There is a toxicity problem in the machine learning community

Toxicity is everywhere.



First , the peer review process has been disrupted. A quarter of papers from the NeurIPS conference are uploaded to arXiv. DeepMind has researchers who publicly harass reviewers who criticize their ICLR submission. In addition, articles from renowned institutions with arXiv are accepted at leading conferences, even if the reviewers decide to reject the work. Conversely, some articles with the most positive reviews are rejected (I don’t want to name any, just take a look at this year's ICRL openreview page).



Second , there is a crisis of reproducibility . Apparently, tuning hyperparameters on a test set has become a standard practice nowadays. Articles that do not outperform the current best method have zero chances of being accepted at a good conference. As a result, hyperparameters are tweaked to achieve performance gains where none is available.



Third , there is the problem of worship... Every article related to Stanford or DeepMind is hailed as a breakthrough. For example, BERT has seven times more citations than ULMfit. Belonging to Google gives the article a lot of credibility and visibility. At every ICML conference, a crowd of people stands in front of every DeepMind poster, regardless of the content of the work. It's the same story with Zoom meetings at the ICLR 2020 virtual conference. Moreover, NeurIPS 2020 received twice as many applications as ICML, although both conferences are of the highest level. Why? Why is the word "neural" so praised? Next, Bengio, Hinton and LeCune [winners of the 2018 Turing Award for AI Research - approx. ln] ​​are indeed the pioneers of deep learning, but calling them the "godfathers of AI" is crazy. This is already becoming a cult.



FourthlyYang LeCun has been rather mild about the themes of bias and fairness. However, in response, received completely inadequate toxicity and negative reaction. Getting rid of LeCoon and shutting the person up is not a solution.



Fifth , machine learning and computer science in general have a huge inequality problem(diversity). In our CS faculty, only 30% of students and 15% of professors are women. Taking parental leave during graduate or doctoral studies usually means the end of an academic career. However, this inequality is often misused as an excuse to shield some people from any form of criticism. Reducing every negative commentary in scientific discussion to race and gender creates a toxic environment. People are afraid to participate in discussions lest they be labeled racists or sexists, which in turn exacerbates the problem of inequality.



Sixth , morality and ethics are set arbitrarily... In any discussion, US domestic politics dominate. At this very moment, thousands of Uyghurs are sent to concentration camps based on computer vision algorithms invented by this community, and no one seems to care at all. Adding a "Wider Impact" section at the end of each paper will not solve this problem. Heaps of crap are raised when some researcher is not mentioned in an article. Meanwhile, the African continent, with a population of 1 billion, is virtually excluded from any meaningful discussion of ML (apart from a few Indaba workshops).



Seventh , the mentality is widespreadtype "publish or die". If you don't publish 5+ articles a year at NeurIPS / ICML conferences, you are a failure. The research teams have grown so large that the supervisor does not even remember the names of all the graduate students. Some submit more than 50 entries to NeurIPS per year. The sole purpose of this article was to add another NeurIPS article to the resume. The quality is secondary; the main goal was to go through the preview stage.



Finally , the discussions became disrespectful... Jurgen Schmidhuber, Helmholtz Prize Winner from the International Society of Neural Networks, calls the Royal Society of London Jeffrey Hinton a thief, Timnit Gebra, Ethiopian-American and co-director of the Ethical Artificial Intelligence Team at Google, calls the "godfather of AI" Jan LeCun, director of I Caltech research, as a white suprematist at Nvidia, Anima Anandkumar calls Geometric Intelligence CEO and book author Gary Markus sexist. Everyone is attacked, but nothing gets better.



Albert Einstein opposed quantum mechanics . Can we please stop demonizing those who do not share our views exactly? Let's allow people to disagree by not cutting their throats.



The moment we shut up people because of their opinion, scientific and social progress simply dies.