我發現聽像詹森這樣的人談論他們的成功/生活哲學以及他們的價值觀等非常吸引人，並且玩一個遊戲，看看其中有多少是「經過精心解讀和重新調整的傳說，附加在一個擁有數十億變量的複雜生活故事上，他們對此毫無控制」與「在經歷了許多偶然事件後，走進一個神聖的位置之前就已經存在」的對比。 幾乎總是屬於第一類，這就是為什麼聽他喋喋不休地談論他曾是洗碗工讓我感到毛骨悚然。

當我花了2分鐘寫一篇帖子，而它沒有立即獲得10個讚時，「Boost」按鈕看著我的樣子

the next crop of >$1B-raised AI labs like Thinky are honestly gonna do incredibly well because they aren’t tied to existing products and caught up in momentum and {{{Business Insights}}} and maintaining DAU etc… they can pick all the most sensical things being done while getting rid of all the immense garbage that was accumulated such as the lovecraftian OG RLHF assistant-mode methods, incredibly narrow use of RL in post training, tokenizers, whatever you can think of the “top AI talent” concept is also immensely overrated as there are brilliant people everywhere and top-end engineers everywhere if anything, not getting some ex-OAI ex-Deepmind ex-whatever seconds might even help as most of these ppl who aren’t in pure research are already partially blind due to being too deep in the existing approaches these labs can come in and (I hope) flesh out and take the best parts of the last few years of research that hasn’t been scaled up and it’s likely they hit gold, be able to think outside the box cuz they don’t have some homonculus CEO trying to grow a product there are endless ways that they can move the needle because so so so much, especially around the models, still has tons of low hanging fruits Then, the cycle will repeat and they will turn into full-on product companies, become chained to their previous work, etc, bc I don’t think more than one clear pure AI lab that’s a titan-level company that is bound to be here in 50yrs atm