Tuesday, November 14, 2017

AAAS Statement on Scientific Freedom and Responsibility

Scientific freedom and scientific responsibility are essential to the advancement of human knowledge for the benefit of all. Scientific freedom is the freedom to engage in scientific inquiry, pursue and apply knowledge, and communicate openly. This freedom is inextricably linked to and must be exercised in accordance with scientific responsibility. Scientific responsibility is the duty to conduct and apply science with integrity, in the interest of humanity, in a spirit of stewardship for the environment, and with respect for human rights.

For more information: https://www.aaas.org/page/aaas-statement-scientific-freedom-responsibility

Camille Flammarion: "Urbi et Orbi”, in L'atmosphère: météorologie populaire, 1888

Where the sky and the Earth touch

Thursday, November 9, 2017

Panasonic buying deep learning startup

Arimo was born Adatao in 2013 and is being acquired by Panasonic. It calls its product Behavioral AI and targets it to machine learning for Industry 4.0.

It started with two tools. pAnalytics is a Spark environment providing an API where developers can work with the data and expose it to the end users with charts and graphs. pInsights is the end user layer, which takes natural language queries. This tool learns from the end user's interactions and can suggest possible queries.

This approach is used to learn from the past behavior of equipment to identify complex anomalies that are hard to predict with traditional statistical modeling. The same deep learning algorithms can also be used to predict retail shopper's behavior to offer them incentives and optimize store inventories. A related solution area is financial services, where the technology can find signals and anomalies in large-scale transactional data to detect fraud, model risk, and predict investor or consumer behavior.

Panasonic first aims to apply the technology to data on business refrigerators for supermarkets and convenience stores. It envisions a service reducing energy consumption for a store chain overall by setting optimal operating patterns for individual stores, based on past data on refrigerators' internal temperature and energy use. Panasonic can then expand the application to industrial air conditioners.

In the future, Panasonic plans services to manage the physical health of the elderly based on data from appliances and a range of sensors. Since Panasonic has few data analysis experts, Arimo will be a training ground for its employees.


Friday, November 3, 2017

3d face recognition

This morning, #45 announced a massive tax relief for the American people. Also as of this morning, the new iPhone X is available for purchase in Apple stores.

If you are investing your massive tax relief in an iPhone X, do not just look at the gorgeous OLED screen, but also at the 3d face recognition sensor, because you have been reading about the underlying physics on this blog.

It has been over a dozen years since Neil J. Gunther of Performance Dynamics, annoyed by a Harvard professor's claim of having disproved Bohr's complementarity principle, proposed to follow the idea of VLSI design rules to formulate practical design rules for quantum communications and quantum imaging devices.

We performed interference experiments in Neil's kitchen using a green laser and a paper clip to form an image. Sergio Magistri noticed that doing physics is good, but creating an artifact that we could sell would be better. He hooked us up with Edoardo Charbon, who had invented a CMOS SPAD array.

After lengthy discussions, Edoardo—who in the meantime had become a professor at EPFL—was willing to reduce our ideas to practice. We received a 500,000 franc grant from the Swiss National Science Foundation to buy the lab equipment and a matching grant from the European Union to hire Dmitri Boiko as a postdoc.

To form the image, we used the metal plate creating the nozzles in an ink jet cartridge to obtain an array of pinholes.

We performed experiments supporting the concept of a g2-camera, summarized on this blog. The statistical post-analysis was so challenging that Neil had to implement it in the fast processor of an oscilloscope. We wrote two papers with the early details:

The blog posts hot body, excited particles, and the north sky and chaotic light sources are the basis for telling apart the sources for the photons reaching the SPAD array.

It is amazing that today the computations can be done on a small, inexpensive smartphone. However, it took 13 years and hundreds if not thousands of people to get to today's device, a simpler version of which, by the way, is also used in Bosch measures.