As artificial intelligence (AI) continues to develop and demonstrate even more intelligent capabilities, the amount of energy, water, and other resources required to make the needed complex computations continues to rise. By 2026, the energy consumption from data centers and AI is expected to be double that of 2022 levels. This rapid increase in consumption is unsustainable and is exacerbating climate change and environmental degradation, defeating the purpose of using AI for the benefit of humanity.
AI chatbots take, on average, 10 times as much electricity to answer questions as a Google search. This is because the constantly undulating nature of an AI model requires the storage of massive training datasets and up to trillions of parameters to process queries. The International Energy Agency projects that an integrated AI model may demand the same amount of electricity consumed by 1.5 million people, the population of Bahrain.
“The biggest players in AI are building new data centers that could tap out entire country’s electricity grids,” theorizes Binoy Syed, co-founder and COO of AI startup Simuli. While AI energy consumption is expected to grow to 160% of its current levels in the next decade, energy grids are struggling to match that pace.
AI data farms not only directly contribute to global warming by releasing heat and utilizing fossil fuels, but they also take away energy from the people who live around them. As the largest issue facing the expansive integration of AI into daily life by far, their energy consumption has already resulted in reports of rising electricity bills and blackout risks in residential areas. “There’s a genuine constraint on everyday lives and infrastructures. The superfluous consumption of power that AI needs is hurting our environment,” Binoy explains.
To drive forward the AI industry and, along with it, the rest of the world, its energy needs must be drastically reduced. This is because right now, the massive data centers they require can only be built by the biggest technology companies in the space. Once they make their expensive gamble, other companies may be less inclined to make the same investment for a relatively small ROI. So, its current energy usage could also seal off the potential for AI’s future.
Their immensely expensive and inefficient use of energy means that AI will be difficult to implement in developing countries, marking the start of an already unequal distribution of innovation and progress. This begs the question, “If the intelligence of these systems needs ridiculously unsustainable amounts of energy, are they really that intelligent?” asks Binoy. Now, Simuli’s Hyperdimensional AI approach has the answer.
To ensure the continued development and scaling of AI while reducing its negative environmental impacts, Simuli is creating a way to exponentially grow computing power without increasing energy consumption at the same rate. Using lossless data-agnostic compression technology, Simuli’s application programming interface (API) will allow AI machines and data centers to directly make computations without the need to decompress the data. Its holographic compression technology is a full preprocessing service, providing a one-stop data orchestration shop.
“Necessity is the mother of invention, and invention is necessarily intelligence,” says Binoy. According to Simuli, compressed computing is more climate-friendly, estimated to reduce power consumption by up to 99% and memory consumption by up to 178 times. This reduction in computing resources would make AI capabilities scalable and more affordable, democratizing the space rather than leaving it solely in control of Big Tech.
“Everyone should have a voice in shaping their futures, and AI will play a significant role in that future,” Rachel St. Clair, PhD, co-founder and CEO of Simuli explains. The goal is for AI to enhance lives, reduce challenges, and alleviate suffering. As demonstrated, this mission is currently infeasible, but Simuli represents a shift. As Rachel says, “At the core of Simuli is our motivation to understand the fundamental truths of reality. We, as human beings, learn in a certain way because of our constrained resources. We can model the future of AGI on those principles to create a resource-efficient, and ultimately more intelligent, system.”
