Every week, my team at Invector Labs publishes a newsletter that covers the most recent developments in AI research and technology. You can find this week’s issue below. You can sign up for it below. Please do so, our guys worked really hard on this:
From the Editor: The Apple Card AI Bias Drama
A bit of a controversial note for today.
The challenges of bias in artificial intelligence(AI) models headlines this week with the controversy surrounding the Apple Card. Based on initial reviews, the new credit card granted higher limits to men than to women without a reasonable explanation. These developments are certainly concerning it even triggered an investigation by the New York State Department of Financial Services. Needles to say that the media had a field day with these news and every tech reporter seemed to be an AI bias expert this week.
The case of Apple’s new credit card shows the risks of relying completely on AI for decisions that affect humans. However, the solution is not as trivial as the media makes it look. Solving AI bias goes beyond data curation techniques and, for now, it might require human intervention. We understand very little about the cognitive processes behind human bias. Many of them are incredibly subjective and related to education or socio-economic aspects which they don’t quite translate easily into data. Algorithms are going to show signs of bias for years to come and the solution will require very strong computer science and social science rigor. I believe that trivializing these matters does a disservice to the AI industry. If anything, AI offers a blank canvas to remove many of the biases present in our everyday lives. Apple and Goldman need to get back to work and fix these issues but if you think this is the only credit card product with a biased selection process, think again
Now let’s take a look at the core developments in AI research and technology this week:
Detecting Harmful Content
Facebook published a new edition of the Community Standards Enforcement Report that details how they are using self-supervised models to detect harmful content.
Generating Training Data for Text Classifiers
IBM AI researchers published a paper proposing a generative method to synthesize new training data for text classification models.
Understanding Human Mobility using AI
Google AI researchers published a paper that provides remarkable insights about human mobility in the last few decades.
Cool AI Tech Releases
Baidu open sourced a new version of PaddlePaddle, the deep learning framework designed powering most of its products.
New Mobile Vision Models
Google introduces the next generation of its on-device vision models based on hardware-aware AutoML.
Crowdsourced AI Training
Applause, a startup focused on automated testing, announced a new service that leverages humans and algorithms to train and test machine learning models.
AI in the Real World
Goldman and Apple Under Fire Over AI Bias
The controversy surrounding the Apple Card approval process has shade new light about the importance of bias training in AI models.
Top US Senate Democrat Pitches a Massive Investment in AI
US Senator Charles Schumer proposes the creation of a new agency that will invest around $100 billion in AI in the next five years.
Oculus CTO leaves Facebook to Pursue AI Project
Legendary programmer John Carmack is leaving Facebook after six years to pursue a project related to artificial general intelligence.