Recently, IBM made an announcement at the IEEE CAS/EDS AI Compute Symposium regarding new partnerships and technologies. New analog AI is being developed at IBM by combining memory and compute in a single device designed to alleviate a limitation that is dictated by conventional hardware technologies. In the conventional systems, the memory is segregated in various locations, with data moving between them when there is an operation. VP of IBM systems stated that the IBM Systems Research is working tirelessly to find new ways to create powerful and more efficient AI systems with analog and digital cores. He further added that the Analog Hardware Acceleration Kit's release as an open-source would help the developer community. The starter kit offered by IBM will have PyTorch, an analog device simulator.
This kit will help in developing new AI apps for NLP and computer vision. Customization options are also flexible for users to configure the material parameters. The starter kit's main mission is to enlist the AI communities to evaluate the potential of the analog hardware. Red Hat has partnered with IBM to integrate the company's AI core into Red Hat's OpenShift. Synopsys will prove to be an EDA partner for the company's Hardware Center.