Uber 四个月烧光年度 AI 预算,体验却没变好
yaroslavvb · x · 2026-11-23
yaroslavvb observes two extremes of LLM impact: on the enterprise side, Uber used up its annual AI budget in 4 months yet the Uber experience 'somehow didn't get better'; on the other, recreational mathematicians are getting 10x'd.
His example: a new website for interactively folding fractal tilings, illustrating the leap in individual capability.
「漫话AGI」频道最新
- 争议:后训练算力已超 75%,用预训练 perplexity 定安全上限站不住 — stochasticchasm · 2026-11-23
- FDA 首次批准人体逆转衰老基因疗法临床试验 — Dr_Singularity · 2026-11-23
- Gary Marcus 回击 LLM 炒作质疑:批评过度宣传是工程严谨 — GaryMarcus · 2026-11-23
- AI 安全圈内讧:学者批评公开点名爆料者是在烧毁公共资源 — sjgadler · 2026-11-23
- AI 证明是另一种暴力:够快够全就不必发明新数学 — moultano · 2026-11-23
- Google 研究员观点:AI 数学证明靠速度与全知 brute force,可绕过人类式的巧思 — moultano · 2026-11-23