Blog

More machine learning – ScaledML

27 – 28 March 2019

The ScaledML conference is growing up; from a Saturday at Stanford to a two day event at the Computer History Museum with sponsors. http://scaledml.org/2019/

Two big new themes emerged

  • Concern for power efficiency (Simon Knowles, Graphcore, talked about Megawatts; Pete Warden, Tensorflow talked about milliwatts and energy harvesting
  • Development platforms – Adam D’Angelo, Quora, was particularly clear on how Quora operate development to efficiently support a small number of good developers

David Paterson gave the first talk on Domain Specific architectures for Neural Networks – an updated version of this talk https://cacm.acm.org/magazines/2018/9/230571-a-domain-specific-architecture-for-deep-neural-networks/fulltext

The roofline performance model is a useful way to visualize comparative performance. For future performance improvements functionally specific architectures are the way forward; this requires both hardware updates (what Google is doing with the TPUs) and improved compiler front and back ends.

Fig 3 from the Domain Specific Architectures paper linked above.

Intel recognizes this trend – Wei Li described the work his team is doing to incorporate domain specific support into Xeon processors. This blog post has the gist of what he presented.

Most of the talks are here on YouTube

More on Micromobility

GoBike
requires rescue

Recently in Los Altos. Reminder that rental bikes require service and support – this one has a flat front tyre, damaged front handlebars, and a pair of spectacles poked into the cracked plastic. Emailing Ford GoBike to suggest they picked it up resulted in rather stilted customer service responses.

There’s lots of work still to do to get to the micromobility transformation in transport for suburbs. That transformation is a reasonable reaction to a poor infrastructure overburdened with vehicles much larger than they need to be for the movement of people and things they perform.

Reference – Horace Dediu https://micromobility.io/blog/2019/2/23/the-micromobility-definition

Mobility’s second great inflection point

Headline on a McKinsey article https://www.mckinsey.com/industries/automotive-and-assembly/our-insights/mobilitys-second-great-inflection-point

However, it entirely misses cycling, ebikes, and scooters. Perhaps the authors all live in snowy cold cities.

” .. the will to fix our cities is getting stronger. Urban populations, historically at the vanguard for change, are already more receptive to solutions such as ridesharing and carsharing, as well as EVs. ”

When the majority of trips are short, and as delivery robots gain adoption, the number of people willing and able to use a bicycle with electric power assist increases substantially. This has significant implications for the transportation business, and especially for automotive manufacturers.

Delivery of things

Another delivery robot company – this one operating at scale in Europe and making campus deliveries in the Valley. Starship raised $25m in June 2018. Runs on paths and sidewalks. No hints in the demo videos about how interacts with opposing direction bicycles and scooters.  Ideally something like this would be able to enter buildings and control lifts for campus deliveries. 

Commercial package delivery launch announcement (Oct 31 18) – in Milton Keynes now, and the SF Bay area before the end of the year. https://www.starship.xyz/press_releases/worlds-first-robot-package-delivery-launched-publicly-today/

Venturebeat review 

https://venturebeat.com/2018/10/31/starship-technologies-launches-commercial-package-delivery-service-using-autonomous-robots/