GAFAM's lust for AI patents

Published on 11 February 2021 at 10:42

The GAFAM (Google, Amazon, Facebook, Apple and Microsoft) have created an empire, within the European Market, through patents. Their data collection and processing power are the cause of their success. Though, the expertise and funding of such companies does not completely explain why they succeeded where European companies didn’t. What they came up with could have been replicated, but it didn’t, and that’s because of patents. 


Patents grant a monopoly of 20 years over an invention. They provide exclusivity regarding a process or a product. 


A technology can have so many patents associated with it that it is impossible for competitors to use it. This happens with “patent clusters”. The incentive function of patents, their promotion of innovation is thus discouraged. The constant grant of patents in an area of technology virtually eliminates competitors. 


To give you an idea: the technology of an iPhone is made of 55 patents. There are Apple stores across 25 countries in the world. So, they have at the very least deposited patents in each of these. 


The GAFAM companies are omnipresent in our society: from retail services, to cloud ownership, television and movies, along with artificial intelligence virtual assistants and food delivery. This protects these large corporations from influence from competitors. They accumulate their inventions and patents, covering a wide net of technology and diverse markets. Always one step ahead, they increasingly expand their business.


There should be a way to limit how these large organizations can cumulate patents. 

This isn’t only an antitrust concern, to maintain competition and growth within our market. This is a matter of citizen’s rights. The more these companies are exclusively able to collect data from their users, through patented technologies, the more users become their product.


Users end up within filter bubbles, are subject to targeted advertising. All their research, purchases, calendar events and conversations get carefully used against them. This profits the companies to whom such data is sold on to, by service providers.


There is an urgent need to provide for concrete regulation, especially concerning cloud ownership. The European Cloud Strategy of 2012, through the ‘Cloud for Europe (C4E)’, aims to, I quote, “build trust in European cloud computing ». There is a real objective of informing European citizens and a hope for transparency. Indeed, the ‘European Cloud Partnership’, aims to bring together the industry and the public sector to, I quote: “work on common procurement requirements for cloud computing in an open and fully transparent way ».


Then one gets struck by reality. I will invite you to look at the most recent patent issued by Amazon to the United States Patent and Trademark Office (USPTO) published on the 4th of February 2021. It is headed “managed distribution of data stream contents”. The abstract is not ‘exactly’ intelligible to the common user. It is convenient to these companies that the common user doesn’t know anything about artificial intelligence. 


Your best friends to understand AI are: neural network, backpropagation and overfitting. 

The MIT news journal defines neural networks as: “means of doing machine learning, in which a computer learns to perform some task by analysing training examples. » 


A neural network is meant to solve problems. For this reason, it gets updated to make better predictions through what is called ‘optimization’. The more it gets trained the better it should perform. However, AI gets also used to solve problems which it hasn’t been trained for. That’s where backpropagation comes in. 


The neural network is composed of its architecture and of weights, the latter are attributed to features surrounding the problem at hand. Optimization will adjust the neuron’s weight over time so that the neural network fit to the data obtained through training. For instance, a robot betters its performance trying, analysing features repeatedly and adjusting its behaviour, until it gets to solve a problem it wasn’t even trained for. Algorithms allow to correct such behaviour, make sure all the neurons that contribute to an error fixed. 


At times, in big datasets, backpropagation will make neural network fit to certain data, collected through training, based on coincidental relationships. Used as patterns those create a danger of ‘overfitting’. The AI ends up assimilating information to tasks that don’t necessarily match because it went beyond what was a simple system.


In other words, users train their phones and computers to know what music or movie they are likely to enjoy, from what they have already listened to and watched. Moreover, devices can accumulate so much information as to easily deduce general behaviours. 


In the last months, US authorities have opened fire. They accused Google in October 2020 and Facebook in December 2020, of abusing their dominant position. An antitrust suit was filed against Google, accusing them of endangering competition since they hugely pay off companies to prioritize their search engine for their products. As for Facebook, the US authorities want to force the company to sell off WhatsApp and Instagram. 


If the European Union wants to chip in though, it will have to do so consistently. National authorities regulating competition cannot investigate these corporations while the EU are investigating as well.


Add comment


There are no comments yet.