Radical Reads

Unpacking the 2021 AI Index Report

By Daniel Mulet, Partner

http://Unpacking%20the%202021%20AI%20Index%20Report

Photo credit: Stanford Institute for Human-Centered AI’s 2021 AI Index Report

The Stanford Institute for Human-Centered AI (HAI) released their 2021 AI Index Report, now in its fourth edition, a very significant effort to collect and distill complex information about trends in AI. At 220 pages, the report is comprehensive and a terrific resource for anyone interested in learning about the field.

Here are key takeaways that caught our attention and underline how fast the sector is advancing:

Research and Development: The report demonstrates an overall increase in AI-related research activities. There was a significant boost for participation in AI research conferences in 2020 as “the number of attendees across nine conferences almost doubled in 2020”, likely because geographical barriers were removed when most conferences took place virtually.¹ Research output has increased overall by more than sixfold in the last five years from 5,478 AI-related publications on arXiv in 2015 to 34,736 in 2020.”²

Technical Performance: AI content generation is more accessible than ever before with systems composing text, audio, and images to a sufficiently high standard that humans have a hard time telling the difference between synthetic and non-synthetic outputs for some constrained applications of the technology.”³ Progress in this area, as is noted for natural language processing (NLP), outpaces the benchmarks set to test the advances.

Investment and the Economy: Acceleration in AI investment has not been slowed by the pandemic. A McKinsey survey showed no negative effect on corporate AI investment based on the recent economic downturn – “more than 25% of businesses actually reported increasing their AI investment in 2020.”⁴ The sector with the largest share of private AI investment was in “Drugs, Cancer, Molecular, Drug Discovery” with more than USD 13.8 billion in 2020 (4.5 times higher than 2019).⁵ These numbers helped AI drug discovery overtake autonomous vehicles as the most funded category in 2020.

AI Education: Corporate AI adoption has provided more industry employment options for new graduates. The annual survey from the Computing Research Association (CRA) showed that fewer graduates are opting for jobs in academia with the “share of new AI PhDs choosing industry jobs increasing by 48%.”⁶

Diversity in AI: “The people building AI systems are not representative of the people those systems are meant to serve.”⁷ We share the concerns raised by the authors regarding the continued lack of diversity in the field. The consequences have wide-ranging implications: “this generalized lack of diversity in race and ethnicity, gender identity, and sexual orientation not only risks creating an uneven distribution of power in the workforce, but also, equally important, reinforces existing inequalities generated by AI systems, reduces the scope of individuals and organizations for whom these systems work, and contributes to unjust outcomes.”⁸

 

1. See Report Highlights, Research & Development, p. 10
2. See Chapter 1, Highlights, p. 17
3. See Report Highlights, Technical Performance, p. 10
4. See Chapter 3, The Effect of COVID-19, p. 103
5. See Chapter 3, Focus Area Analysis, p. 97
6. See Chapter 4, Industry vs. Academia, p. 118
7. See Chapter 6, Overview, p. 137
8. See Chapter 6, Overview, p. 137

AI News This Week

  • Could The Simpsons Replace Its Voice Actors with AI?   (Wired)

    As the show pushes through its fourth decade, it’s the iconic voices on The Simpsons that could pose the biggest threat to its staying power as its actors reach retirement age. Advances in computing power and ‘deepfake’ technology can make convincing lookalikes – or in this case soundalikes – from a limited amount of training data. The producers of the show have 30 years worth of audio to work from.

  • China Targets AI, Chips Among Seven Battlefronts in Tech Race With US  (Wall Street Journal)

    China upped the stakes in its tech race with the U.S. as leaders laid out plans to speed up development of advanced technologies such as chips, artificial intelligence and quantum computing over five years. China’s leaders are pushing to rival the U.S. in cutting-edge technologies and in developing an independent supply chain to wean companies off dependence on American suppliers. The Biden administration is conducting a broad review of U.S. policy on China technology and is seeking to enlist allies to stay ahead of Chinese advances.

  • Drive-Throughs That Predict Your Order? Restaurants Are Thinking Fast   (New York Times)

    Inspired by pandemic lessons that kept customers in cars, restaurant chains are adding more lanes and curbside pickup, improving apps and testing menu boards that use AI. The experience of a drive-through is poised to be altered for the first time in decades. AI is helping chains suggest foods that are particularly popular in the area that day and use outside factors, including the weather, to highlight items like an iced coffee on a hot day. Chains such as McDonald’s and Burger King are using predictive personalized digital promotions to customers based on previous purchases.

  • Algorithm helps artificial intelligence systems dodge “adversarial” inputs  (MIT News)

    A new deep-learning algorithm developed by MIT researchers is designed to help machines navigate in the real, imperfect world, by building a healthy “skepticism” of the measurements and inputs they receive. The team combined a reinforcement-learning algorithm with a deep neural network to build an approach they call CARRL, for Certified Adversarial Robustness for Deep Reinforcement Learning. This approach may help robots safely handle unpredictable interactions in the real world like in self-driving cars.

  • Hundreds of sewage leaks detected thanks to AI  (BBC News)

    Scientists identified 926 spills from two wastewater treatment plants over an 11-year period by employing machine learning. The researchers trained a pattern recognition algorithm to recognize, through the pattern of flow through a treatment plant, when a spill was happening. The researchers say that water companies around the UK could put a similar approach in place to prevent future spills.

Radical Reads is edited by Leah Morris (Senior Director, Velocity Program, Radical Ventures).