Massively parallel computing on GPUs, popular in AI neural networks, may be replaced by graph processors, as many AI datasets aren’t dense and require lower precision. The US Defense Advanced Research Projects Agency is exploring this through the Hierarchical Identify Verify Exploit project, which includes Intel creating a massively parallel graph processor codenamed Puma. Now, Puma provides upgraded performance for graph analytics, using custom RISC instruction set cores interconnected via Intel’s royalty-free chiplet interconnecting Advanced Interface Bus.
Open source leaders are pushing for permanent government financing to maintain open source projects. They argue that as government oversight is vital to the national