This month: crunch time for quantum computers, an electric car with a quick charging battery (at a price), bias in artificial intelligence, Elon Musk enters the broadband business (via LEO satellites), online grocery orders delivered in 1 hour and more.

Computing: Is it crunch time for quantum computers?

What: Google, Microsoft and Intel are all in the final stages of producing 49 or 50 qubit quantum computers that should achieve ‘quantum supremacy’ – the ability to perform a calculation that a classical computer could not complete in a realistic time frame. However, a mathematician at the Hebrew University of Jerusalem (which has produced 8 Nobel Prize winners and a Fields Medallist in the last decade alone) says that he has modelled the noise processes in quantum computers and that they will correlate in a way that means quantum supremacy will never be achieved, irrespective of how well the qubits are engineered. He believes that the quantum supremacy experiments that we can expect this year could provide confirmation of his theories [1].

Implications: If the professor is proved correct (and it is a very big IF), it could bring the multi-billion dollar investments in quantum computers to a crashing halt. However, given the different forms of quantum computers and the many different types of applications possible, even if he is correct, other approaches may still be valid, so it is unlikely to be the end of the line for quantum computers.

[1] Qanta Magazine

Big Data: Time series database company attracts investors

What: A startup called InfluxData has raised $35m in a Series C round led by the venture arm of software company SAP. InfluxData has created a software platform for recording timeseries data and the tools to analyse that data in real-time. Understanding time series has a myriad of applications (e.g. analysing IoT sensor data, looking for anomalies in computer security information, visualising trends in business performance, etc). The InfluxData software was released as an open source project on Github in 2014 and now has 120,000 users. While the open source version of the software is free (to attract developers), InfluxData’s business model is to offer a closed source paid version with greater scalability and security to enterprises. So far, they have 400 paying customers including the likes of SAP, Cisco, Paypal, Tesla and Siemens [1].

Implications: The importance of analysing streaming data has not been adequately recognised in the hype around Big Data. The investment in InfluxData is an indication that this is changing and that this type of analysis will play a big role in the future of most organisations.

[1] TechCrunch

Transport: New electric car with quick charging battery

What: Californian company Fisker Inc announced a new luxury electric car, the Fisker Emotive, with a new solid-state battery that they say can achieve a 200km range after just 9 minutes of charging. The battery is based on graphene and has been patented by Fisker, who say it has 2.5 times the energy density of lithium ion batteries (it is not clear if this by weight or volume). The total battery capacity of the car is a range of 650km. The car is priced at US$129,900 [1].

Implications: Given that charging time is the last serious drawback of battery cars, the ability to offer a quick charge of this nature is important. However, the claims of charging time may need to be treated with some scepticism until they are proven – Fisker’s claim went from a charge time of 1 minute when they announced the battery patent in November [2], to 9 minutes when they announced the car in January [1]. Given the price of the car, it is also highly likely that the battery is also much more expensive than lithium ion.

[1] Autocar

[2] HybridCar

Broadband: Another LEO satellite constellation gets approval

What: Elon Musk’s SpaceX has received FCC approval to launch 2 experimental satellites as a precursor to creating a low earth orbit (LEO) satellite constellation for broadband communications [1].

Implications: SpaceX joins a growing list of LEO satellite constellation broadband proposals, the most advanced of which is Greg Wyler’s OneWeb system (which will be in service in 2020). The SpaceX proposal is light on relevant details e.g. how many satellites in the constellation, does it include inter-satellite communications, what frequencies will it use, whether the FCC approval is for the 2 experimental satellites or the whole system. However, LEO constellations offer 100% global coverage from pole to pole, and performance equivalent to today’s terrestrial broadband technologies (DSL, cable, fixed wireless but not FTTP). Unlike GEO satellites (such as used by NBN) they don’t incur a latency penalty and will deliver a transformed experience to broadband users out of the range of terrestrial connections.

[1] TechCrunch

Artificial Intelligence: Image recognition works really well – if you are a white male

What: A MIT researcher has tested how well face recognition software from IBM, Microsoft and Chinese company Face++ was able to identify the gender of a person from a photograph. She found that the programs all recognised white males with less than 1% error whereas the error rate for black females was 35% [1].

Implications: The problem here is not the algorithms but the data used to train the algorithms. Training databases are predominantly white and male, leading to biases in the algorithms. The lesson is that effective use of machine learning requires judicious choice of databases to train them.

[1] New York Times

Retail: The race to 1-hour online grocery delivery

What: Today, according to Nielsen, almost a quarter of Americans buy groceries online and they expect this number to triple over the next decade. However, customers want shorter delivery times with an expectation that this will get down to 1 hour, especially for fresh produce. For retailers, the fulfillment centres, however, need to be where land is cheap, so are somewhat distant from the customers. Israeli startup, CommonSense Robotics, is developing a solution based on ‘micro-fulfilment centres’ which is a small building in an urban location with the necessary inventory to supply that area. CommonSense Robotics provide robotics to automate the pick and packing and have just raised a Series A round of $20m to expand their solution which also uses “AI” (presumably to help the grocer decide on inventory levels in the micro-fulfillment centre) [1].

Implications: If three quarters of the population is going to buy their groceries online, retailers will need to focus on their delivery supply chain to be competitive. With Amazon and Walmart dominating this space (in the US), companies such as CommonSense Robotics offer opportunity to other players.

[1] ZDnet

Agtech: Are the cattle running scared?

What: Actually, it appears to the US Cattlemen’s (sic) Association that is running scared. Last week they took the interesting step of petitioning the Department of Agriculture to prohibit artificial beef producers from labelling their products as ‘beef’ or ‘meat’. They want the Department to only allow the produce of slaughtered animals to be labelled with these terms.

Implications: Clearly the US Cattlemen’s Association believe that synthetic proteins and factory produced meats are close to arriving in the market. Numerous companies, including Impossible Foods, Beyond Meat, and Memphis Meats are working on these genetically produced meat substitutes [2]. To date, they have not been price competitive with farmed meat, but the dynamics of the industry are changing fast.

[1] US Cattlemen’s Association

[2] CBInsights

Healthcare: Inferred measurements using AI

What: For some time, wearables have existed that provide basic heart rate measurements. These devices have focussed on health and fitness, not medical information, because of the regulatory hurdles and the limited nature of the measurements – it is just a heart rate, not even a heart wave. Furthermore, the accuracy of the output from the photodiode-based heart rate measurements can be quite variable. However, a US startup, Cardiogram, has shown that by applying machine learning algorithms to the heart rate measurements of the Apple Watch, they can infer a diagnosis for a number of conditions, including atrial fibrillation (with 97% accuracy), sleep apnoea (with 90% accuracy) and hypertension (with 92% accuracy). Their latest result, a study done on 14,000 patients, has also shown they can diagnose diabetes with 85% accuracy [1].

Implications: Studies such as these show that there is potential to use simple wearable devices as diagnostic and monitoring tools in medicine. However, the results need to be treated with caution: the diabetes measurement is not a glucose measurement but an inference as to whether the wearer has diabetes or not. It is not even clear from the results if it can be used to do early detection of diabetes (which would indeed be useful) or will just determine what the wearer knows already (i.e. that they have diabetes).
[1] TechCrunch

Professor Hugh Bradlow is President of the Australian Academy of Technology and Engineering. He is recognised as a global leader in telecommunications technology, including being named by Global Telecom’s Business in the top 100 most powerful executives in the global telecoms Industry two years in a row, and by Smart Company as one of the 12 most influential people in Australian ICT.


Subscribe

to get daily updates on what's happening in the world of Australian Higher Education