Original post · AI translation
Sam Altman 称 token 是“一种虚假的单位”：你可以数出 token 的数量，但这个数量并不能告诉企业花的钱换来了什么。
为 token 辩护的唯一理由来自信息论：模型在传递有效信息的同时仍会输出噪声，而一旦去除这种浪费，token 数量就会接近香农下限，即传递实际信息所需的最低数量。
----
完整视频见“YouTube 频道 Cheeky Pint”（链接在评论中）
Original
Sam Altman calls tokens “a kind of fake unit”: you can count them, but a token count does not tell a business what it got for the money. The one defense of tokens is information-theoretic: models still stream noise along with signal, and once that waste is removed, token counts approach Shannon’s floor, the minimum needed to carry the actual information. ---- Full video on "Cheeky Pint" YouTube channel, (link in comment)
完整视频见 YouTube 频道“Cheeky Pint”
https://t.co/HSkuaiJDkY
Original
Full video on "Cheeky Pint" YouTube channel https://t.co/HSkuaiJDkY
Sam Altman 表示，世界可能需要为单个 AI 模型所能达到的智能程度设定上限
> “我有点觉得，至少在一段时间内，也许是很长一段时间内，我们必须就允许这些单个模型变得多聪明达成一些限制。
但这一决定会导致进展不足。原因不是 AI 没有能力开展出色的研究。”
----
完整视频见 YouTube 频道“Cheeky Pint”（链接在评论中）
Original
Sam Altman says the world may need to cap how smart a single AI model is allowed to get > "I kind of think that, at least for a while, maybe for a long while, we are going to have to agree to some limits about how smart we let these single models get. But, a lack of progress will come from that decision. Not from the AI's inability to do good research." ---- Full video on "Cheeky Pint" YouTube channel, (link in comment)