What if I told you that Alan Turing’s work started impacting today’s study of AI before computers were invented?
If you watched The Imitation Game starring Benedict Cumberbatch, the name of Alan Turing should ring a bell: this British scientist is best known for solving the Enigma code which the Germans employed during WWII to transmit coded messages. Not only did this breakthrough shorten the war by at least two years but it also is estimated to have saved more than fourteen million lives.
Who is Alan Turing? How Did He Impact the Computer World?
Here is how it happened. The Enigma code was developed for the Enigma Machine, a German electromechanical rotor ciphering device. The machine was the first one to have a plugboard, which increased the complexity of the cipher, making it nearly impossible to crack.
Pre-war Poland had its own team of engineers and mathematicians that created the Bombe machine. The problem was that the Germans used to create new ciphers daily, so their messages could not be intercepted in the beginning of World War II.
This is where Alan Turing came in. Together with his fellow code-breaker Gordon Welchman, he improved the Polish Bombe, which had been decoding all messages sent by the Enigma machines since late 1940.
Then Turing moved on to the more complex German naval signals. Together with his Hut 8 team at Bletchley Park, he managed to decrypt these as well in 1941, contributing significantly to Allied victory in the Battle of the Atlantic and, subsequently, in the war.
This might already sound astounding, but there is even more to the genius of Alan Turing. Many people believe him to be the founder of computer science – and not without reason.
Turing laid the foundation for theoretical computing and artificial intelligence and came up with the unprecedented concepts of algorithms and computation. He created the Turing machine, the prototype of the modern-day computer. Moreover, he envisioned all of this without actually seeing a computer!
So, what is the Turing machine? In his article On Computable Numbers…(published as early as 1937), Turing considered whether a method or process could be devised that could decide whether a given mathematical assertion was provable.
Turing analyzed the methodical process, focusing on logical instructions, the action of the mind, and a machine that could be embodied in a physical form. He developed the proof that automatic computation cannot solve all mathematical problems. This concept became known as the Turing machine.
This idealized computing device consists of a read and writer head. It operates on an infinite strip of tape divided into squares. Each square contains a different symbol written on it (expressed as a septtuple ‘Q, T, B, ∑, δ, q0, B, F’). The machine might resemble a scanner but it is intended to store memory. It acts as an input/output vehicle, which is fundamental for solving a problem in programming. This became the stepping stone for modern-day problem-solving.
Turing understood the idea of controlling the function of a computing machine by using memory and storing encoded instructions. These are the very same concepts of modern-day low-level computing.
Next, Turing took this idea and imagined the possibility of multiple Turing machines, each corresponding to a different method or algorithm. Each algorithm would be written out as a set of standard instructions, and the actual interpretation would be considered a mechanical process.
So each particular Turing machine embodied the algorithm, and a universal Turing machine could do all possible tasks. Essentially, through this theorizing, Turing created the computer: a single machine that can be adapted to any well-defined task by supplying an algorithm, or a program.
Alan Turing’s AI Test Origins
But Alan Turing did not stop at that. In fact, he went even further by shaping the future of AI for decades to come. In 1950, at the dawn of computing, he was already grappling with the question: “Can machines think?”.
At the time, the term artificial intelligence had not even been coined: John McCarthy would come up with it only in 1956. In an attempt to answer this question, in his paper Computing Machinery and Intelligence (1950), Turing described a procedure that is now known as the Turing Test.
The test was an adaptation of a Victorian-style parlor game called the Imitation game. It involves secluding a man and woman from an interrogator who has to guess which is which by asking questions and studying typewritten replies. The man aims to fool the interrogator, while the woman tries to help him.
In the Turing Test, a computer program replaces the man. He asks: “Will the interrogator decide wrongly as often when the game is played like this as he does when the game is played between a man and a woman?”.
As such, the test studies whether the interrogator can determine which is a computer and which is human. Turing did not explicitly say that the interrogator should be told one of the respondents was a computer, but it seems clear from his example questions. The idea was that if the questioner is not able to tell the difference between a human and a machine, the computer will be considered to be thinking.
Many researchers have taken great interest in passing the Test. In 1990, businessman Hugh Loebner set up the annual Loebner Prize competition with a prize of $100,000 to the creator of a machine that could pass the Turing Test. Turing himself believed that by the year 2000 computers would be able to pass the test with flying colors.
In reality, however, no one has yet managed to create such a machine to pass the Turing test to the full extent. But this ongoing contest has most certainly been one of the driving forces behind AI research. Scientists use it for a lot of tasks, from robots assessing the dynamics of climate change to computers running the world’s finances.
It is hard to say how any of these achievements would have been possible without the continued inspiration from Turing’s fresh and bold ideas.
“At Dasha, Our Voice AI Passes a Limited Turing Test”
That’s the basis of the Turing AI test, a way of inquiring if computers can think as humans do. During interactions with conversational AI, the test has been framed around the question of whether humans can distinguish if they are speaking to an AI or a fellow human being.
At Dasha, we are creating a voice AI that already passes a limited Turing test . In about 20 different scenarios, including additional sales, delivery confirmation, scheduling doctor appointments, NPS surveys etc., 98% of people speaking to Dasha can’t tell it from a human. Dasha combines human-like sound, fast response time and natural language understanding algorithms to provide a superior user experience.
However, we never stop dreaming big: 70% of our team are engineers (among them 21 ACM ICPC participants), and with such a team passing the full Turing test is only a matter of time.