and I wrote to you that it's just your hypothesis "as the traits increase, the results will get worse".
I stated my hypothesis. no one here in MO has tried to test it yet because of the cost of experiments. but I remind you, some people did it for GPT, there was a jump in the quality of connections of heterogeneous information to such an extent that it became possible to create new connections and conclusions.
I wrote to you that this is your hypothesis, I didn't hypothesise. It is rigorously proven.
Don't compare finger to finger, large language models are trained exactly the same way. And you have no idea what the learning error is.
Don't compare finger to finger, large language models are trained exactly the same way. And you have no idea what kind of learning error there is.
What finger to finger?
Exactly, large language models are trained exactly the same way, and they use optimisation algorithms (you can ask GPT what algorithms he was trained with, a few months ago he answered unambiguously, now he's humouring, but I'll just say that ADAM is one of them). and I have no idea what kind of learning error there is, just like you have no idea. The authors are good just because they were able to build evaluation criteria for a large model, which is very difficult, because it is not enough to collect information, you need to be able to correctly evaluate the quality of training (as I said, building evaluation criteria is not less important).
You like to measure yourself. I'm not teaching you, you should understand these things if you think you're a super pro.
You're the one who thinks I'm a super pro and you're off-topic. I don't like blabbering, a mush of unrelated arguments, sprinkled with psychological tricks like citing authorities. If you're too lazy to read the evidence, I can't help you any more.
Maxim Dmitrievsky #: You're the one who thinks I'm a super pro and writes off-topic. I don't like blabbering, a mush of unrelated arguments, sprinkled with psychological tricks, like references to authorities. If you're too lazy to read the evidence, I can't help you any further.
What authorities did I cite? I think you're usually flaunting references to authorities and their articles.
I am not an expert in MO, but I know a lot about optimisation. And since MO is nowhere without optimisation, I can participate in any discussion here without hesitation.
What authorities did I refer to? I think you usually flaunt references to authorities and their articles.
I'm not an expert in MO, but I know a lot about optimisation. And since there is no MO without optimisation, I can participate in any discussion here without hesitation.
You've cited GPT as some kind of proof of something you don't understand. You write for the sake of writing. There is no meaningful message. I'm not interested in optimisation, that's the 3rd question. I did not write about optimisation and did not ask about it. If learning includes optimisation, it doesn't mean that learning is optimisation. That's not what the conversation was about at all.
You just didn't understand what I wrote about and started writing about the sick stuff that is closer to you.
Maxim Dmitrievsky #: You cited GPT as some kind of proof of no idea what. You're just writing for the sake of writing. There is no meaningful message. I'm not interested in optimisation, that's the 3rd question. I did not write about optimisation and did not ask about it. If learning includes optimisation, it doesn't mean that learning is optimisation. That's not what the conversation was about at all.
You didn't write about optimisation, so why are you poking around?
I didn't write to you.
And, learning is a special case of optimisation, remember that.
You see, he didn't write! Huh, so now I have to ask for permission to write about what to write about and what not to write about? Calm down already.
Allow the use of cookies to log in to the MQL5.com website.
Please enable the necessary setting in your browser, otherwise you will not be able to log in.
You are signed in. Two-factor authentication has been disabled. Set it up again and scan a new QR code, or continue with two-factor authentication disabled.
and I wrote to you that it's just your hypothesis "as the traits increase, the results will get worse".
I stated my hypothesis. no one here in MO has tried to test it yet because of the cost of experiments. but I remind you, some people did it for GPT, there was a jump in the quality of connections of heterogeneous information to such an extent that it became possible to create new connections and conclusions.
I wrote to you that this is your hypothesis, I didn't hypothesise. It's strictly proven.
What you said is NOT proven, it is an empirical judgement, therefore your statement is a hypothesis.
I had no questions for you.
what you said is NOT proven, they are empirical judgements, so your statement is a hypothesis.
I had no questions for you.
What finger to finger?
Exactly, large language models are trained exactly the same way, and they use optimisation algorithms (you can ask GPT what algorithms he was trained with, a few months ago he answered unambiguously, now he's humouring, but I'll just say that ADAM is one of them). and I have no idea what kind of learning error there is, just like you have no idea. The authors are good just because they were able to build evaluation criteria for a large model, which is very difficult, because it is not enough to collect information, you need to be able to correctly evaluate the quality of training (as I said, building evaluation criteria is not less important).
You like to measure yourself. I'm not teaching you, you should understand these things if you think you're a super-professional.
You like to measure yourself. I'm not teaching you, you should understand these things if you think you're a super pro.
You're the one who thinks I'm a super pro and writes off-topic. I don't like blabbering, a mush of unrelated arguments, sprinkled with psychological tricks, like references to authorities. If you're too lazy to read the evidence, I can't help you any further.
You cited GPT as some kind of proof of no idea what. You're just writing for the sake of writing. There is no meaningful message. I'm not interested in optimisation, that's the 3rd question. I did not write about optimisation and did not ask about it. If learning includes optimisation, it doesn't mean that learning is optimisation. That's not what the conversation was about at all.