SingularityNET is placing its bets on a network of cutting-edge supercomputers to bring us closer to Artificial General Intelligence (AGI), with the first of these powerful machines set to go online this September.
While today's AI can achieve impressive feats—like GPT-4 crafting poetry or DeepMind's AlphaFold unlocking the secrets of protein structures—it still falls short of true human-like intelligence.
"Even though the novel neural-symbolic AI approaches we've been developing at SingularityNET reduce the demands for data, processing power, and energy compared to standard deep neural networks, we still require substantial supercomputing capabilities," SingularityNET CEO Ben Goertzel mentioned in a recent statement to LiveScience.
This is where SingularityNET’s bold plan comes into play: a “multi-level cognitive computing network” designed to support and train the extraordinarily complex AI systems needed for AGI. Picture deep neural networks that emulate the human brain, vast language models (LLMs) trained on massive datasets, and systems capable of integrating human-like behaviors such as speech and movement with multimedia outputs.
However, achieving this level of sophistication is far from cheap. The first supercomputer, expected to be ready by early 2025, will be a monstrous assembly of top-tier hardware, including Nvidia GPUs, AMD processors, and Tenstorrent server racks—a true Frankenstein’s monster of cutting-edge technology.
