查尔斯·伊斯贝尔
立场时间线 6 条 · 6 个议题
同一议题按时间排列,说法变化一目了然。
个体行为高度重复,用简单计数统计即可高精度预测人的下一步行为
“任何个体都具有惊人的可预测性,因为你总是在重复做同样的事情。(“Any given individual is remarkably predictable. Because you keep doing the same things over and over again.”)”
来源访谈 →真正的智能存在于与他人的互动之中,孤立个体的智能毫无意义
“孤立的自我智能毫无意义。正确的答案是,你必须通过与他人互动的方式来展现智能。(“Being intelligent in and of myself in isolation is a meaningless act. The correct answer is you have to be intelligent in the way that you interact with others.”)”
来源访谈 →计算学科的独特世界观在于认识到模型、语言和机器三者是等价的
“区分计算科学家的特质是...他理解模型、语言和机器是等价的,它们是同一回事。(“The thing that distinguishes the computationalist... is that he or she understands that models, languages, and machines are equivalent. They're the same thing.”)”
来源访谈 →认为大众科学传播粗俗的自我叙事是错误的,应当被纠正
“我们给自己讲了一个故事、一个叙事,认为[大众科学传播]是粗俗的...我认为这是错的。(“We have told ourselves a story, a narrative, that [popular science communication] is vulgar... I think that's wrong.”)”
来源访谈 →即便是客观的东西,最终也取决于讲故事、做决策和权衡取舍
“即使是客观的东西也成了目标,因为最终你必须讲故事、做决策、进行权衡。(“Even the objective is an objective because at the end, you've got to tell a story, you've got to make decisions, you've got to make trade-offs.”)”
来源访谈 →当前机器学习研究过度工程化、局限于孤立任务,AI系统必须在真实混沌环境中长期生存学习
“要实现真正的进步,必须让AI系统在真实、混沌、不可控的环境中长期(数月)生存和学习,即使这意味着研究周期会被拉得很长。”
来源访谈 →金句墙 10 条
“Statistics is how you're going to keep from lying to yourself.”
统计学是让你不自欺欺人的方法。
机器学习不仅是统计:Isbell与Littman的跨学科对话 · 2020/12/26“ICML was machine learning done by computer scientists, and NeurIPS was machine learning done by computer scientists trying to impress statisticians.”
ICML是计算机科学家做的机器学习,而NeurIPS是计算机科学家试图给统计学家留下深刻印象的机器学习。
机器学习不仅是统计:Isbell与Littman的跨学科对话 · 2020/12/26“Chemistry is just physics, but I don't think it's as useful to think about chemistry as being just physics.”
化学只是物理学,但我认为将化学仅仅视为物理学并不那么有用。
机器学习不仅是统计:Isbell与Littman的跨学科对话 · 2020/12/26“You have to give them what they need without bending to their will.”
你必须给予他们所需的东西,同时不屈服于他们的意志。
机器学习不仅是统计:Isbell与Littman的跨学科对话 · 2020/12/26“The reward for good work is more work. The reward for bad work is less work.”
出色工作的奖励是更多的工作,糟糕工作的奖励是更少的工作。
机器学习不仅是统计:Isbell与Littman的跨学科对话 · 2020/12/26“Any given individual is remarkably predictable. Because you keep doing the same things over and over again.”
任何个体都具有惊人的可预测性,因为你总是在重复做同样的事情。
计算的本质是模型、语言与机器的等价 · 2020/11/2“Being intelligent in and of myself in isolation is a meaningless act. The correct answer is you have to be intelligent in the way that you interact with others.”
孤立的自我智能毫无意义。正确的答案是,你必须通过与他人互动的方式来展现智能。
计算的本质是模型、语言与机器的等价 · 2020/11/2“The thing that distinguishes the computationalist... is that he or she understands that models, languages, and machines are equivalent. They're the same thing.”
区分计算科学家的特质是...他理解模型、语言和机器是等价的,它们是同一回事。
计算的本质是模型、语言与机器的等价 · 2020/11/2“We have told ourselves a story, a narrative, that [popular science communication] is vulgar... I think that's wrong.”
我们给自己讲了一个故事、一个叙事,认为[大众科学传播]是粗俗的...我认为这是错的。
计算的本质是模型、语言与机器的等价 · 2020/11/2“Even the objective is an objective because at the end, you've got to tell a story, you've got to make decisions, you've got to make trade-offs.”
即使是客观的东西也成了目标,因为最终你必须讲故事、做决策、进行权衡。
计算的本质是模型、语言与机器的等价 · 2020/11/2访谈收录

机器学习不仅是统计:Isbell与Littman的跨学科对话
乔治亚理工学院的Charles Isbell与布朗大学的Michael Littman在Lex Fridman播客中展开了一场深度对话。他们探讨了机器学习的本质是否仅仅是“计算统计”,分享了贝尔实验室的黄金时代经历,并深入讨论了在线教育、教学合作以及学术生涯中的自我批判与成长。

计算的本质是模型、语言与机器的等价
佐治亚理工学院教授Charles Isbell在访谈中探讨了交互式人工智能的核心理念,认为智能本质上是与他人和社会的互动过程。他区分了AI与机器学习的不同追求,并指出计算学科的独特世界观在于理解模型、语言和机器的等价性。他还讨论了终身学习、预测人类行为的实验,以及学术界如何平衡专业深度与公共影响力。