Inside AI - November 12th, 2019

Inside AI (Nov 12th, 2019)

BERT bias problems / McDonald's AI drive-thrus / "In the Age of AI"

Subscribe | View in browser

1. Famed computer scientist Robert Munro fed words into Google's BERT and found that the AI associated 99 percent of words with men rather than women. The only outlier was the word "mom," according to Munro, who discussed the inherent biases of BERT and other universal language models in a recent interview with The New York Times. In a blog post, Munro described how he looked into cloud-computing services from Google and AWS that assist in adding language skills to new applications. Both services "failed to recognize the word 'hers' as a pronoun," but correctly identified the word “his," according to The Times. A Google spokesman said the company is aware of the issue and is taking steps to address it. Amazon said it “dedicates significant resources to ensuring our technology is highly accurate and reduces bias." - NYTIMES

2. Recent actions by the Trump administration, such as the U.S blacklist targeting China tech firms, show that the U.S. is aware of China's stronghold in the AI industry, although the efforts could backfire, experts say. China has a number of advantages when it comes to AI implementation, including billions poured into startups and a powerful central government with access to massive amounts of data, according to analysts interviewed by the BBC. But the experts warn that U.S. efforts to stymie China's advancements could backfire. By blacklisting China tech firms, the U.S. could actually be encouraging China to develop its own microchips that compete with Nvidia and others. "To start thinking about putting up export control walls around the U.S. could be as damaging to the U.S. research enterprise as anything that a foreign adversary might try to do to us," says AI expert Tom Mitchell. - BBC

3. The computational power required to train AI is rising seven times faster than previous rates, according to an updated analysis by OpenAI. The research lab added new data to its previous 2018 analysis, which found that the power used to train large AI models doubled every 3.4 months since 2012. New data shows that the present doubling time is more than seven times the rate that occurred from 1959 to 2012, when the amount of needed power doubled every two years, which is in accordance with Moore’s Law. - MIT TECH REVIEW

4. Leaders in retail and other industries recently weighed in on McDonald's plans to start using AI-based predictive suggestions at its drive-thrus. In an online discussion, Retail Wire asked the experts a number of questions about the technology, including if they have privacy concerns about how McDonald's would scan people's license plates to recognize customers and recommend food items. Here are some responses:

"It will be interesting to see the reaction of customers. Some will see it as a convenience while others will think it is down right creepy." - Michael Terpkosh, president, City Square Partners LLC.

"I would regard this less as a convenience and more as an invasion of my privacy." - Steve Montgomery, president, b2b Solutions, LLC.

"Half the time I go to McDonald’s I order something different so do I now need to correct the screen before my order is processed? I guess this could be cool but it doesn’t sound very appetizing to me." - Georganne Bender, consumer anthropologist, Kizer & Bender. - RETAIL WIRE

5. Government is trailing industry in the adoption of AI technologies, according to a newly released report from Deloitte. The accounting firm surveyed 1,100 U.S. executives from firms that have adopted AI. It found that governments are "on the lower end of the AI maturity curve compared to other industries," said Bill Eggers, executive director of Deloitte’s Center for Government Insights. Other key findings:

  • Nearly three-quarters of respondents said AI would be “very” or “critically” important within the next two years.
  • Governments reported the highest percentage of “starter” adopters of AI but had the lowest percentage of “seasoned” adopters.
  • Just over 70 percent of respondents cited a "moderate-to-extreme" skills gap as an obstacle for AI adoption. - FED SCOOP

6. Royal Dutch Shell last year began using cloud-supported artificial intelligence and machine learning to detect breakdowns —sometimes months before they occurred— resulting in a cost-savings of millions so far. The Pernis refinery, Europe's largest, processes 20 million tons of crude oil annually, and glitches can be costly. So the company installed 50,000 sensors that generate 100,000 measurements a minute to better analyze and process data. The model was designed to predict failures in control valves, and it allowed workers to carry out maintenance or adjust operating conditions as needed. - FINANCIAL TIMES

A version of this story first appeared in Inside Cloud.

7. Kristin Tynski, VP of digital marketing firm Fractl, used publicly available AI tools to create fake blog posts and author headshots. The goal of the project - which can be viewed at - is to highlight how easy it is to generate fake posts and even fake people, potentially unleashing a wave of propaganda online, according to Tynski. She told Venture Beat: "What is alarming to me about this new era of high-quality, AI-generated text content is that it could pollute search engine results and clog the internet with a bunch of garbage." - VENTURE BEAT

8. Futurist and AI author Martin Ford says Andrew Yang is the only Democratic presidential candidate to directly engage with voters about how AI and automation will disrupt the job market. In an opinion piece for The Hill, Ford describes how many executives, technologists, and other "elite" experts are very aware of the impending challenges surrounding AI, which is still considered a fringe issue in most political circles. Yang is the only Democratic candidate to propose solutions, he argues, though he did not specify exactly what those are. - THE HILL

9. The PBS/Frontline film "In the Age of AI" is now available to watch online. The nearly two-hour investigation explores the new industrial revolution and the "promises and perils" of AI, according to its web description. The film has five distinct areas that it examines: China’s AI plan, the promise of AI, the future of work, surveillance capitalism, and the surveillance state. - FORBES

10. Ping, which uses AI to automate legal work, raised $13.2 million in a recent funding round. The Series A was led by Upfront Ventures as well as First Round, BoxGroup, Initialized, and Ulu Ventures. According to Ping co-founder and CEO Ryan Alshak, the startup will use the cash infusion to scale up its enterprise distribution. The startup's AI and machine learning software automatically fills out timesheets and logs work for attorneys. - TECHCRUNCH

Written and curated by Beth Duckett in Orange County. Beth is a former reporter for The Arizona Republic who has written for USA Today, Get Out magazine and other publications. Follow her tweets about breaking news and other topics in southern California here.

Editor: Kim Lyons (Pittsburgh-based journalist and managing editor at Inside).

Copyright © 2020, All rights reserved.

Our mailing address is:
767 Bryant St. #203
San Francisco, CA 94107

Did someone forward this email to you? Head over to to get your very own free subscription!

You received this email because you subscribed to Inside AI. Click here to unsubscribe from Inside AI list or manage your subscriptions.

Subscribe to Inside AI