It has been just revealed the fact that IBM made a new and effective AI Processor. Check out the latest reports about this below.
IBM and its brand-new processor
As the use of AI systems has grown, so has their energy demand. Training new systems requires massive data sets and a lot of processor time, making it incredibly energy-intensive. However, executing a trained system tends to be less intensive and can be easily managed by smartphones in some cases. Nevertheless, due to the frequency of execution, the energy use can add up quickly.
We are excited to announce that there are numerous ideas on how to decrease the energy consumption of running these systems. IBM and Intel have conducted experiments with processors that imitate the behavior of real neurons.
IBM has even tested performing Neural network calculations in phase change memory to prevent the need for repeating trips to RAM.
Now, IBM has come up with a new approach, called the NorthPole processor.
It combines some of the concepts behind all of these approaches with a simplified method of running calculations, creating a highly power-efficient chip that can execute inference-based Neural Networks more efficiently.
For tasks like image classification or audio transcription, the chip can be up to 35 times more efficient than relying on a GPU. It’s important to clarify a few things at the beginning of this discussion.
It’s important to note that NorthPole is not designed to assist with the energy demand in training a neural network. Its main purpose is to execute trained neural networks.
It’s also worth noting the fact that NorthPole is not a general AI processor. It is specifically designed for neural networks that focus on inference. Inference involves identifying the content of an image or audio clip, which can be useful in a variety of applications.
On the other hand, according to the online publication Ars Technica, if you require a large language model, NorthPole will not be able to meet your needs.
More than that, it is also important to reveal the fact that NorthPole draws inspiration from neuromorphic computing chips, such as IBM’s TrueNorth.
According to the same publication mentioned above, it has been revealed the fact that, unlike neuromorphic hardware, NorthPole’s processing units perform calculations instead of attempting to mimic the spiking communications of actual neurons.
We suggest that you check out the complete original article in order to learn more details about this.
The post IBM Makes New Efficient AI Processor appeared first on Health Thoroughfare.
