Well isn't that pretty obvious from interacting with e.g. chatGPT and Bard?
Or maybe the issue is a lot of people don't understand how smart dogs are?
They have an ability to recombine text they've ingested into coherent-sounding sentences and paragraphs that initially seems uncanny to most people, the first time they see it, but it only takes an hour or so to realize they output utter nonsense a lot of the time, unless you ask them about things that many actual people have written about, and which most of those people haven't been wrong or lying about.
If I had to compare the current crop of LLMs to a living creature, it would be a lot closer to a colony of ants. Amazing ability to do certain things, but not a lot of that nebulous thing we call "intelligence".
Of course they're not... Dogs are one of the smartest creatures we know of. Dogs and humans might as well be considered the same level of intelligence compared to something like an LLM.
It would be more interesting to hear the discussion of a more realistic comparison. What about a fly? A Venus fly trap? A dandelion? A river flowing downhill?
Those feel more in the absolute best case neighborhood when considering the potential "intelligence" of the current state of AI.
Our understanding of what intelligence even is is so limited we barely have the right words, or "thinking tools" as Daniel Dennet calls them, to even have this discussion.
The problem is that "smart" is overly broad for this discussion. We need to be much more precise. Clearly some LLMs have far greater reasoning ability or poetry writing skills than dogs.
But also LLMs don't possess many types of cognition that are fundamental abilities of animals. Such as visual/spatial modalities (although some models are multimodal -- still not in the same way animals are).
But none of that means that LLMs can't be extremely useful.
What I find most fascinating about LLMs isn’t that they are smart but that they are different from the smarts I’m used to. It’s not that smart but it knows far more than any human ever could. It’s not smart, but it is well-read. It has read more than anyone and it makes up for its lack of smarts with that, making it a different sort of intelligence, if you can even call it that.
5 comments
[ 5.2 ms ] story [ 24.6 ms ] threadOr maybe the issue is a lot of people don't understand how smart dogs are?
They have an ability to recombine text they've ingested into coherent-sounding sentences and paragraphs that initially seems uncanny to most people, the first time they see it, but it only takes an hour or so to realize they output utter nonsense a lot of the time, unless you ask them about things that many actual people have written about, and which most of those people haven't been wrong or lying about.
If I had to compare the current crop of LLMs to a living creature, it would be a lot closer to a colony of ants. Amazing ability to do certain things, but not a lot of that nebulous thing we call "intelligence".
It would be more interesting to hear the discussion of a more realistic comparison. What about a fly? A Venus fly trap? A dandelion? A river flowing downhill?
Those feel more in the absolute best case neighborhood when considering the potential "intelligence" of the current state of AI.
Our understanding of what intelligence even is is so limited we barely have the right words, or "thinking tools" as Daniel Dennet calls them, to even have this discussion.
Is a computer that can multiply two really big numbers smarter than a dog? Or a human for that matter? Because we sure can't do that.
But also LLMs don't possess many types of cognition that are fundamental abilities of animals. Such as visual/spatial modalities (although some models are multimodal -- still not in the same way animals are).
But none of that means that LLMs can't be extremely useful.