Using AI art to promote an ultra-low-power speech-recognition processor?

Last year, Mr. X showed great enthusiasm for Syntiant’s NDP101 Neural Decision Processor, which is featured in two different boards, the NDP9101B0 and the TinyML Development Board.

Both are built to test Syntiant’s deep-learning speech-recognition chip, which boasts a 100x efficiency improvement when processing neural network instructions over traditional CPUs and DSPs without requiring a connection to the cloud, and consumes <140 µW when recognising words. The NDP9101B0 is implemented as a Raspberry Pi shield for easy testing of microphones and sensors. The TinyML Development Board, on the other hand, is a 24 mm x 28 mm self-contained platform that provides an ideal environment for performing low-power voice and acoustic event detection.

For a detailed rundown of either of these boards, and the NDP101 processor itself, follow the links below. Here, you can apply for a sample if you’ve got a commercially viable project that requires speech recognition, whether that be a smart home device, automotive design, or another HMI-heavy application. NDP9101B0 and TinyML

The main image for this post were generated by DALL.E, a free website-based AI image generator that can be found here for your curiosity and amusement.

The prompts used to generate these images reflect how we at ipXchange have come to view Syntiant’s NDP101. These included phrases such as:

-> “An artificial intelligence microchip mounted on a (printed) circuit board listening to [the world / people talk] (with a ghostly electronic ear), 3d render, red and black (colours)”
-> “A(n) (curious) artificial intelligence microchip mounted on a printed circuit board listening to the world (around it and understanding what it hears), anime”
-> or, much more directly: “Syntiant NDP chip” – Delete brackets as appropriate.

It’s perhaps a little unsettling how good AI art is becoming, but one would expect that DALL.E’s images are generated using vast, power-hungry servers that are constantly training to get better at what they do.

The NDP101 consumes far less power than cloud-based systems, but still manages to produce surprisingly high performance in certain AI applications. 100x efficiency improvement in a self-contained package is certainly an intriguing boast from Syntiant, so Mr. X reckons the NDP101 is definitely worth a closer look.

Keep designing!