digests /

Last Week in AI #44

Biased machines vs biased humans, the AI Index Report, drone racing and more!

Mini Briefs

Biased Algorithms Are Easier to Fix Than Biased People

Sendhil Mullainathan makes the case that bias in algorithmic systems is easier to fix, than fixing bias in people. This analogy from the article succinctly makes the point:

It is much easier to fix a camera that does not register dark skin than to fix a photographer who fails to see dark skinned people.

We used fictitious resumes to find employer bias.
We used statistics to find bias in a healthcare algorithm.

Side by side these studies show why discrimination perpetrated by algorithms is so very different from discrimination perpetrated by peoplehttps://t.co/lhYEPhSgnl
— 𝐒𝐞𝐧𝐝𝐡𝐢𝐥 𝐌𝐮𝐥𝐥𝐚𝐢𝐧𝐚𝐭𝐡𝐚𝐧 (@m_sendhil) December 6, 2019

As they correctly point out, fixing bias in algorithmic systems requires proper regulation, which does not exist yet. Kristian Lum (@KL Divergence) adds:

However, the idea that it’s easy to fix a biased algorithm takes the part of the process that is **really** hard— making value-laden decisions about what kinds of "bias" and disparities are and are not acceptable — as given.
— Kristian Lum (@KLdivergence) December 7, 2019

We aspire that algorithmic systems can be better than humans not just at scale and speed, but also in terms of fairness and bias. However, the fact that diversity and inclusion needs to improve to achieved this aspiration does not change. We need to have a diverse set of people at the table when discussing what bias even looks like in the context of an algorithmic system.

Introducing the AI Index 2019 Report

The AI Index Report for 2019 is out. It provides data about AI spanning multiple disciplines and industries.

The purpose of the project is to ground the discussion on AI in data, serving practitioners, industry leaders, policymakers and funders, the general public and the media that informs it.

This year, they have tripled the number of datasets included, and additionally created the Global AI Vibrancy Tool, an interactive tool that compares countries’ global activities. As a means to track technical progress in the field, they have released the AI Index arXiv Monitor, that enables a better search experience on the arXiv preprint repository.

Advances & Business

DeepMind’s Dreamer AI learns from the past to predict the future - Researchers hailing from Google, Alphabet subsidiary DeepMind, and the University of Toronto sought to exploit this with an agent — Dreamer — designed to internalize a world model and plan ahead to select actions by “imagining” their long-term outcomes.
A self-driving truck delivered butter from California to Pennsylvania in three days - A Silicon Valley startup has completed what appears to be the first commercial freight cross-country trip by an autonomous truck, which finished a 2,800-mile-run from Tulare, California to Quakertown, Pennsylvania for Land O’Lakes in under three days.
Google Assistant’s interpreter mode is coming to phones today - Interpreter mode, the feature that allows Google Assistant to translate your conversations in real time, is coming to phones. Google says it will work with 44 languages and can be invoked by saying commands like “Hey Google, help me speak Thai” or “Hey Google, be my German translator.”
Hair-Brushing Robot Shows How Artificial Intelligence May Help the Disabled - A team of researchers have created a hair-combing robot as possible aide to stroke victims and the disabled.

Concerns & Hype

Humans best AI in first-ever Drone Racing League showdown - The Drone Racing League recently held its first Human VS AI match, pitting a drone helmed by AI developed by Delft University of Technology’s MavLab against one flown by human pilot Gabriel Kocher.
Emotion recognition technology should be banned, says an AI research institute - There’s little scientific basis to emotion recognition technology, so it should be banned from use in decisions that affect people’s lives, says research institute AI Now in its annual report.
A Sobering Message About the Future at AI’s Biggest Party - More than 13,000 artificial intelligence mavens flocked to Vancouver this week for the world’s leading academic AI conference, NeurIPS. The venue included a maze of colorful corporate booths aiming to lure recruits for projects like software that plays doctor.
AI R&D is booming, but general intelligence is still out of reach - Trying to get a handle on the progress of artificial intelligence is a daunting task, even for those enmeshed in the AI community.
Facial Recognition Is Everywhere at China’s New Mega Airport - Flying in and out of Beijing may soon be a Minority Report-like experience.

Analysis & Policy

Public opinion lessons for AI regulation - An overwhelming majority of the American public believes that artificial intelligence (AI) should be carefully managed. Nevertheless, as the three case studies in this brief show, the public does not agree on the proper regulation of AI applications.
The first effort to regulate AI was a spectacular failure - Albert Fox Cahn participated in New York City’s process to understand how automated decision systems are impacting its citizens. But, he writes, the historic, well-intentioned initiative went “horribly wrong.”
Making deepfake tools doesn’t have to be irresponsible. Here’s how. - It’s possible to limit the harm synthetic media tools might cause—but it won’t happen without effort.
This Year’s Hottest Job Involves Artificial Intelligence - A.I. specialist is the fastest growing job in terms of number hires, at least according to LinkedIn.
In the EU, facial recognition in schools gets an F in data protection - The eruption of scandals and debate about facial recognition has become almost everyday news around the world and Europe is no exception. Pilot projects and the testing of systems are widespread, and in the case of France and Sweden, these trials are happening in schools.
Artificial Intelligence Isn’t an Arms Race - And by treating it like one, the United States could miss out on its real potential.

Expert Opinions & Discussion within the field

The AI community needs to take responsibility for its technology and its actions - On Monday, at the opening of one of the world’s largest gatherings of AI researchers, Celeste Kidd addressed thousands of attendees in a room nearly twice the size of a football field. She was not pulling her punches.
A.I. Meets #MeToo At Prominent Computer Science Conference - Celeste Kidd, a cognitive psychologist, used her keynote talk at a prominent artificial intelligence conference to highlight issues around sexual harassment that have roiled the field over the past two years.
Yoshua Bengio, Revered Architect of AI, Has Some Ideas About What to Build Next - The Turing Award winner wants AI systems that can reason, plan, and imagine.
An AI conference once known for blowout parties is finally growing up - At the most prominent AI research gathering of the year, invited speakers and attendees are grappling with how to make technology that better serves the world.
AI Safety: Charting out the High Road - This past year, revelations about the plight of Muslim Uighurs in China have come to light, with massive-scale detentions and human rights violations of this ethnic minority of the Chinese population.

Explainers

Towards Neuroscience-Grounded Artificial Intelligence - Why There Will be no Human-Level Artificial Intelligence Before Understanding Biological Intelligence First.
Model-Based Reinforcement Learning: Theory and Practice - Reinforcement learning systems can make decisions in one of two ways. In the model-based approach, a system uses a predictive model of the world to ask questions of the form “what will happen if I do x?” to choose the best x.
Why are so many AI systems named after Muppets? - One of the biggest trends in AI recently has been the creation of machine learning models that can generate the written word with unprecedented fluidity. These programs are game-changers, potentially supercharging computers’ ability to parse and produce language.