Artificial intelligence will have a profound impact on the everyday lives of humans and will also open up brand new business models and revenue streams for companies. Companies like Facebook, Amazon, Microsoft, Google, Alibaba, Tencent and Baidu are creating massive databases of faces, locations, habits, social graphs, etc. The Internet of Things is here and growing exponentially. Self-driving cars are on the market. Cloud computing is pervasive. And yet we have just begun to imagine how to leverage those inventions to create this new world of artificial intelligence, delivering new business models, new productivity paradigms, new ways of entertaining the masses that will not only have profound effects on mankind – but also generate enormous amounts of wealth for the innovators who can monetize that data.

Phase 1 was the gathering and storage of “big” data. But it’s a lot more than how big the data is that will drive the next wave of innovation. The winners in the race for AI will need four things: 1) access to lots of data, 2) access to the right data, 3) powerful algorithms that turn data into actionable insights and 4) the speed and scale to outdistance the competition. Today, data has the opportunity to serve as a profit center as it trains algorithms and creates new applications. The more data, the smarter the AI algorithms will be.

Only with new memory technologies such as Crossbar ReRAM memory technologies, can the true potential of artificial intelligence be realized, delivering the speed, low power/energy efficiency, storage capacity and manufacturability to make this new world a reality.

ReThink Artificial Intelligence with Memory Centric Computing

The radical shift from big data to AI comes when objects transition from simply fetching lines of code to processing, analyzing and acting on massive amounts of external data from sensors in real-time. Data coming from sensors can be stored on-chip and directly fed through deep neural networks to take direct actions. Non-volatile memory technologies such as Crossbar ReRAM are helping to address the performance and energy challenges by delivering lower power, faster read and byte-addressable writes. Crossbar ReRAM technology can be directly integrated on-chip with the processing logic enabling brand new memory-centric SoC architectures. On-chip integration with processor cores on a single chip solution makes ReRAM adequate to accelerate deep neural network algorithms. ReRAM technology is an important innovation to accelerate the potential of artificial intelligence universe, enabling a multitude of applications that can speed up performance, dramatically improve energy efficiency, enable advanced security and reduce chip count and size.

High performance computing applications, such as artificial intelligence require high-bandwidth low latency data accesses across processors, storage and IOs. Crossbar ReRAM memory technologies provide significant improvements by reducing the performance gap between storage and computing.

We are thrilled to see Crossbar achieve a significant industry milestone and produce its first working array in CMOS production facility. We are one step closer to achieving our goal of commercializing a completely new class of memory using a much simpler cell structure than traditional flash alternatives. Artiman is excited to continue working with Crossbar to revolutionize the memory industry.
Yatin Mundkur
Partner, Artiman Ventures