AI-Powered Solutions for a Sustainable Future | Intel
853
12.9
Intel560 тыс
Следующее
Популярные
86 дней – 1 40723:34
140 дней – 7673:48Опубликовано 18 июля 2024, 23:22
As AI lays its roots deeper into our modern society, scientists and big-thinkers have begun to clamor over its potential world benefits. One of the more exciting opportunities growing out of this is the way in which AI could help us combat climate change. Many experts believe that deep learning AI algorithms can play a key role in tackling environmental concerns, from designing more energy-efficient cities, to large-scale monitoring of global emissions, and even optimizing our renewable energy deployment.
One specific use of AI is to help us cultivate beneficial marine algae such as seaweed and kelp. These algae not only provide tremendous potential as a sustainable food source but can also work to trap carbon from our atmosphere and bury it deep beneath the ocean floor. In fact, some studies estimate that kelp forests alone capture around 4.5 million metric tons of carbon dioxide from seawater every year.
But as sea temperatures rise and kelp forests decline, scientists are flocking to find ways to restore our dwindling kelp supplies. By applying the use of AI technologies, conservationists are able to supercharge their efforts, parsing through oceans of data in order to generate predictive sea maps, identify key sites for future cultivation, and detect and prevent diseases before they can occur.
But the use of AI itself may come with its own environmental cost, with the energy required to train and run these algorithms often hitting staggering levels. Recent studies have shown a worrying trend amongst developers to focus on AI solutions that favor accuracy over efficiency, with many turning to power-hungry solutions such as neural network architecture to drive their models. In fact, one study calculated that training a single neural net transformer model generates around 626 thousand pounds of CO2 emissions - roughly 5 times the lifetime emissions of the average American car (including its manufacture!). And as researchers continue to test, develop, and retrain their algorithms, it’s easy to see how these numbers can quickly add up.
So, what are sustainably-minded technologists to do?
Fortunately, some researchers are looking to quell this trend, by advocating for efficiency to be used as a key evaluation criterion in measuring a model’s success. Supporting this, solutions-minded companies like Intel are developing more ‘power-efficient machine learning’ techniques, that work by weeding out unnecessary content from training datasets while they are still in a compressed form. The result is a vast improvement in power efficiency when training a model, while still being able to maintain accuracy.
“The solution to making more sustainable AI is both a software and a hardware solution”, says Merlin Kister, Senior Director at Intel. “It’s also taking a look at the model that you have to start with and make sure that gets right sized.”
Another exciting possibility could come from the development of more probabilistic-based AI models. These new models, such as those being developed at Intel, would allow for much more flexibility in an AI’s ability to adopt abstract reasoning, allowing for the expression of large amounts of logical processing, but without the need for brute-forced inference that relies on thousands or millions of data points.
As Merlin Kister puts it, “Intel engineers are working hard every day to develop innovative new features and technologies to put into our processors to make them more energy efficient so they can be more sustainable.”
Learn how to make AI work for your organization: intel.com/content/www/us/en/ar...
Subscribe now to Intel on YouTube: intel.ly/3IX1bN2
About Intel:
Intel, the world leader in silicon innovation, develops technologies, products and initiatives to continually advance how people work and live. Founded in 1968 to build semiconductor memory products, Intel introduced the world's first microprocessor in 1971. This decade, our mission is to create and extend computing technology to connect and enrich the lives of every person on Earth.
Connect with Intel:
Visit Intel WEBSITE: intel.ly/Intel
Follow Intel on X: intel.ly/Twitter
Follow Intel on INSTAGRAM: intel.ly/Instagram
Follow Intel on LINKEDIN: intel.ly/LinkedIn
Follow Intel on TIKTOK: intel.ly/TikTok
AI-Powered Solutions for a Sustainable Future | Intel
youtube.com/intel
One specific use of AI is to help us cultivate beneficial marine algae such as seaweed and kelp. These algae not only provide tremendous potential as a sustainable food source but can also work to trap carbon from our atmosphere and bury it deep beneath the ocean floor. In fact, some studies estimate that kelp forests alone capture around 4.5 million metric tons of carbon dioxide from seawater every year.
But as sea temperatures rise and kelp forests decline, scientists are flocking to find ways to restore our dwindling kelp supplies. By applying the use of AI technologies, conservationists are able to supercharge their efforts, parsing through oceans of data in order to generate predictive sea maps, identify key sites for future cultivation, and detect and prevent diseases before they can occur.
But the use of AI itself may come with its own environmental cost, with the energy required to train and run these algorithms often hitting staggering levels. Recent studies have shown a worrying trend amongst developers to focus on AI solutions that favor accuracy over efficiency, with many turning to power-hungry solutions such as neural network architecture to drive their models. In fact, one study calculated that training a single neural net transformer model generates around 626 thousand pounds of CO2 emissions - roughly 5 times the lifetime emissions of the average American car (including its manufacture!). And as researchers continue to test, develop, and retrain their algorithms, it’s easy to see how these numbers can quickly add up.
So, what are sustainably-minded technologists to do?
Fortunately, some researchers are looking to quell this trend, by advocating for efficiency to be used as a key evaluation criterion in measuring a model’s success. Supporting this, solutions-minded companies like Intel are developing more ‘power-efficient machine learning’ techniques, that work by weeding out unnecessary content from training datasets while they are still in a compressed form. The result is a vast improvement in power efficiency when training a model, while still being able to maintain accuracy.
“The solution to making more sustainable AI is both a software and a hardware solution”, says Merlin Kister, Senior Director at Intel. “It’s also taking a look at the model that you have to start with and make sure that gets right sized.”
Another exciting possibility could come from the development of more probabilistic-based AI models. These new models, such as those being developed at Intel, would allow for much more flexibility in an AI’s ability to adopt abstract reasoning, allowing for the expression of large amounts of logical processing, but without the need for brute-forced inference that relies on thousands or millions of data points.
As Merlin Kister puts it, “Intel engineers are working hard every day to develop innovative new features and technologies to put into our processors to make them more energy efficient so they can be more sustainable.”
Learn how to make AI work for your organization: intel.com/content/www/us/en/ar...
Subscribe now to Intel on YouTube: intel.ly/3IX1bN2
About Intel:
Intel, the world leader in silicon innovation, develops technologies, products and initiatives to continually advance how people work and live. Founded in 1968 to build semiconductor memory products, Intel introduced the world's first microprocessor in 1971. This decade, our mission is to create and extend computing technology to connect and enrich the lives of every person on Earth.
Connect with Intel:
Visit Intel WEBSITE: intel.ly/Intel
Follow Intel on X: intel.ly/Twitter
Follow Intel on INSTAGRAM: intel.ly/Instagram
Follow Intel on LINKEDIN: intel.ly/LinkedIn
Follow Intel on TIKTOK: intel.ly/TikTok
AI-Powered Solutions for a Sustainable Future | Intel
youtube.com/intel
Свежие видео
10 дней – 3064:33
16 дней – 4 4303:09
17 дней – 1 5740:14
Случайные видео
05.04.23 – 32 1501:00
25.03.23 – 38 2223:42
22.01.22 – 888 41310:20
20.10.20 – 3 385 31314:00
09.12.16 – 2 1513:03
3 дня – 1170:24
4 дня – 35940:23
4 дня – 2988:04
5 дней – 112 07414:24
10 дней – 1 0190:15
6 дней – 125 55814:42
8 дней – 1 1344:11