Mercedes uses Akida, page-403

In addition to @@cyberworks post https://hotcopper.com.au/posts/58648035/single, a few thoughts of mine on this announcement from Daimler as referenced here https://media.daimler.com/marsMediaSite/en/instance/ko/VISION-EQXX--taking-electric-range-and-efficiency-to-an-entirely-new-level.xhtml

At first when it was mentioned (Rob?) that Brainchip was working with an automaker from Stuttgart, the first company that came to mind was Porsche (still can't rule them out I suppose).

1. Besides extolling the advancements made by their team internally, the press release references several suppliers they were working with -- HPP for batteries, Fraunhofer Institute for Solar Energy Systems ISE for solar cells, Bridgestone for tyres, AMsilk / Mylo / UBQ material for materials, Salzgitter Flachstahl GmbH / Rheinmetall Automotive for structural materials, NAVIS Automotive Systems, inc. for navigation systems, and of course Brainchip for neuromorphic processing.

If you care to research the above companies, you would find that the suppliers they have named and work with are leaders in their respective fields, and being named amongst them implies that Brainchip on top of our neuromorphic processor segment. This is despite of the fact and in light of the announcement that Mercedes-Benz is a partner in Intel's INRC neuromorphic community project, and that Loihi research chips are available to developers.

2. Looking at the portion of the press release dedicated to Brainchip (emphasis mine);

"

Neuromorphic computing – a car that thinks like you

Another key efficiency feature of the VISION EQXX that takes its cue from nature is the way it thinks. It uses an innovative form of information processing called neuromorphic computing. The hardware runs so-called spiking neural networks. Information is coded in discrete spikes and energy is only consumed when a spike occurs, which reduces energy consumption by orders of magnitude.

Working with California-based artificial intelligence experts BrainChip, Mercedes-Benz engineers developed systems based on BrainChip’s Akida hardware and software. The example in the VISION EQXX is the “Hey Mercedes” hot-word detection. Structured along neuromorphic principles, it is five to ten times more efficient than conventional voice control.

Although neuromorphic computing is still in its infancy, systems like these will be available on the market in just a few years. When applied on scale throughout a vehicle, they have the potential to radically reduce the energy needed to run the latest AI technologies.

"

To me, there's a lot to be read between the lines, and that's coming from someone who's not particularly a fan of wild speculation.

Working with California-based artificial intelligence experts BrainChip,

I think this speaks for itself, to me the DIRECT collaboration between the two companies (and not through a 3rd party like Cerence or Valeo) is clear and undisputable. For instance if I purchase a timing logic board containing a Texas Instrument chip for a product I'm manufacturing from Acme Ltd, I'd say I'm working with Acme, and it would be untrue to claim I'm working with TI .

Another key efficiency feature of the VISION EQXX that takes its cue from nature is the way it thinks. It uses an innovative form of information processing called neuromorphic computing. The hardware runs so-called spiking neural networks. Information is coded in discrete spikes and energy is only consumed when a spike occurs, which reduces energy consumption by orders of magnitude.

The focus of the press release was about efficiency -- to tout merely a hot-word wake feature as "another key efficiency feature" is overselling it, even from a marketing standpoint, so I think Akida is involved in more than that. This is supported by the fact that they gave one "example" of the hot-word detection -- which in actual fact would make much an impact on the car's range, as well as the final sentence "When applied on scale throughout a vehicle, they have the potential to radically reduce the energy needed to run the latest AI technologies." Also remember that in the video below, Rob demonstrated 3 automotive systems (models) running on a single Akida 1000 and a total power consumption in the mW range.


Why is this claim being made? In my mind Daimler engineers have already been experimenting with / have developed systems that have already achieved this, given that there is always a time lag between developing a product and showcasing it.

3. Daimler has already successfully integrated Akida in a hot-word detection system, and have stated that it consumes much less power. This is a validated use by a commercial 3rd party in this function of an Intelligent Virtual / Personal Assistant (IVA / IPA) device market. This the same market of smart home / office automation shared by Amazon's Alexa, Google Assistant, Apple's Siri, and related devices -- mobile phones included.

Let that sink in for a while.