4 ms·
> ChatGPT is, in fact, incapable of knowing anything at all. this is foundamentally wrong. I've been instructing chat gpt about wargame rules, and it's unders
by avereveard 4y ago
> ChatGPT is, in fact, incapable of knowing anything at all.
this is foundamentally wrong. I've been instructing chat gpt about wargame rules, and it's understanding those rules well enough to run rounds of combat. that is far beyond what this article pretends chatgpt to be (a massive pretrained markov chain) as all that learning is happening after training
heck, you can make chatgpt pretend to be something that isn't, exhibiting creativity that is well beyond previous llm.
> All it knows, all it knows are the statistical relationships of how words appear together
again, you can ask nonsense, and it will not spew nonsense back. it has a cursory understanding of what nonsense is, and will not fall for easy traps (i.e. what year apollo 7 landed on the moon) that would trick straighforward probabilistic models
all in all, this article toned down my respect for the blog down two pegs, if he's so wrong about this topic, how much was he wrong on other topics where I had no understanding and took his word for good?
- plutonorm 4y agoAgree. The guy is in deep cope territory.
- blackbear_ 4y ago> if he's so wrong about this topic This sounds very arrogant. What is your experience grading college essays? (In other words, you are talking about completely differet use-cases)
- avereveard 4y agohis cursory understanding of chatgpt is easily falsiable
- mejutoco 4y agoKnowing is putting in relationship with, IMHO. You know what a table is, by were it usually is placed, differences with a chair, differences with a cat, materials it can have, usage given by people. In summary, how it exists in the context of something else. If we imagine a universe with only one thing (difficult because we would be there too, but bear with me), what could we know about that one thing? It would be the whole universe, all and nothing at the same time. I would like to propose an alternative take on the same: we also are incapable of knowing. We are just putting in relationship one thing with others, in a similar way as ChatGPT, but on a much bigger scale. And that is fine. We are still special: we have sensors (senses), we can move, feel pain, but from the point of view of knowledge we are still comparing things and putting them in relationship to one another.
- ZephyrBlu 4y ago> if he's so wrong about this topic, how much was he wrong on other topics where I had no understanding and took his word for good? No value judgement on this specific case, but this is commonly known as Gell-Mann amnesia: http://www.wikibin.org/articles/gell-mann-amnesia-effect.html http://www.wikibin.org/articles/gell-mann-amnesia-effect.htm...
- BlueTemplar 4y agoEh. Not really ? Because while grading history essays is his domain of expertise, ChatGPT and LLMs are not.
- BlueTemplar 4y agoSo, would you say that the author here is displaying known AI pitfalls : extrapolating nonsense from a too limited corpus of information, overconfidence while being very good at sounding confident ? ;p More seriously, I have so far no reason to believe that he's wrong about neither him nor his students having managed to make ChatGPT produce an even passable essay, so if you want to prove him wrong...