[Amdnl] Academic Dishonesty in the Many Deep Convolutional Networks?

Juyang Weng weng at cse.msu.edu
Wed Feb 22 10:44:21 EST 2017


Dear ALL:

The ImageNet contests stirred up public interests in AI, which is 
positive. However, Contest practices that violate well-recognized 
engineering concepts gave false messages that misled government funding 
agencies, private foundations, industrial investors, and young or 
inexperienced new researchers. The associated fundamental limitations of 
the Contests will wane public interests sooner or later.

The moderators of connectionists at cmu appear to have operated in black 
box, imposed academic speech censorship, retaliated a post author by 
depriving him for two years of not only the freedom to post but also the 
rights to be notified by posts from others, and covered up academic 
dishonesty under a conflict-of-interest situation when a graduate 
student acted as a moderator.

More information:
English:
Many Deep Convolutional Networks Have Academic Dishonesty?
https://www.facebook.com/juyang.weng/posts/10155115304114783
Chinese:
生活是科学(13):人工智能的深度学习神经网络多是学术作假?
http://blog.sciencenet.cn/blog-395089-1034721.html

-John

-- 
--
Juyang (John) Weng
------------------- Work ---------------------  ---- Technology Transfer ----
Professor                                       Founder
Department of Computer Science and Engineering  GENISAMA LLC
MSU Cognitive Science Program                   Okemos, MI 48864 USA
and MSU Neuroscience Program                    Tel: 517-980-6270
428 S Shaw Ln Rm 3115                           Web: genisama.com
Michigan State University                       --------- Outreach ----------
East Lansing, MI 48824 USA                      Founder
Tel: 517-353-4388                               Brain-Mind Institute
Fax: 517-432-1061                               Web: brain-mind-institute.org
Email: weng at cse.msu.edu                         Brain-Mind Magazine
Web: http://www.cse.msu.edu/~weng/              Web: brain-mind-magazine.org
----------------------------------------------  -----------------------------





More information about the Amdnl mailing list