The future of artificial intelligence (AI) could depend on an ambitious, yet uncertain form of energy that doesn’t quite exist today, at least according to OpenAI CEO Sam Altman.
Speaking during events at the World Economic forum in Davos, Altman revealed that it was his belief that future AI would require massive amounts of power and that “there’s no way to get there without a breakthrough.”
He went on to explain that this challenge “motivates us to go invest more in fusion.”
Current methods for developing the world’s most popular large language AI models, such as OpenAI’s ChatGPT and Anthropic’s Claude, involve the combined use of 10s of thousands of graphical processing units (GPUs), similar to the ones used in gaming PCs.
The power requirements for these training cycles are incredibly high compared to most similar applications. Even post-training, the largest LLMs use up a significant amount of resources.
Related: Elon Musk ‘uncomfortable growing Tesla to be a leader in AI’ without 25% voting control
Experts, including Sam Altman, predict that we’ll either need to revolutionize how we train and run the largest AI models or we’ll need to find a way to create more sustainable energy than is currently possible using today’s technologies.
The concept of fusing atoms together to create energy goes back to 1879 and the discovery of plasma, a form of ionized gas, by British chemist Sir William Crookes.
Fusion is achieved when two lighter atoms are, essentially, fused together to form a heavier atom. When this occurs, the atom releases energy.
The current challenge in the field is to get fusion reactors to produce more energy than it takes to power them.
Fusion promises to be a zero-carbon solution for upscaling power grids. However, much like human-level AI, there’s no consensus on when or if a working fusion
Researchers believe that it will be achieved sometime between the next few years and, perhaps, not within our lifetimes.
Altman invested $375 million of his personal funds in a fusion startup called Helion in 2021. Initially, Helion claimed it would have a working fusion reactor prototype producing positive energy by 2024, but it’s unclear if this estimate still holds.
At the time, it was the biggest investment Altman had ever made and represented the lion’s share of Helion’s total $500 million fundraise.