
So would you describe “statistically likely” to be the same as “intelligent”?
Statistically likely to…what? I don’t understand what you’re trying to say.
A large language model is giving you the statistically most likely words to your prompt, weighted by prettymuch everything written online. You would describe that as “intelligent”?
Intelligence is a description of capability, not the means by which the capability is achieved. So if the output looks intelligent then the process is intelligent regardless of how it works. The difference between natural and artificial intelligence is how the intelligence is achieved - what you’re describing doesn’t match any intelligence found in nature so if it produces intelligent output then it’s artificially intelligent.
Most every other computer program made will also meet this definition. Hell, this definition is so loose, you can use it to describe evolution as an intelligent process.
What if I push it one step further back the chain? In so far as these programs are recognizably intelligent, it is only because conscious people put a lot of time and work into making them. Into setting up the systems that statistically weighed the models. The model is intelligent, but it’s not artificially so. It’s just an expression of plain old human intelligence, obfuscated through sci-fi terminology.
Normal computer programs only look intelligent in very narrow areas, like number crunching, which is why we don’t tend to call them intelligent. Their general intelligence is next to zero. Even if we were delusional enough to think life came from non-life and developed intelligence by random chance and natural selection, you have the same thing there where you get much more non-intelligent output than intelligent output. Monkeys on typewriters could also look intelligent some of the time, but looking at the totality of output they wouldn’t.
All artificial things are expressions of human behaviors. That’s kind of the definition of artificial.
I think what this conversation is dancing around is that we have a colloquial definition of “Artificial Intelligence”, inherited from science fiction and broadly (albeit with much specific variation) understood as “conscious machine”…
…and then we have a set of computer programs that are - fundamentally - no different from any other program (one Turing machine can, in principle, run all the same algorithms as any other Turing machine). Yes, we can technically describe these programs with the words “artificial” and “intelligent”, but doing so is kind of disingenuous, given the cultural association that predates any use of the term in comp sci fields.
totally different conversation, but its also a fun one:
yeah I won’t discount the possibility that life has other origins. At the same time, you gotta deal with Occam’s Razor: working with what we currently know about the history of the planet, life emerging from non-life requires fewer assumptions. It also cannot be discounted as a possibility.
We don’t have a mathematical definition for intelligence or artificial like we do for a Turing machine, but most useful concepts don’t have precise definitions, like human, air, porn, drugs, medicine. I don’t see how there is anything disingenuous about calling computer programs artificially intelligent because that same term describes sci-fi computer programs which display intelligence. A lot of robots in sci-fi aren’t superintelligent and they usually have robotic voices and vocabulary choices and sometimes difficulty understanding words that aren’t in the dictionary or answering slightly vague questions, things which LLMs have no trouble with.
Saying the mount Rushmore carvings was formed by erosion is a simple explanation with few assumptions that is possible if you assume erosion happens sufficiently randomly, but it’s an absurd explanation nonetheless. Occam’s razor is just a rule of thumb to apply to competing explanations which are similarly reasonable and even then it’s not telling you which is more likely to be true but which is the simplest model to work with.
In a vacuum, sure. But in the context where a small number of companies are circulating billions of dollars between each other and have a vested interest in overstating what their product is capable of? Very disingenuous. It’s the big financial bubble of the 21st century. We can have a conversation where we ignore the cultural and historical context, but frankly its not an interesting conversation.
I don’t think the Rushmore example… works? You’re trying to make an example where Occam Razor points to something we obviously know is wrong, but you’re first assuming a lack of context. Erosion is not the simplest explanation given everything we already know. I know Occam’s Razor is a rule of thumb. It doesn’t dismiss any other origin for life. But the same time, you cannot dismiss life emerging out of physics and chemistry just because you feel it is absurd.
I mean, you can dismiss it. Follow you gut and stick to what you believe. Quite a lot of good scientists have made important discoveries doing just that! But more than a gut feeling is needed to actually convince anyone else.