I really think that every marginalized group that suffered #discrimination and systemic oppression should join the discussion about #OpenSourceAI

If a black box will be distributed, regulated and trusted as #OpenSource, no human right organization will be able to inspect the dataset used to train it.

And if you can implant undetectable backdoors in a machine learning model (see https://arxiv.org/abs/2204.06974), you can much more easily implant undetectable bias that will hurt some groups of people to benefit others.

#LGBTQ people and victims of systemic #racism should really demand transparency,...before it's too late!

Planting Undetectable Backdoors in Machine Learning Models

Given the computational cost and technical expertise required to train machine learning models, users may delegate the task of learning to a service provider. We show how a malicious learner can plant an undetectable backdoor into a classifier.

^arXiv.org

TIL that a #BlackBox #AI can be defined as #OpenSource.

Granting the “freedom to fine tune” cannot be enough to qualify as Open Source AI just like granting the “freedom to configure” cannot be enough to qualify as Open Source Software.

Yet, for #OSI it's enough.

And you can't provide effective solutions to the problem they pretend to see into making training data available.

Or you will be silenced.

https://opensource.org/blog/copyright-law-makes-a-case-for-requiring-data-information-rather-than-open-datasets-for-open-source-ai

#OpenSourceAI #OpenWashing #OpenWashingAI
#OSS #OSAI #AIAct #BigTech #GAFAM

Copyright law makes a case for requiring data information rather than open datasets for Open Source AI

When we look at applying Open Source principles to the subject of AI, copyright law comes into play, especially for the topic of training data access.

^{Open Source Initiative}