The Download: Feature Articles
The Names and Science Behind the Turing Award
By Nicolas Lebrun | June 26, 2019
Award Pays Homage to a Persecuted Pioneer of Computer Science
On June 15, 2019, Yann LeCun, Silver Professor of Computer Science at NYU’s Courant Institute of Mathematical Sciences, received the 2018 A.M. Turing Award from the Association for Computing Machinery (ACM). The ACM presented LeCun and two of his colleagues, Yoshua Bengio and Geoffrey Hinton, what is often referred to as the “Nobel Prize of Computing” for conceptual and engineering advancements that led to the deep learning revolution in artificial intelligence (AI).1
The award’s namesake, Alan Mathison Turing (1912-1954), was a British mathematician and theoretical computer scientist, perhaps most widely known as the subject of the 2014 film, The Imitation Game. The film primarily dramatizes Turing’s interest in cryptography (the study of code breaking), beginning in his adolescence and continuing through his time at Bletchley Park during the Second World War. At Bletchly, he was responsible for decoding the Enigma machine, a highly-advanced German encryption device capable of generating over 159 quintillion permutations of a code.
Although these efforts led to information that ostensibly precipitated the end of the war, Turing’s code breaking by no means represents the extent of his contribution to the field of computing. Considered one of the founders of AI, Turing was deeply interested in the potential of this technology. He was one of the first to outline the fundamental limitations of mechanical computation, and even expressed an interest in the possibility of “building a brain.”
This preoccupation led him to propose a method of determining whether a computer is capable of human-level cognition. The Turing test, as it’s commonly known, envisions two participants, one human and one machine, communicating with one another via the exchange of written or electronic text. If a human evaluator reading these communications is unable to determine which of the pair is the machine, then, Turing argues, the machine should be considered intelligent.2
Currently, one of the most common applications of AI resembles an inversion of the Turing test, as machines have become increasingly reliable at verifying the identity of humans online. Such is the purported function of CAPTCHA (Completely Automatic Public Turing Test to tell Computers and Humans Apart), frequently added to websites to prevent bots from creating accounts and other automated behaviors The service’s ulterior function is that it uses humans to help solve text- and image-based problems too difficult for AI’s algorithms, to the financial benefit of the service’s owners.3 Although the mildly-distorted segments of text that users are presented with when trying to access online services may not be as integral to national security as highly-encrypted, war-time communications, decoding them does serve a purpose — namely, improving the state of artificial intelligence.
Shortly after being purchased by Google in 2009, reCAPTCHA (a popular deployment of CAPTCHA technology) was implemented as a means of deciphering distorted text that Google Books’ algorithms were unable to decode. Since then, the service has begun displaying images used to train Google’s driverless cars. Thus, the reason CAPTCHA validation of humanness is becoming increasingly difficult is not that AI is becoming indistinguishable from human intelligence (yet), but that text- and image-reading algorithms are getting more competent, and human intervention is only required for progressively ambiguous cases.
This is in no small part due to contributions made by LeCun and his colleagues. In the 1980s, LeCun began proposing efficient Deep Learning methods that would lay the foundation for the resurgence of interest in this technology. Deep Learning is a process by which deep neural networks (computing systems composed of millions of virtual neurons) process multiple layers of data inputs in order to form complex internal representations of the data with which they can evaluate further inputs.
One of LeCun’s biggest insights was the development of the Convolutional Neural Network (CNN), a network that mimics the organizational structure of the human visual cortex by limiting the neurons to a restricted receptive field. This contribution has enabled breakthroughs in fields such as computer vision, speech recognition, and natural language processing4, and represents the closest modern computing has come to realizing Turing’s aspirations.
In the years following the Second World War, Turing continued pushing computation forward. In 1945, he presented a paper outlining the first detailed study of a stored-program computer, an architecture that would remain standard until well into the 1960s. However, Turing’s full potential would remain unrealized.
During an investigation into a home invasion in 1952, police officers uncovered evidence of Turing’s involvement in a homosexual relationship — illegal at the time in Great Britain. Faced with the charge of “gross indecency,” Turing was presented with an ultimatum: either receive hormonal treatment designed to reduce his libido or serve a prison sentence. Turing opted for the former.
His security clearance was revoked and his work in cryptography ceased. He retained an academic post but suffered mental and physical degradation over the following years, resulting in his death at the age of 41. Officially, his death had been ruled a suicide by cyanide poisoning, although some reports postulate that it may have been accidental.
The Turing award was presented to LeCun, Hinton, and Bengio in San Francisco on June 15, 2019. Perhaps simply by chance, the gala took place in the middle of a month honoring the 50th-anniversary of the Stonewall riots, a watershed moment in LGBTQ rights. It was the 52nd time the award was presented, the first being in 1966, a mere 12 years after Turing’s passing and one year before homosexuality was decriminalized in England. Given these circumstances— this year more than any other—this award serves as an invitation to consider the legacy of Turing in a wider context, and the tragedy of a man so devoted to the future, who became the victim of a society unable to move on from the past.
- “A.M. Turing Award,” Association for Computing Machinery, accessed June 17, 2019.
- Alan Mathison Turing, “Computing Machinery and Intelligence,” Mind 59, 236 (1950): 433-460.
- “Teaching computers to read: Google acquires reCAPTCHA,” Official Google Blog, accessed June 17, 2019.
- Yan LeCun, Yoshua Bengio & Geoffrey Hinton, “Deep Learning,” Nature 521 (2015): 436-444. (PDF)