What’s new? : Deep learning is the reason behind the most notable success of machine learning. But this incredible performance comes at a price: training deep learning models requires a lot of energy. Now, new research shows that scientists who train algorithms using a cloud platform can significantly reduce the energy they use, and therefore the emissions they produce.
How can they do it? : Simple changes to cloud settings are key. The researchers created a tool that can measure the electricity usage of any machine learning program running on Microsoft’s cloud service Azure at every stage of the project. They estimated emissions based on the zip codes of the servers running the 11 machine learning models, and found that emissions could be reduced significantly if the researchers adjusted the settings to use the servers in specific geographic locations and at specific times of day.
Big picture: Giving people the option to adjust their settings is an uphill battle. Since its launch in October, only 13% of Azure users running machine learning programs have used the energy measurement tool, so the next step will be to convince other users. Read the full article.
– Xu Ming
The world will need dozens of breakthrough climate technologies in the next decade
We are living in a pivotal decade. Global emissions must be cut in half by 2030, largely through the large-scale deployment of existing technologies such as wind turbines, solar panels and electric vehicles. But emerging climate technologies must also hit the market within this decade, even if they don’t have much immediate impact on emissions.
those of this year list MIT Technology Review 35 or less innovators are seizing the opportunity to decarbonize the economy and make the clean energy transition affordable.Read more about their work and what it takes to help them succeed this essay The author is Varun Sivaram, senior director of clean energy and innovation for the U.S. President’s special envoy on climate, John Kerry.
I combed the internet to find you the funniest/most important/scariest/most fascinating tech stories of the day.
11 billion Chinese leaked data online for more than a year
It was left unnoticed in an insecure database until hackers offered to sell it. (CNN)
2 The Large Hadron Collider helps scientists discover three new particles
These combinations have never been seen before. (motherboard)
+ Don’t believe these misconceptions about the capabilities of the Hadron Collider. (think big)
3 How Wall Street emerged from the cryptocurrency carnage unscathed
It turns out that regulation is very convenient after all. (New York Times $)
+ It will also be used for cryptocurrencies. (wired $)
+ The cryptocurrency crash could be a setback for web3. (foot $)
+ Venture capitalists burned by the frenzied growth of the past decade are cautious. (motherboard)
+ Black investors suffer the most. (foot $)
+ It’s OK to Opt Out of the Crypto Revolution. (MIT Technology Review)
4 Europe has the green light for big tech regulation
However, it will be some time before the new law takes effect. (Love News)
+ The UK’s Online Safety Act has been tweaked to prioritise the detection of child abuse material. (protector)
+ The Supreme Court’s EPA ruling last week did not bode well for U.S. regulation. (protocol)
5 Microsoft is still using emotion-detecting AI
An app for the visually impaired — despite widespread doubts about the technology’s accuracy. (protocol)
+ Emotional AI researchers say exaggerated claims give their work a bad name. (MIT Technology Review)
6 How tech is saving Sri Lanka’s struggling tourism industry
Includes virtual cheetah tour. (Rest of the world)
7 Humans Shouldn’t Hibernate
But a few cases suggest that it is possible to enter a numbness-like state. (network)
8 Everything is an atmosphere these days
This suggests that it may be time to change the vibe – away from the vibe itself. (atlantic organization $)
10 The Sadness of Robot Existence
Maybe it’s time to stop projecting our own emotions onto them. (protector)
+ How we feel about robots. (MIT Technology Review)
+ That said, they are very good surgeons. (Spectrum IEEE)