Discover more from Last Week in AI

Weekly text and audio summaries of the most interesting AI news, as well as editorials commenting on recent events

Over 20,000 subscribers

Last Week in AI News #67

Last Week in AI

Jun 1, 2020

Subscribe for future emails here!

Check out our new audio digest!

Mini Briefs

Robots could help save your local store from going out of business

The Covid-19 pandemic has dramatically accelerated the growth of e-commerce and adoption of automation technologies in retail. While the biggest users of robots for order fulfilment are large companies like Amazon, this is starting to shift in the last couple of months. Efficient fulilment of online orders is crucial for retailers to survive the current and worst drop in retails sales since records began. As a result, robotics startups working in fulfillment, providing solutions for bin picking, sensing, and task planning, are seeing a boom, and many are working to transform smaller retailers to compete with larger players.

Adopting new automation technologies will not be “plug-and-play” for everyone, and there are many challenges in working with legacy systems. Still, the current pandemic has forced many companies to step into the future:

“There’s always a lot of aspirational talk about the future and what companies would like to do,” says Gravelle. Suddenly there are fewer reasons to put off those plans: “Now they have to do it.”

Eye-catching advances in some AI fields are not real

Advances in AI are usually measured by performance on some benchmark tasks, and newer AI algorithms often appear to supersede the performance of their predecessors. However, recent studies have shown that in many subfields of AI, old algorithms, when properly tuned, perform just as well as the newer ones. This brings doubt on the current perception of progress in AI. In some subfields, there is “no clear evidence of performance improvements over a 10-year period.”

There seem to be multiple sources of misleading progress in AI research. One is that researchers are incentived to publish new algorithms instead of tweaking old ones for improvements. Another is that when training a new algorithm takes days or longer on large compute clusters, it is very expensive in both time and money to perform thorough and fair comparisons with previous approaches. Lastly, due to the recent explosion of interest in AI, new researchers far outnumber experienced ones who can provide adequate review and feedback.

Podcast

Check out our weekly podcast covering these stories!
Website | RSS | iTunes | Spotify | YouTube

Advances & Business

Google AI researchers want to teach robots tasks through self-supervised reverse engineering - A preprint paper published by Stanford University and Google researchers proposes an AI technique that predicts how goals were achieved, effectively learning to reverse-engineer tasks.
AI could help scientists fact-check covid claims amid a deluge of research - An experimental tool helps researchers wade through the overwhelming amount of coronavirus literature to check whether emerging studies follow scientific consensus.
Scale AI releases free lidar dataset to power self-driving car development - High quality data is the fuel that powers AI algorithms. Without a continual flow of labeled data, bottlenecks can occur and the algorithm will slowly get worse and add risk to the system.
The Air Force’s AI-Powered ‘Skyborg’ Drones Could Fly as Early as 2023 - The drones would fly alongside Air Force warplanes, doing jobs too dangerous or dull for pilots.
How Army Futures Command plans to grow soldiers’ artificial intelligence skills - With artificial intelligence expected to form the backbone of the U.S. military in the coming decades, the Army is launching a trio of new efforts to ensure it doesn’t get left behind, according to the head of Army Futures Command.

Concerns & Hype

Walmart Employees Are Out to Show Its Anti-Theft AI Doesn’t Work - The retailer denies there is any widespread issue with the software, but a group expressed frustration—and public health concerns.
Top safety official at Waymo self-driving unit stepping down - Alphabet Inc’s Waymo self-driving unit said on Thursday that its chief safety officer, Debbie Hersman, was stepping down but would remain as a consultant to the company. Hersman, the former chair of the U.S.
A robot sheepdog? “No one wants this,” says one shepherd - It’s certainly an arresting image: a four-legged robot trots across a grassy hillside, steering a herd of sheep without a human in sight.
U.S. Blacklists Dozens of Chinese Companies Working on AI, Face Recognition Tech - The U.S. Commerce Department has put another 33 Chinese businesses - many of which develop artificial intelligence and face recognition tech - on its economic blacklist as a punitive measure for purportedly conspiring with Beijing and the government’s brutal crackdown on Muslim minorities.
Twitter billionaire Jack Dorsey: Automation will even put tech jobs in jeopardy - The rise of artificial intelligence will make even software engineers less sought after. That’s because artificial intelligence will soon write its own software, according to Jack Dorsey, the tech billionaire boss of Twitter and Square.
Deepfakes Are Going To Wreak Havoc On Society. We Are Not Prepared. - Last month during ESPN’s hit documentary series The Last Dance, State Farm debuted a TV commercial that has become one of the most widely discussed ads in recent memory. It appeared to show footage from 1998 of an ESPN analyst making shockingly accurate predictions about the year 2020.
A fight for the soul of machine learning - Last Tuesday, Google shared a blog post highlighting the perspectives of three women of color employees on fairness and machine learning.

Explainers

Research summary: Troubling Trends in Machine Learning Scholarship - With the explosion of people working in the domain of ML and the prevalence of the use of preprint servers like arXiv, there are some troubling trends that have been identified by the authors of this paper when it comes to scholarship in ML.
Research summary: Warning Signs: The Future of Privacy and Security in the Age of Machine Learning - There are no widely accepted best practices for mitigating security and privacy issues related to machine learning (ML) systems.