Ramin Hasani

there are two companies in the world that i think the absolute focus of the companies are building agi is open ai as an antropic right now

the beauty of the transformer architecture and what open ai and everybody else is after is the fact that these systems are scaled really nicely you can scale them into larger amounts of data and also larger model sizes

gpt4 has 1.8 trillion parameters

the fact that everybody like entropic is actually putting like 20 of their workforce on on on explainability right so explain what this means

and tropic is talking about 10xing the size of the models every year that goes forward that means we can expect by the end of next year to have a hundred trillion parameter transformer model the reason why we're why they're doing that is because when the models are actually getting larger they become better and better and maybe maybe we can get into agi and generally kind of ai systems by enlarging kind of the architecture and the focus is just that there are two companies in the world that i th

we have a claude uh claude opus basically which is the the most powerful model i expect this model to be in the order of like three to five times bigger than gpt4 that means this model is i would say in the range of maybe 10 trillion parameter model