Untether AI CEO Arun Iyengar touts an in-memory computing chip aimed at reducing AI’s energy usage. CITYNEWS/Daniel Berry
Artificial intelligence (AI) has become a part of our daily lives and even more mainstream, with applications like ChatGPT, used in the automotive industry for streaming platforms, and more. But for all its benefits, AI also has a noticeable impact on the environment.
Arun Iyengar, CEO of Toronto-based technology company Untethered AI, explains that traditional AI methods are a huge waste of the world’s energy resources.
“If you look at just January and look at how many people used ChatGPT and how much energy it takes to serve them, you can see how much energy it takes to fully service a town of 175,000 people. are the same,” he says.
That’s because it uses methods established by John von Neumann in the 1940s. John von Neumann is widely regarded as the father of modern AI technology and has powered every computer his chip has to date.
“Every time you move data, you consume energy. So the longer you move data, the more power it takes to make it happen. That’s why today’s implementations consume so much of electricity will be consumed.”
Iyengar’s company, a semiconductor startup, has created a first-of-its-kind chip aimed at dramatically reducing AI’s energy use.
“We basically put processing and memory next to each other. So moving data becomes contained within a very short distance, and the power or energy required to run artificial intelligence is 6 It will be reduced by a factor of ~10,” he says.
“We design the chip, we manufacture it, we deliver it to large customers, who can implement it in servers if they are in a ChatGPT type environment for example, in self-driving cars, etc. can.”
He explains that companies can also increase what and how much they can do with AI.
“With the traditional approach, 90% of the energy was wasted moving data.
“As an example, we’ve worked with General Motors. In an environment like a car, there’s a limit to how much power you have. more and more [processing capability] can provide the same amount of power [companies] We can expand,” he says.
In other words, you get more bells and whistles without increasing the amount of energy you use.
But why should ordinary Canadians invest in these technical aspects of AI?
Iyengar says that as artificial intelligence becomes more prevalent, the amount of power it uses will ultimately affect us all.
“We have to be very responsible for how we actually power up. Right now about 1% of the world’s energy goes to our servers. [used for AI]With continued AI adoption it will be around 15% [by the end of the decade]Where does it come from? Of course, to families in colder parts of the country, I can’t say, ‘Sorry, I’m powering his ChatGPT over there, so it won’t be powered today,'” he said. is joking.
Iyengar adds that big tech companies have an important role to play in reducing the environmental impact of AI.
“What big tech companies are trying to do today is bring their products to market,” he says.
“When starting a very large deployment, it’s important to have a clear picture of the environmental impact, and it’s the duty of any large company to say, ‘Hey, I know how to deploy, so I’m just going to deploy , let us unfold in a sustainable way.” ”
