AI and ML could save the planet – or add more fuel to the climate fire

'Staggering amount of computation' deployed to solve big problems uses a lot of electricity

AI is killing the planet. Wait, no – it's going to save it. According to Hewlett Packard Enterprise VP of AI and HPC Evan Sparks and professor of machine learning Ameet Talwalkar from Carnegie Mellon University, it's not entirely clear just what AI might do for – or to – our home planet.

Speaking at the SixFive Summit this week, the duo discussed one of the more controversial challenges facing AI/ML: the technology's impact on the climate.

"What we've seen over the last few years is that really computationally demanding machine learning technology has become increasingly prominent in the industry," Sparks said. "This has resulted in increasing concerns about the associated rise in energy usage and correlated – not always cleanly – concerns about carbon emissions and carbon footprint of these workloads."

Sparks estimates that AI/ML workloads account for more than half of all compute demand today.

"The big issue is that many high-profile ML advances just require a staggering amount of computation," said Talwalkar, who also works as an AI researcher at HPE, citing an OpenAI blog post from 2018 that showed that the compute and energy requirements for the model had increased more than 300,000 times since 2012.

"That's a figure, at this point, that's almost four years old, but I think the trend is continuing along similar directions," he added.

However, the fact the broader machine learning community is even thinking about the effect of AI on the climate is a promising sign, Talwalkar noted.

"This wasn't something that we were really thinking about in the machine learning community a few years ago," he said. "It's good to get ahead of this issue and put pressure on ourselves as a community."

It's not too late

Confronting the environmental ramifications of AI proliferation first requires a better understanding of the problem itself, Talwalkar explained.

"This means learning to accurately measure the exact degree to which this is a problem both in terms of the energy requirements of current AI workloads, as well as coming up with accurate predictions of what we expect future requirements to look like," he said, adding that these insights will not only help researchers understand the true cost of a workload, but also take steps to develop more efficient hardware and improve the algorithms.

"We're actively in the midst of hardware proliferation in terms of specialized hardware specifically designed for training and/or deployment of machine-learning models," he said, citing Google's tensor processing unit as an early example and pointing to ongoing efforts by Nvidia, Graphcore, Cerebras, and others to develop novel hardware for machine learning and AI workloads.

"It's tempting to throw more hardware at the problem, but I think simultaneously as a research community we're really pushing the envelope as well on the algorithmic advances," Sparks noted, highlighting the equal importance of software.

In this regard, Talwalkar argues a better understanding of how and why deep learning models work could bear fruit for optimizing the algorithms to eke out more performance from the compute resources available.

AI is in its infancy

Despite the challenges, Talwalkar remains optimistic that the community will rise to the occasion, and, as the technology matures, place less emphasis on what we can do with these workloads and increase efforts to optimize them.

"We're certainly in the early days of AI progress," he explained. "It seems like we’re seeing new applications showing up daily that are pretty amazing."

Talwalkar believes AI and ML will follow a path not unlike that of the Human Genome Project – a massively expensive endeavor that laid the groundwork for low-cost gene sequencing that has proven enormously beneficial.

And while Sparks expressed similar optimism, he doesn't expect AI/ML growth to abate any time soon. "At least for the next few years, we're going to see a lot more – not a lot less." ®

Broader topics

Other stories you might like

  • Is computer vision the cure for school shootings? Likely not
    Gun-detecting AI outfits want to help while root causes need tackling

    Comment More than 250 mass shootings have occurred in the US so far this year, and AI advocates think they have the solution. Not gun control, but better tech, unsurprisingly.

    Machine-learning biz Kogniz announced on Tuesday it was adding a ready-to-deploy gun detection model to its computer-vision platform. The system, we're told, can detect guns seen by security cameras and send notifications to those at risk, notifying police, locking down buildings, and performing other security tasks. 

    In addition to spotting firearms, Kogniz uses its other computer-vision modules to notice unusual behavior, such as children sprinting down hallways or someone climbing in through a window, which could indicate an active shooter.

    Continue reading
  • Microsoft promises to tighten access to AI it now deems too risky for some devs
    Deep-fake voices, face recognition, emotion, age and gender prediction ... A toolbox of theoretical tech tyranny

    Microsoft has pledged to clamp down on access to AI tools designed to predict emotions, gender, and age from images, and will restrict the usage of its facial recognition and generative audio models in Azure.

    The Windows giant made the promise on Tuesday while also sharing its so-called Responsible AI Standard, a document [PDF] in which the US corporation vowed to minimize any harm inflicted by its machine-learning software. This pledge included assurances that the biz will assess the impact of its technologies, document models' data and capabilities, and enforce stricter use guidelines.

    This is needed because – and let's just check the notes here – there are apparently not enough laws yet regulating machine-learning technology use. Thus, in the absence of this legislation, Microsoft will just have to force itself to do the right thing.

    Continue reading
  • Cerebras sets record for 'largest AI model' on a single chip
    Plus: Yandex releases 100-billion-parameter language model for free, and more

    In brief US hardware startup Cerebras claims to have trained the largest AI model on a single device powered by the world's largest Wafer Scale Engine 2 chip the size of a plate.

    "Using the Cerebras Software Platform (CSoft), our customers can easily train state-of-the-art GPT language models (such as GPT-3 and GPT-J) with up to 20 billion parameters on a single CS-2 system," the company claimed this week. "Running on a single CS-2, these models take minutes to set up and users can quickly move between models with just a few keystrokes."

    The CS-2 packs a whopping 850,000 cores, and has 40GB of on-chip memory capable of reaching 20 PB/sec memory bandwidth. The specs on other types of AI accelerators and GPUs pale in comparison, meaning machine learning engineers have to train huge AI models with billions of parameters across more servers.

    Continue reading

Biting the hand that feeds IT © 1998–2022