Chess not checkers
The patents of the tech built to outsmart us
Do you remember when you only needed to decipher a couple of blurry words to complete a CAPTCHA? It seemed so unnecessarily complicated at the time. But then in 2012, Google upped the ante and began including snippets from Google Street View, requiring web users to transcribe door numbers and signage. Despite the CAPTCHA being conceived as a way to to sort the biological from the robotic, the inventor of the tantrum-inducing check, Louis von Ahn, quickly realised that the collective hundreds of thousands of hours spent everyday typing out jumbled letters could be used to transcribe and digitalise old newspapers and books, which means there’s every chance you (and thousands of others) helped put a 19th century romance novella on an online bookshelf without even knowing it. These days, Google continues to harness your free labour; every time you pick out the bridges or traffic lights for a CAPTCHA, you are helping Google train the exact sort of technology that they are trying to beat: artificial intelligence.
AI has become somewhat of a buzzword given its reach into film, news headlines and even the aforementioned CAPTCHA. When you consider how the public have been informed on AI with movies like iRobot and Ex Machina and news headlines like Elon Musk wanting to insert computer hardware into our minds with his OpenAI company, you can hardly blame someone for having a quixotic expectations. But the AI being wielded today is still relatively primitive and far from the super intelligence powering HAL 9000.
Research into AI only really began in the 1950s and, while AI has since been used to beat humans at a number of board and reality TV games, it has also experienced periods of stagnation, known as AI winters. First in 1974–1980, in which time unrealistic expectations and limited computing hardware tempered the ambition of computer scientists, and a second winter followed in 1987–1993 as a result of curbed investment in specialised computer hardware.
But then the 90s brought a renaissance of AI with IBM developing Deep Blue: an AI machine trained to beat humans at chess. The world’s best chess player at the time, Garry Kasparov, famously lost a six game match against IBM’s supercomputer in 1997 — a symbolic moment in history which drew attention to the fast closing gap between machine and human intelligence. The public recognition of AI’s potential fostered a spur of investment that has continued over the last 22 years.
AI is nowadays a highly commercialised tool with conglomerates like Google, Amazon, Baidu and others currently employing machine learning to leverage their big data into a multiplicity of profitable products and services. In fact, Google is a top filer of AI-related patents with two thirds of its patent applications being granted — an impressive rate, given how notoriously tough examiners are on software patents. And although companies like Google use AI for marketing and consumer insight applications, over half of last year’s AI-related patents were actually for computer vision — the field of AI that trains computers to interpret the visual world in order to make vehicles autonomous. To be fair, the prospect of being driven around by a digital chauffeur easily justifies the disproportionate number of patent filings.
Evidently, the current frontier of AI technology isn’t quite on the same level as what Hollywood would have us believe, which leaves us wondering when we will see patents for the Holy Grail of AI: generalised artificial intelligence — the stage of AI in which computers will be able to think and perform tasks at human level. This intersection of intelligence, known as singularity, raises many philosophical questions. But there is one question in particular already being asked in the patent community, and it’s not uncontroversial: when AI itself inevitably begins to create IP, who is considered the inventor on the patent?
Under the UK Patents Act 1977 and the European Patent convention, only ‘natural persons’ are recognised as inventors. Although, this was challenged last month by a team of AI researchers, which filed patents for two inventions in the UK, US and EU patent offices. What they revealed to the patent board post filing was that the inventions were actually the product of an AI system they built called DABUS, and that the inventions (a food container capable of changing shape and a flashlight system capable of drawing attention in emergencies) were technologies that the researchers themselves had no background in. The three patent offices are in the process of reviewing the applications.
Should AI be considered inventors? Provide us with your best argument and follow our social media.
We have a lot more to say about jet suits and patents. For information on this and any other patent information study, please visit patently.com and email us at ask@patently.com.
