Category: All Languages

Fairness and privacy, Removing context encourages unfairness

Han Zhao @hanzhao_ml How to ensure fairness (statistical parity) and privacy (DP) simultaneously? What are the costs of privacy and fairness upon accuracy? Excited to share our #ICML2024 work answering the two questions above! paper: https://arxiv.org/pdf/2405.04034 code: https://github.com/rxian/fair-regression Replying to @hanzhao_ml Han, Link to abstracts to give context. If I could, I would outlaw use
Read More »

The survival of the human and related species

Borriss @_Borriss_ People like critiquing OpenAI for many things.. (no problem)  But the reality is that they’re pushing AI forward.  It’s been 17 months since they launched ChatGPT.  Why didn’t Apple or Google or Meta leapfrog and present an AI assistant first?  If we had to wait for them to… Replying to @_Borriss_ Borriss, the problem is that OpenAI
Read More »

Chaos at Cambridge

You really ought to think, not in terms of “publication in 90 days” but “enabling real time collaboration globally”. The “90 days” was fine in the 1800’s, but today adding quarter year delays into global critical developments is like adding a mlll stone to a Formula 1 race car.   If authors are chasing after
Read More »

XFinity, like all organizations now, has a lossy fragmented distributed information system.

#XfinitySupport I had to rest for a while, but did go over my experiences with XFinity in recent days and weeks and years. It is typical of what most large corporations face now – trying to piece together many divisions and workers with a bewildering variety of computers and information systems, rules and automated systems.
Read More »

Global open resources – discussions, suggestions, links, knowledge – where all share in exploring

https://cosmosmagazine.com/history/archaeology/oldest-neanderthal-cave-engravings-france/ Cosmos offers newsletters, but still seems to separate the whole world into publishers, authors, readers in a strict hierarchy. The articles are difficult to read, not using explicit references with title, authors and affiliations. Just hyperlinks hidden under fairly generic words like “recent studies”, “published”. The users experience is jarring. Going from Cosmos style
Read More »