Skynet This Week #11: Biased recruiting tools, explainable AI, AI pop music and more!
Our bi-weekly quick take on a bunch of the most important recent media stories about AI for the period 8th October 2018 - 22nd October 2018.
Advances & Business
Philip E. Ross, IEEE Spectrum
There is a clear conflict of interest in this piece, as Gill Pratt is trying to sell his own system in competition with self-driving cars. He has a rock solid argument though: his driving assistance system promises as much safety as self-driving cars, much earlier.
Will Knight, MIT Technology Review
A look at how Waymo tries to trip up it’s self-driving cars in a virtual world so that they don’t trip up in the real world. Testing in a virtual world is seen as a vital part of the validation process for software updates to self-driving cars.
Concerns & Hype
Jeffrey Dastin, Reuters
Amazon has been developing an AI recruiting tool that rates eligible candidates on a scale of one to five. Amazon’s researchers found that the tool, was biased against women, scoring resumes with words like “women’s” lower. Even forcing the tool to ignore words that indicate gender, the biases continued to be seen. Amazon eventually shut down the experiment and a new team has been tasked with employment screening, with a focus on diversity. For a general explanation of bias in the field of AI, take a look at our recent editorial.
Amazon’s computer models were trained to vet applicants by observing patterns in resumes submitted to the company over a 10-year period. Most came from men, a reflection of male dominance across the tech industry.
Melissa Avdeeff, The Conversation
The group Skygge has released a pop album using parts of tracks generated using a probabilistic model. We think that it is far too early to talk about a “new era” that would see machine generated music dominate the market; algorithms and machine learning have been used before to generate music, and there is no indication yet that this will become a generality.
Caroline Haskins, Motherboard
Despite the alarming headline, the idea behind the machine learning powered app is to help people fill specific lawsuits at the Small Claims court (US) against abusive behavior by various big corporations. Rather than “suing anyone”, this looks like a step forward in helping people who cannot afford a lawyer win small fights against existing power structures.
Analysis & Policy
AI Now Institute, AI Now Institute*
The AI Now Institute reports on their workshop hosted at NYU Law’s Center on Race, Inequality, and the Law about algorithm-based decisions in criminal justice, health care, education and employment. The discussions on specific cases are especially enlightening. A must read not only for anyone involved in decision making, but for those who have been confronted to algorithm-based decisions.
GoodAI Blog, Medium
Unlike the Pew survey covered in our previous brief, this article is about a survey that took place at a summit for proponents of the “singularity” (an hypothesised point in time where AI will surpass humans in terms of general intelligence). Unsurprisingly, the participants were overwhelmingly positive in their view of the future of AI. The surprise rather comes from the difference in perception between the general public and people who participate to these events.
Philip E. Ross, IEEE Spectrum
William Nordhaus, who won the Nobel Prize in economics this month, does not believe that the singularity is near. This brief article summarizes his argument: to the extent that a event like the singularity would be predictable, nothing indicates that it is about to happen.
Brian Bergstein, MIT Technology Review
This article is a level-headed summary of what machines currently can do, what they cannot do, and what risks are are more urgent than others.
That second idea [we’d better hope that machines don’t eliminate us altogether], despite being an obsession of some very knowledgeable and thoughtful people, is based on huge assumptions. If anything, it’s a diversion from taking more responsibility for the effects of today’s level of automation and dealing with the concentration of power in the technology industry.
James Vincent, The Verge
The Partnership on AI (PAI), which is a US based consortium for the development of safe AI added it’s first Chinese member, Baidu. It will be interesting to see how companies from different parts of the world deal with the core tenets of the PAI.
“We cannot have a comprehensive and global conversation on AI development unless China has a seat at the table.”
Expert Opinions & Discussion within the field
New York Times
The New York Times as part of their AI special report, spoke to five different experts from different fields about AI. It offers an interesting look into how experts from different backgrounds regard AI.
Vincent Vanhoucke, Medium
Vincent Vanhoucke, Principal Scientist at Google offers his take on building interpretable and explainable AI systems. He argues that evidence from neuroscience should guide how the issue of explainability should be tackled for AI systems.
Limor Gultchin, Skynet Today
Our latest editorial that gives a summary of what AI bias is, why we should care, and what we can and are doing about it.
Sanders Kleinfeld, Google Blog
Along with their recent Inclusive Images Kaggle competition, Google has launched a new course to help everyone better understand fairness in machine learning.
Vasily Zubarev, vas3k blog
A great explainer of Machine Translation covering its history and the progression to current state of the art.