"Revolution in AI Energy Consumption: BitEnergy AI's Breakthrough and the Future of Tech Giants"

On today's episode of Demo, we dive into the emerging realm of eco-friendly Artificial Intelligence (AI), as energy consumption by AI skyrockets to an alarming 134 TWh per year. We unveil how tech giants like Microsoft and OpenAI are grappling with the financial and reputational threat this poses. We spotlight BitEnergy AI's innovative algorithm, L-Mul, while discussing how it could decrease AI's energy utilization by a staggering 95%. The trick? Swapping energy-draining floating-point multiplications with simpler integer additions. Preliminary tests indicate AI performance remains steady, suffering from only a minute 0.07% drop in accuracy. Sometimes, there's even a minor enhancement. However, with GPUs currently optimized for floating-point multiplications, this change might ruffle some feathers, especially for firms like xAI that have invested heavily into their datacenter units. Finally, we speculate on whether L-Mul, compatible with liquid neural networks, could be a new pathway to curb AI's exponential adoption and its mounting energy consumption or whether deeper changes to our usage of technology need to be considered. Tune in for this energy-charged episode of Demo.