Forbes reads 12 pain points in the O'Reilly Artificial Intelligence Conference

September 06, 2024

What is the status of artificial intelligence development in the future and what are the difficulties faced by it? What are the future development potentials? Gll Press, an industry expert who participated in the O'Reilly artificial intelligence conference, brought his 12 observations. The views he extracted mainly come from top AI experts such as Peter Norvig and Yann LeCun. In addition, there are executive-level experts from Microsoft, Nvidia, and Allen Artificial Intelligence Institute. 12 observations include: AI is a black box; AI has high technical difficulty; unmanned vehicles may make driving a hobby; AI needs to consider culture and background - training affects learning; AI does not replace all the work of people; AI Will not destroy humanity; AI is not magic, deep learning is useful, but there are limitations; AI is enhanced intelligence, and has the strength of humans and machines; AI changes the way we interact with computers; AI needs to be smarter Testing, Turing test is not enough; Churchill's view of AI; the futile pursuit of human intelligence at the materialistic paradigm may hinder the development of AI.

At this year's O'Reilly Artificial Intelligence Conference, 66 AI practitioners from 39 organizations introduced the current state of AI development: from chatbots to deep learning, to automatic technology and emotional recognition, and to automated work and AI progress. Barriers, and even life-saving and business opportunities, are involved. The meeting was chaired by Peter Norvig and Tim O'Reilly as the Program Committee. The following is the participants' observation of Gll Press.

Gll Press is a managing partner of a consulting company and author of Forbes website. His article is a summary after attending the O'Reilly artificial intelligence conference, refining 12 observations of the current AI industry, such as the black box problem, where the limitations of deep learning, human-computer interaction issues, AI winter, etc. .

1.AI is a black box: cracking method of trust-but-verify

Peter Norvig, research director of Google, once said, "In contrast to traditional software, machine learning does not produce code, it is more like a black box. You can get a glimpse of it and have some understanding of what is going on inside, but you don't see it. The whole picture."

Tim O'Reilly wrote recently in one of his blog posts: "Because many algorithms that are reshaping our society are black boxes... In the field of deep learning, because they are difficult for even their creators to understand, The key to the problem is therefore trust. An important discipline in today's world is understanding how to evaluate algorithms without knowing the exact rules that the algorithm follows."

O'Reilly proposes a trust-but-verify method for the algorithm, four rules: the expected result is known, external observations can verify the result; how to measure "success" is clear; the creator of the algorithm is The user's goals are consistent; and the algorithm helps creators and users make better long-term decisions.

Aquarium Equipment

Aquarium Equipment,Aquarium Air Pump,Submersible Pump,Intelligent Aquarium Lamp

Sensen Group Co., Ltd.   , https://www.sunsunglobal.com