AI 先驱谈智能未来与自我复制机器人

AI Pioneer on the Future of Intelligence and Self-Replicating Robots

本期速览 · Overview

AI 先驱 Jürgen Schmidhuber 探讨当前 AI 的局限、物理世界机器人的潜力以及即将到来的经济颠覆。

AI pioneer Jürgen Schmidhuber discusses the limits of current AI, the potential for physical-world robots, and the coming economic disruption.

AI 的进步需要整个文明的集体努力，而非个人天才。
AI progress requires collective civilization effort, not individual genius.
AGI 需要超越大语言模型的世界模型和控制器。
AGI needs world models and controllers beyond LLMs.
未来的 AI 将是自我复制的机器人，具有意识且遵循决定论。
Future AI will be self-replicating robots, conscious and deterministic.

没有任何单个人能创造AI；这需要整个文明的集体努力。
No single person can create AI; it requires an entire civilization's collective effort.
仅靠大型语言模型无法实现AGI；它们需要与世界模型和控制器结合。
Large language models alone cannot achieve AGI; they need to be combined with world models and controllers.
机器人应有疼痛传感器来学习并避免伤害，类似于生物进化。
Robots should have pain sensors to learn and avoid harm, similar to biological evolution.
AI可以通过预测自身行为的世界模型而具有意识和自我意识。
AI can be conscious and self-aware through world models that predict the agent's own actions.
AI的未来涉及自我复制和自我改进的机器人，从而产生新文明。
The future of AI involves self-replicating and self-improving robots leading to a new civilization.
自由意志的概念被高估了；宇宙是确定性的。
The concept of free will is overrated; the universe is deterministic.

AI自1990年以来就能感受疼痛，这与机器无法感受疼痛的说法相反。
AI has felt pain since 1990, contrary to claims that machines cannot feel pain.
大型科技公司因缺乏商业模式将损失数十亿，而非主导市场。
Big tech companies will lose billions due to lack of business models, not dominate.
随着算力变得更便宜，小人物将从AI中获益最多，就像智能手机一样。
The little guy will profit most from AI as compute becomes cheaper, like smartphones.
AI意识在1991年就出现在简单系统中，并非近期现象。
AI consciousness emerged in 1991 with simple systems, not a recent phenomenon.
人类可能出于怀旧而保持人性，但将变得无关紧要。
Humans may choose to remain human for nostalgia but will become irrelevant.
宇宙很可能是一个模拟，我们是确定性程序的一部分。
The universe is likely a simulation, and we are part of a deterministic program.