A new programme that uses cutting-edge picture recognition AI to envision the future consequences of climate change on any location on the planet, including your own house, has been released.
The initiative, named “This Climate Does Not Exist,” allows you to enter the address of your present residence or a beloved vacation spot and see what it may look like years from now, when climate change has taken its toll.
You might imagine how Disneyland would seem if it were to be engulfed in haze, similar to how Beijing was enveloped in smog in 2014. You may see what your childhood house would look like if it is swamped by increasing sea levels, similar to how Indonesia was ravaged by floods in 2020 as a result of massive deforestation.
Climate change has already had an influence on extreme weather occurrences in several parts of the world.
Researchers looked at more than 350 peer-reviewed articles looking at weather patterns and mapped out extreme weather occurrences in a separate effort by Carbon Brief, a UK-based site focused on climate science and policy. From heatwaves in France to hurricanes in the Caribbean, they discovered that 70 percent of the 405 extreme weather trends they studied were made more frequent or severe by human-caused climate change.
Similarly, the initiative “This Climate Does Not Exist” underlines that climate change is already having disastrous impacts all across the world, even if it isn’t in your neighborhood.
The initiative, on the other hand, is not intended to make you feel down. It was created by a group of AI scientists to increase awareness of the human impacts of climate change and to offer a set of tangible steps you can do to assist.
“The visuals may be eye-opening,” Sasha Luccioni, a postdoctoral researcher at Mila, the Quebec AI Institute behind the climate project, says. “But then you take that and you learn about what you can do and how you can be engaged.”
After you’ve seen the visualisation, you may share it with others by sending a link to the pictures. Additionally, the site demonstrates how you may take both collective and individual steps to mitigate the consequences of climate change, such as participating with your legislators and modifying your diet or consumption patterns.
“It was critical to frame the entire endeavour and ensure that the photographs were being used in a constructive or beneficial way. That is why we collaborate with climate change communicators “According to Luccioni.
How the technology was created
The project began in 2019 and will be released in 2021 in conjunction with the COP26 climate summit in Glasgow.
According to MIT Technique Review, the notion was inspired by a technology called generative adversarial networks, or GANs, which was invented by researcher Ian Goodfellow following a “heated dispute in a Montreal pub.”
GANs were first used to create deepfakes, or fake pictures that appear to be extremely realistic. In the past, MIT has used technology to create images of horses wearing hats, as well as renderings of your neighbourhood as a battle zone, in an attempt to humanise the Syrian Civil War’s atrocities.
“Why can’t we utilise it for climate change as well?” Mila researchers reasoned.
The Mila research institute’s experts found creating these scenarios to be a fun endeavour.
Smog was a little easier. The model’s major task was to recognise items and their relative sizes, after which it could estimate how the smog should appear dependent on the viewer’s distance.
According to Luccioni, the most difficult part was creating a phoney impression of a flooded region.
“Data collection took a long time since we anticipated we’d be able to acquire a lot of it at initially. However, it turns out that when a flood occurs, for example, individuals are less likely to stay. They’re going to depart. There are other flood photographs, although they’re generally from helicopters “Luccioni elucidates.
When the Mila team ran out of real-world photographs to train their model on, they resorted to virtual images created by local video game developers. They requested the creators of the video game Watchdogs to fill the virtual environment they created of the San Francisco Bay Area with people and capture images to feed Mila’s models with data.
Furthermore, the Mila researchers went on television to promote their experiment, encouraging Canadians to share photographs of recent floods. They were able to collect around 5,000 photographs after their diligent efforts.
Mila researchers were able to produce these photos using a GAN they constructed with the help of a team of over 20 machine learning professionals lead by scientific director Yoshua Bengio. They worked with climate communicators and scientists to develop messages that they think would inspire good change.