Prof. Geoffrey Hinton's speech for the IEEE/RSE Wolfson James Clerk Maxwell Award

翻译 王彦晶 原文听写见下

视频链接

五十年前,人工智能的创始人们说服大家相信逻辑是人工智能的关键,我们得想办法让机器能做逻辑推理。一个被这些大佬认为是疯了的替代的进路是选择忘掉逻辑,试图去理解大脑神经元网络是如何学习的。有趣的是,两位曾经反对AI的逻辑方法的人恰恰是图灵和冯诺依曼。我想如果他们中有一个活的更长些的话,恐怕现实会很不一样。过去35年来,我很幸运的和一群极富天赋的研究生和博后一起工作,就在几年前,我们终于让神经网方法能够突破现有的语音及物体识别的技术瓶颈,取得了更好的结果。现在神经网络已经被广泛应用于世界的各个角落。看来,那个“疯狂”的方法赢了哦 :) 谢谢!

50 years ago the fathers of Artificial Intelligence convinced everybody that logic was the key to artificial intelligence. Somehow we have to get computers to do logical reasoning. The alternative approach which they thought was crazy was to forget logic and try to understand how networks of brain cells learn things. Curiously two people who rejected the logic based approach to AI were Turing and von Neumann. If either of them had lived I think things would have turned out differently. Over the last 35 years I’ve the good fortune to work with an incredibly talented collection of graduate students and postdocs. And few years ago we finally got neural networks to work better than the existing technologies of speech recognition and object recognition. Now neural networks are everywhere, and it looks like the crazy approach is winning. Thank you!


Pasted Graphic 1

作为做AI的逻辑学家,我们要实事求是的面对神经网这些年的进步,学习Hinton坚持不懈的精神(神经网那么多年没啥funding都扛过来),更加open的取长补短,找到逻辑和统计学习办法的结合点,吸引真正有创造力的年轻人,而不是自说自话。三十年河东三十年河西,我相信逻辑在未来的AI里会有不一样的作用。We will be back (in a different shape)!

BTW 人工神经网的第一篇文章可算是一篇逻辑论文:McCulloch, Warren; Walter Pitts (1943). "A Logical Calculus of Ideas Immanent in Nervous Activity". Bulletin of Mathematical Biophysics. 5 (4): 115–133

另外, Turing 和 von Neumann的态度有根据么?