I don't understand why you say OpenAI can't be more efficient, the methods are in the scientific papers they released and the models are open source. They have enough smart people left at OpenAI that can do that I believe, or at least some of the western competitors can. The real innovation that very few are talking about is how they did it and I'm no expert, but it's how they combined Reinforcement learning (RL) for creating LLMs. 3:11 this is probably in part BECAUSE of the sanctions, Chinese knew they had to do with less so they focused on making it more efficient. 3:50 yes and I do think this was always known to happen, the sanctions are only a delay tactic, never a full solution. But I think they underestimated the Chinese ability to innovate and how much can be improved if you focus on efficiency instead of everything else. I also do think the Chinese were happily surprised by the results they got, probably worked even better than expected (kind of get that from the tone of the paper of the few parts I've seen of it).
imagine how much humanity would be better if China had no sanctions, we would have 100$ super computers and flying cars today, imagine the scientific progress would China had if it could do business with everyone without sanctions, we could have cured every disease and turn deserts into forest, we would have colonized the moon if it wasn't for the sanctions
I enjoy your grounded delivery, a focus on the facts and your insights. Subscribed.
Thanks for that! Great to hear
business model of OpenAI is fucked. HAHAHAHA
why use them when i can put a Free, unrestricted, uncensored OpenSource version in my basement
I don't understand why you say OpenAI can't be more efficient, the methods are in the scientific papers they released and the models are open source. They have enough smart people left at OpenAI that can do that I believe, or at least some of the western competitors can. The real innovation that very few are talking about is how they did it and I'm no expert, but it's how they combined Reinforcement learning (RL) for creating LLMs.
3:11 this is probably in part BECAUSE of the sanctions, Chinese knew they had to do with less so they focused on making it more efficient.
3:50 yes and I do think this was always known to happen, the sanctions are only a delay tactic, never a full solution. But I think they underestimated the Chinese ability to innovate and how much can be improved if you focus on efficiency instead of everything else. I also do think the Chinese were happily surprised by the results they got, probably worked even better than expected (kind of get that from the tone of the paper of the few parts I've seen of it).
imagine how much humanity would be better if China had no sanctions, we would have 100$ super computers and flying cars today, imagine the scientific progress would China had if it could do business with everyone without sanctions, we could have cured every disease and turn deserts into forest, we would have colonized the moon if it wasn't for the sanctions
Nice video.
Everything is clear.
I heard that what Trump spent for the mexican wall, China spent similar amount to land its rover on the far side of the moon. 😮