Last June, the U.S. Patent and Trademark Office published U.S. Patent Application No. 20170164007, titled Mixed Boolean-Token ANS Coefficient Coding. Originally filed by Google with the USPTO in December 2016, the patent application claims a method for decoding an encoded bitstream using a computing device, the bitstream including frames having blocks of pixels, which involves the use of an entropy decoder state machine including a Boolean asymmetric numeral system (ANS) decoder and a symbol ANS decoder. The claimed invention is intended to improve the transmission of digital video streams with the use of the entropy decoder to reduce the amount of data involved in the transmission of digital streams.
Listed as the assignee on the ‘007 patent application is Mountain View, CA-based Internet giant Google and the only inventor named is an Alexander Jay Converse of Oakland, CA. However, news reports indicate that a Polish computer scientist named Jarek Duda has publicly claimed that he is the inventor of the ANS compression technique and that Google is attempting to patent his technique despite his own desire to maintain that technique as open source. An article published online by The Inquirer indicates Google has responded to this claim by claiming that Duda’s invention was more theoretical in nature and that the company is trying to patent a specific application of Duda’s concept.
Duda’s lamentations regarding Google’s patent application covering the ANS technique can be read in a June 2017 post authored by Duda and found on the forum of data compression website Encode:
“Nice ‘thank you’ from a multibillion ‘don’t be evil’ corporation to a poor academic whose work they use for free and who has helped them with it for the last three years (e.g. through https://groups.google.com/a/webmproj…um/codec-devel ) – there was a moment they gave me hope for a formal collaboration with my University so I could build a team, but then silence … probably due to this patent application.”
Pardon our French, but it’s très intérresante to hear an inventor claim that Google is in essence stealing a technology after having worked with that inventor on the development of that technology. The link provided by Duda brings a person to a Google Groups page for codec developers and the thread about the entropy coding technique that was authored by Duda can be found by searching that website. The thread was started in January 2014 by Duda and includes 63 posts from eight authors. Other post authors on the thread include Google employees such as Paul Wilkins, Aki Kuusela and, interestingly enough, Alex Converse, the named inventor on the ‘007 patent application. Converse is, in fact, one of the last people other than Duda to comment in the thread, which he did back in October 2016, two months before the ‘007 patent application was filed. Between November 2016 and this May, Duda posted 13 times in the thread on various topics, including to inform Google of a patent application filed in the U.S. and UK to cover an ANS compression technique where he asked Google if they would fight that patent application. On June 11th, 2017, Duda posted a link to the ‘007 patent application in the thread, noting that “Now I understand why there was no feedback” from Google. On June 16th, Duda posted his analysis of the ‘007 patent application claims where he said that he couldn’t find anything new in the claims aside from his own invention.
“This patent application and general situation is completely ridiculous … Therefore, I would like to propose not wasting more time with it – please just remove specifying ANS in this patent, and let us forget about this situation – I can further help you with making the best of ANS, and in case of a formal collaboration I am still open for (which e.g. would allow me to build a team or e.g. just through working remotely and summer breaks for Google), in future bring you many new ideas indeed worth patenting.”
The Encode forum post authored by Duda has been updated several times to note actions that Duda has taken with the U.S. Patent and Trademark Office and other patent offices around the world to prevent the issuance of a patent on any of Google’s patents making use of the ANS technology developed by Duda. This includes the filing of a third-party preissuance submission with the USPTO in August 2017 in which he notified the agency that he is the author of the ANS family of entropy encoders and that it was his will that the technology should be in the public domain. “Google did neither respect my will to prevent legal restrictions for using ANS, nor did consult with me or inform me about this patent application, whose claimed innovations can be found, among others, in my well documented work for and with them,” Duda wrote.
Obviously an inventor is going to want credit for his or her work, and if Duda wants to open source his own technological advance, that is by all rights his prerogative. However, although Google’s actions in this saga appear to be fairly dubious, Duda has viewpoints of his own which deserve to be critiqued under the cold light of day. In an e-mail exchange with Duda, he provided a link to a piece of text which he authored and has been published on a website titled End Software Patents. In that piece, Duda makes the following mental blunder:
“Theoretically, the patent system should not allow [software] patents. They ought to be considered ‘obvious.’ In practice, though, these patents can be issued, because the patent office’s criterion for ‘unobvious’ has historically been astoundingly weak.”
The idea that all software is obvious is a theoretical argument that doesn’t just border on the scattological, it wades right into the sewer. Consider artificial intelligence. If AI, which requires the use of software algorithms, is supposed to augment human intelligence and provide us with answers to questions we can’t figure out without the use of AI, how is that at all obvious? What about IBM’s Watson cognitive computing platform? As of January 2011, Watson could perform 80 trillion operations per second. This year’s Wimbledon tennis tournament will use a poster created through the use of Watson’s visual image recognition application programming interface (API). Yes, humans have designed posters for as long as posters have been used for events, but the idea that a human designer could, in a time- and cost-effective manner, go through the historic archive of photos held by the All-England Lawn Tennis Club to create a composite from 8,400 images is real cause for Duda’s opinion on software patentability to lose any and all credibility.
However, Mr. Duda, we have fantastic news for you. Thanks to U.S. Supreme Court schmatta which can be found in the pages of court opinions from cases such as Alice Corp. v. CLS Bank International, challenges to software patents are strong in the United States. When the highest court in the land incorporates such backward-minded patterns of thought which allows them to say that “At its most basic, a computer is just a calculator capable of performing mental steps faster than a human could,” the U.S. patent system must be a relative paradise to Duda and other anti-patent Luddites who believe that software inventions cannot and should not be patentable at all.
One has to wonder whether Google will in the future enjoy the patent system they have so carefully cultivated over the past decade, or whether Google will come to regret their shortsighted strategy to weaken the patent system for all— including itself.
Image Source: Deposit Photos.