Automation and Jobs
Read our latest special issue.
A lot happened this week in the AI space. The Guardian wrote an article with GPT-3 and again demonstrated that no matter what OpenAI paid to train and create the language model, the free marketing might be worth more. After losing the JEDI cloud contract appeal with the Pentagon, Amazon appointed to its board Keith Alexander, who oversaw the National Security Agency mass surveillance revealed by Edward Snowden leaks in 2013. And Portland passed the strictest facial recognition bans in U.S. history, outlawing government and business use of the technology.
However, AI Weekly attempts to reach into the zeitgeist and highlight the issues on people’s minds. This week, it’s without question the smoke that has hung over the western United States and the underlying problem of climate change.
In the Bay Area Wednesday, the density of the smoke effectively blocked out the sun and cast a dark orange or red light. The eerie scene drew comparisons to Blade Runner 2049, among other sci-fi parallels.
Someone put Bladerunner 2049 music to drone footage of San Francisco and at first I didn’t know whether to be amazed or horrified. This is very much horrifying. pic.twitter.com/XQTv4qrE93
— Omar Jimenez (@OmarJimenez) September 10, 2020
The same day, the temperature was forecast to reach the high 80s or 90s but instead dropped to around 60, a nearly 30-degree decline in a single day. Street lights stayed on for most of the daylight hours, and disoriented people felt like it was night during the middle of the day. For some, San Francisco turning into Mars marked a breaking point, the final stressor after weeks of smoke and a heat wave — on top of ongoing efforts to establish racial justice, economic recovery, and a cure to COVID-19.
A quick recap: The August complex fire in northern California is now the largest in state history. CalFire said in its daily report today that 26 times as many acres have burned this year as the same time period in 2019. At one point on Wednesday, National Weather Service Bay Area tweeted that conditions were “beyond our models.” In Oregon, Governor Kate Brown said yesterday that 900,000 acres burned in three days. In a typical year, the state loses 500,000 acres to fire, she said. Today, unhealthy air quality stretched across the entirety of the West Coast, from San Diego to Seattle. On Friday, Portland recorded the worst air quality in the world. In neighboring states, Salt Lake City recorded historically bad air quality, and Denver residents were impacted as well.
The destruction and health risks caused by natural disasters are just one indication of climate change. The global environment is also showing strain: The World Wildlife Fund said in its annual Living Planet report monitoring 21,000 species of animals that wildlife population levels are down nearly 70% in the past 50 years. That loss of biodiversity poses a threat to global food supplies.
So what can we do about it? Climate Change AI is one of the major initiatives seeking to apply machine learning to the problem. The group of AI researchers is exploring solutions to climate change and adjacent problems like human displacement and food insecurity. In the past month, group members have discussed the kinds of startups that can combat climate change, and they are assembling a data set wish list from researchers to inform the data used to train models and solve problems.
In June 2019, researchers from more than a dozen organizations teamed up to release a paper with over 800 references that attempts to cover the myriad ways machine learning can help combat climate change. Areas of focus in the full paper and interactive summary include practical models for electric systems and smart cities, as well as long-term projects with uncertain impact, like CO2 sequestration or engineering a planetary control system.
At a Climate Change AI workshop held at the NeurIPS conference last year, researchers talked about the possibility of making AI a zero-carbon industry and looked at cultural changes that would allow the machine learning community to focus more of its attention on climate change. Researchers also circulated calculators for figuring out the carbon footprint of a machine learning model. Another Climate Change AI workshop is scheduled to take place at NeurIPS in December.
There’s also the work of WattTime, a nonprofit organization that reduces a household’s carbon footprint by automating when electric vehicles, thermostats, and appliances are active based on when renewable energy is available. Algorithms to determine those times are trained using data from the EPA’s continuous pollution monitoring system. The tech is currently available in California, where about 33% of power today comes from renewable energy as part of the Self-Generation Incentive Program, WattTime creator Gavin McCormick told VentureBeat.
“Nobody knows about the U.S. continuous emissions monitoring system, but it’s been live since the ’70s, and it’s why organizations like mine can write increasingly sophisticated AI algorithms to integrate more renewable energy and do what we do,” McCormick told VentureBeat in a phone call.
Last year, WattTime received a grant from Google.org’s AI Impact Challenge to see whether computer vision can track power plant emissions outside the U.S. from satellite imagery. In July, WattTime joined nine organizations and Al Gore to form Climate Trace, a group that wants to trace global emissions in major sectors of the economy, like power plants and shipping. Climate Trace’s goal is to make that data available to the public by June 2021, leading up to the next round of international climate negotiations.
When I was growing up, the idea that technology could change the world was an idyllic dream. In years past, there’s been a fair share of disillusionment, in part due to startups solving problems that don’t exist, mass surveillance, a general lack of funding for diverse startup founders, and a laundry list of transgressions by Big Tech companies too long to name here.
But if you’re feeling helpless about all the recent disasters and want to do something to change things, numerous projects (including the above) need volunteers. According to a Yale Climate Change Communication survey released in April, people are ready to get involved in efforts to compel elected officials to address climate change.
The word apocalypse came up a little too often for my liking this week. It’s easy to feel things are dire, chaotic, and out of control — because they are — but nobody should believe there’s nothing they can do to affect change. Technology alone won’t save us. We have to vote for elected officials whose policies take climate change seriously, and we also need to take individual action to support that change.
Know any other projects at the intersection of climate change and machine learning? Send news tips to Khari Johnson and Kyle Wiggers and AI editor Seth Colaner — and be sure to subscribe to the AI Weekly newsletter and bookmark our AI Channel.
Thanks for reading,
Senior AI Staff Writer