Role: Design lead
Analog AI is a demo created for the annual Neural Information Processing System conference. The demo showcases the Fusion chip, developed by IBM Research Zurich, which uses physical attributes of materials to increase the speed and energy-efficiency needed for the next steps in AI.
The Fusion chip implements an artificial neural network in a piece of in-memory computing hardware. It does this exploiting the storage capability and physical attributes of phase-change memory (PCM) devices.
The demo received a Research Division Award for giving Analog AI high visibility leading to continued engagement between the research team and outside clients.
Designed at IBM Research
Role: Design lead
Analog AI is a demo created for the annual Neural Information Processing System conference. The demo showcases the Fusion chip, developed by IBM Research Zurich, which uses physical attributes of materials to increase the speed and energy-efficiency needed for the next steps in AI.
The Fusion chip implements an artificial neural network in a piece of in-memory computing hardware. It does this exploiting the storage capability and physical attributes of phase-change memory (PCM) devices.
The demo received a Research Division Award for giving Analog AI high visibility leading to continued engagement between the research team and outside clients.
Designed at IBM Research
Visualizing the process
There are many demos illustrating how neural networks guess a number drawn by a user. What’s unique about Analog AI is the number is processed on the actual chip at IBM’s lab in Zurich. To reinforce that this was real, we show the actual electrical path of different numbers as they’re processed on the chip. The simple interface first prompts a user to draw a number, then they see their number processed in the chip and the system’s guess.
How do PCMs work?
A PCM device is a nanometric volume of phase-change material that is sandwiched between two electrodes. With PCM, when an electrical pulse is applied to the material, it changes the conductance of the device by switching the material between amorphous and crystalline phases. A low electrical pulse will make the PCM device more crystalline (less resistance). A high electrical pulse will make the device more amorphous (more resistance). Therefore, instead of recording a 0 or 1 like in the digital world, it records the states as a continuum of values between the two–the analog world.
By leveraging the physical properties of PCM devices, computation happens at the same place where the data is stored, drastically reducing energy consumption. Because there is no movement of data, tasks can be performed in a fraction of the time and with much less energy. Also, PCM does not consume power when the devices are inactive, and the data will be retained for up to 10 years even when the power supply is turned off.
Analog AI and neural networks
In deep learning inference, data propagation through multiple layers of a neural network involves a sequence of matrix multiplications, as each layer can be represented as a matrix of synaptic weights. On the Fusion chip, these weights are stored in the conductance states of PCM devices. The devices are arranged in crossbar arrays, creating an artificial neural network where all matrix multiplications are performed in-place in an analog manner. This structure allows inference to be performed using little energy with high areal density of synapses.