Amazon GO, retail on AI steroids, Open training environments for AIs - Last Week's Tech News Roundup

Please feel free to subscribe to my newsletter

1. Amazon GO, retail on AI steroids

Last week Amazon announced its secretive project Amazon Go with an exhilarating video, heralding the arrival of future we have been talking about (could be reframed as future we have been fearing of). It is one of the best examples of AI creating a tangible efficiency gains I’ve ever seen. Technology behind Amazon’s out of this realm retail concept is AI powered by deep learning and sensor technologies.

Although Amazon hasn’t provided specifics, readily available technology at the market is, in my opinion, sufficient to replicate the proof of concept video Amazon has provided. The weight behind the strategic momentum in fact is not from the technology itself, rather Amazon’s technology and ability to scale it without collapsing under its own weight.

Two weeks ago in Las Vegas, Amazon announced new capabilities of its AWS platform. Amazon made clear that it is very very very serious about artificial intelligence (including digital assistant) applications. It was so serious that they showed off what it is capable of in a retail store by giving us Amazon Go.

I believe Amazon will position it “as a Service” product to extend its operations to offline stores rather than push its way into physical retail space. Benefits of selling it to other stores include: scaling Amazon Go and AWS, better margins for AWS, an ecosystem to other retail verticals. Also there is a hidden game with very big returns.

I am cutting this news short, because there are lots of ground to cover. Please consider this as a teaser. I will be sharing my detailed analysis (underlying technology, Amazon’s angle, hidden game, competition) tomorrow. Please stay tuned. 

2. Google’s DeepMind and Open AI releases their AI training environments

Last week NIPS (Neural Information Processing Systems) Foundation held 2016 conference in Barcelona. During the conference Google’s DeepMind and Open AI (a non-profit platform backed by Elon Musk, Peter Thiel etc.) released their AI training environments, DeepMind Lab and Universe – respectively.

AI research community has been known for its openness. Google, Facebook, Microsoft, Amazon etc. release tools core to their systems for community to use and fork. Now both open environment platforms provide AI research community and enthusiasts effective ways to train and test AI agents’ performances.

The momentum behind sharing is

• Both DeepMind and Open AI expect communities shaped around their initiatives to accelerate research through interaction and help platforms to evolve
• It is very important to get people on board with own platform, help users familiarize, teach them how to use and embed it deep into their systems, just enough for other services, like cloud infrastructure to sell itself
• Retain talent. Researchers flock to those companies to use research resources – money and most importantly data. As an end product, they write a research report and share their findings with the community. This has been the way, as far as I have observed. Even Apple couldn’t change modus operandi and recently announced that it is going to let its researchers publish despite previously Apple executives were boasting Apple was not the place for people with publishing motivations. Talent nourished with the right tools is ripe for a company to add to its payroll with minimum down time
• Now the scarcity is on the data side. I know current AIs are weak and far from a general AI, but this is a limiting strategic parameter for everyone. However, data is propriety and most of the platforms’ business model is already in place to leverage this competitive advantage. Google holds the data to questions we ask, Facebook keeps data of who we befriend and how we interact, Amazon is basically online possession vault etc. Every company approaches question from a different perspective but the data is the constant competition driver. Knowing that, Uber last week announced Uber AI labs, new division of Uber dedicated to AI research, and acquisition of Geometric AI, which approaches AI with less data intensive methods to overcome its shortcomings
• M&A. Since 2011 140 companies on AI have been acquired, nearly 30% in 2016. Big firms gobble up small prominent newcomers, like Google did on DeepMind. Opening up platforms and underling algorithms will create more vibrant research environment therefore leading to more firms emerge. Those training environments, DeepMind Lab and Universe, present a concise way to compare capabilities of an AI system to others. Comparison also brings learning through interaction. Think of it as a house party. Hosting children will open her/his toys to other kids. Those kids surely will show new ways to play with those toys, therefore increasing imagination and mental capacity of the child. It is the same for AI. Not only people will contribute to those systems by improving with crowdsourced efforts, but also letting AIs play each other, mostly with the host. Another upside is if a new firm emerges with technologies of those firms, acquiring and assimilating would be easier

(I’ll be releasing “AI handbook for executives” in 1Q17, subscribe to receive a copy)

3. Google will fully operate on renewable energy in 2017

Google, on Tuesday, announced its long-time commitment to clean energy has elevated to a point that the company will fully run on renewable energy in 2017. The renewable energy in fact is not directly fed into Google operations, rather it is “plugged into the grid, so that Google’s usage presents no net consumption of fossil fuels.” The company, declared itself as the biggest corporate buyer of renewable power, gets into agreements with renewable energy producers to guarantee it will buy energy, therefore helping producers to find financial resources to fund their projects. Currently there are 20 of those agreements resulting in $3.5 billion infrastructure investment in 5 countries across 3 continents to feed 13 large scale data centers around the world.

Other global technology companies have their own commitments to get rid of fossil fuels. Microsoft which has been carbon neutral since 2012 and powered by 100% renewable energy since 2014. Apple last week participated into a joint venture with the world’s largest wind turbine maker to supply Apple’s production partners in China.

4. Biggest headset manufacturers founded Global Virtual Reality Association (GVRA)

Google, HTC, Facebook, Samsung and Acer have partnered to develop best practices on VR and share them openly. GVRA aims to promote the worldwide growth and development of the VR industry, by fostering dialogue between public and private stakeholders in VR around the world and make education and training material available to the public.

GVRA might standardize certain aspects of hardware and software development. Currently headset developers burn lots of precious R&D money on the same set of problems which won’t bring any significant competitive advantage, on the contrary delays progress. Harmonizing the industry with standards will accelerate growth of the market.

One peculiar thing though, organization currently seems to accept only international headset manufacturers to the association. I am not sure what to make out of it, but it makes me wonder whether this is some kind of collusion by which companies with high vertical integration, starting from hardware, will shape the industry. 

5. Google, Microsoft, Facebook and Twitter plan to fight hate speech together

They take EU Commission’s warnings seriously when it issues an ultimatum. The commission reminded technology giants about their unfulfilled promises on hate speech. The companies pledged to “take action in Europe within 24 hours, following rising concerns triggered by the refugee crisis and terror attacks.” A report compiled by EU Justice Commissioner showed only 40% of the cases were reviewed within 24 hours which increased to 80% within 48 hours. The commissioner warned companies to take immediate action in order to save themselves from a legislative approach.

Upon receiving the news the companies rolled out an initiative to create a shared database of “unique digital fingerprints for images and videos that promote terrorism.” Each company will contribute images and videos that violate their content policies. Collaboration will be limited to only sharing content, companies will act on content independently.