Cutting snippets of paragraphs out of existing books and pasting them together into a conglomeration that vaguely resembles a novel does not make me an author. Pattern recognition and matching is not original creation.
The original source material was still human-generated. When a computer is able to imagine a totally new concept out of thin air, then I will be impressed.
I’m no defender of AI, but this is a bad argument. We are neurologically wired to detect patterns. It’s how experience works. Intuition. Differential diagnoses.
There’s a decent body of research within Cognitive Science on creativity. Ideas don’t burst into existence from nothing, they develop. The statement that everything is derivative speaks from reality, not just angst. Artists themselves hang out together. The French Impressionists. Hemingway and his peers. Their communities are about more than just not being lonely. Oppenheimer anyone? Scientists think tank for a reason.
Yeah, it’s an uncomfortable thing to admit, but we don’t actually know if it’s true that “real” intelligence works fundamentally differently than modern ML techniques. The argument that generative AI isn’t “real” creativity could still turn out to be cope.
Name any concept that you think was imagined out of thin air. Talk to the creator, and you will almost certainly get a list of other sources that led to it.
These sorts of tests are really tricky to figure out. Just what the hell is intelligence and consciousness and creativity, anyway? The most useful thing AI can do is pin down some tests of what they actually mean.
Though some of the most important work on consciousness is actually coming out of anesthesiology.
That’s why I said “most humans”. If you take an average person, chances are they are unable to produce a song, if not replicate one they have heard. If that person is a musician, if they make an original song it’s likely similar in concept, execution and technique to other songs they have experienced (because human learning is largely, though not entirely, consumption of previous knowledge and retransformation). Only a minuscule minority of people would be able to produce truly novel music, with rules that are not and have never been used before.
Does it mean only a person that is exceptional in a field can be considered human and intelligent?
Don’t get me wrong, I’m not saying that LLM learn in the same ways as humans do (even if the principle is similar) and that there is any " intelligence" in what they make. But plenty of people enjoy AI gemerated content, sometimes without noticing (and AI generated songs are the most likely to be unidentified by the average persone in my opinion).
But the examples picked by the captain are objectively bad arguments to define intelligence and coscence, as we are being clearly demonstrated in these recent years. Current AI models are pretty darn good at transformative art, probably more than the average person, and that ok, just like a car is objectively faster than any person could possibly be, or a robotic arm is infitely more precise.
Not defending AI, but humans also learn from the works of others. So if we want to mimic or recreate our own intelligence it makes sense that it should “learn” from similar material in a similar way.
One thing that’s not similar is that we take time to learn and grow. And through that growth over time we develop connections between these concepts we have learned from those materials. Plus a bunch of other stuff we do that is kissing from the current state of AI. Yet many people don’t understand that or choose to ignore it.
The irony is that nowadays, something that is universally considered non-human is able to do these things, arguably better than the average human.
Cutting snippets of paragraphs out of existing books and pasting them together into a conglomeration that vaguely resembles a novel does not make me an author. Pattern recognition and matching is not original creation.
The original source material was still human-generated. When a computer is able to imagine a totally new concept out of thin air, then I will be impressed.
I’m no defender of AI, but this is a bad argument. We are neurologically wired to detect patterns. It’s how experience works. Intuition. Differential diagnoses.
There’s a decent body of research within Cognitive Science on creativity. Ideas don’t burst into existence from nothing, they develop. The statement that everything is derivative speaks from reality, not just angst. Artists themselves hang out together. The French Impressionists. Hemingway and his peers. Their communities are about more than just not being lonely. Oppenheimer anyone? Scientists think tank for a reason.
Here’s one article to my point: https://pulpfest.com/2022/07/11/influence-or-coincidence-hemingways-fiction-and-hammetts-hardboiled-pulp-2/
Yeah, it’s an uncomfortable thing to admit, but we don’t actually know if it’s true that “real” intelligence works fundamentally differently than modern ML techniques. The argument that generative AI isn’t “real” creativity could still turn out to be cope.
Name any concept that you think was imagined out of thin air. Talk to the creator, and you will almost certainly get a list of other sources that led to it.
These sorts of tests are really tricky to figure out. Just what the hell is intelligence and consciousness and creativity, anyway? The most useful thing AI can do is pin down some tests of what they actually mean.
Though some of the most important work on consciousness is actually coming out of anesthesiology.
Everything is a remix
This but unironically
You might argue that humans are just pattern recognition or matching, just with a wider variety of inputs than the typical LLM
That’s why I said “most humans”. If you take an average person, chances are they are unable to produce a song, if not replicate one they have heard. If that person is a musician, if they make an original song it’s likely similar in concept, execution and technique to other songs they have experienced (because human learning is largely, though not entirely, consumption of previous knowledge and retransformation). Only a minuscule minority of people would be able to produce truly novel music, with rules that are not and have never been used before.
Does it mean only a person that is exceptional in a field can be considered human and intelligent?
Don’t get me wrong, I’m not saying that LLM learn in the same ways as humans do (even if the principle is similar) and that there is any " intelligence" in what they make. But plenty of people enjoy AI gemerated content, sometimes without noticing (and AI generated songs are the most likely to be unidentified by the average persone in my opinion).
But the examples picked by the captain are objectively bad arguments to define intelligence and coscence, as we are being clearly demonstrated in these recent years. Current AI models are pretty darn good at transformative art, probably more than the average person, and that ok, just like a car is objectively faster than any person could possibly be, or a robotic arm is infitely more precise.
Not defending AI, but humans also learn from the works of others. So if we want to mimic or recreate our own intelligence it makes sense that it should “learn” from similar material in a similar way.
One thing that’s not similar is that we take time to learn and grow. And through that growth over time we develop connections between these concepts we have learned from those materials. Plus a bunch of other stuff we do that is kissing from the current state of AI. Yet many people don’t understand that or choose to ignore it.