Israel is considered among world leaders in AI advancement, however, data shows that the “Startup Nation” still has a ways to go to make it to the top. And now it seems as if the country is making initial moves to do so, by injecting over $1.5 billion, spread over 5 years, into the Israeli AI ecosystem.

First payment of $150M

The Israel Academy of Sciences authorized recommendations from a committee led by Dr. Orna Berry regarding the status of Israel’s AI sector and the need of government intervention. Due to the country not currently having a dedicated budget, and the  urgency of the matter, the committee recommended allocating an initial $150 million payment. The initial sum is part of an over $1.6 billion investment plan in the AI sector, with around 80% of the cash injection designated to close the technological gap between Israel and more advanced AI countries.

As part of the committee’s recommendations, urgent remediation to the sector has been defined and is aiming for immediate implementation. Among the decisions are the establishment of an over $80 million HPC infrastructure, $55 million for numerous R&D projects with a focus on NLP capabilities, and $31 million to train personnel and acquire advanced technologies from academia. In addition, the committee also recommended creating a supporting regulatory environment for promoting AI and for securing access to Israel’s public cloud project - Nimbus.

The committee also called for the development of an Israel based cloud service infrastructure, which will store sensitive information, complying with privacy, regulation, security, operational standards. Furthermore, the committee’s decisions will be presented to both the professional and political echelon in relevant government ministries, in an attempt to advance the most urgent recommendations by the beginning of 2021.

Super Computer: multi-processing core & powerful GPU

Phase 10 of the recommended program presents specs for the ideal “Super Computer”. The project demands 88,000 multi-processing cores (CPU) and 1,000 graphic processing units (GPU), at the level similar to NVIDIA’s A100 GPU server. In addition, the specs include dedicated AI acceleration units from various manufacturers, which will provide at least one petaFLOP performance and a rapid communication network. The network will allow to run parallelism Models as well as fast communication for CPUs and storage that supports GPU files.