This post was originally published by Vishnu Kumar V H at Medium [AI]
Briefing you on the first ever paper on AI
Turing in 1950 published the first ever article on Artificial Intelligence which he then called ‘Computing Machinery and Intelligence’. This summary/ article will give you an idea of what he wrote in world’s first ever interpretation on Artificial Intelligence in this highly philosophical paper, his views and the predictions he made on AI which stand still even today.
‘Can Machines Think?’, Alan Turing ponders as he tries to convince us that Machines can indeed think but he takes his time to do so. He starts off by suggesting a game, a game called the imitation game and it goes as follows:
Three individuals are involved. A man, a woman and an interrogator. The job of the interrogator is to question them and identify the man and the woman. Well of course, the interrogator can’t see them nor speak to them. He/she can question them using nothing but typewritten letters. Labelled X or Y, the man and woman will have to answer the questions and the interrogator will have to identify them. This was the game. Plain and simple but this is when it gets interesting, Turing aligns it straight towards the dead center of the core of this paper and brings in a twist, ‘what if one of X or Y was replaced by a machine’, he says. ‘Will the interrogator still be able to identify them?’
The paper starts shifting gears, Turing has our attention and wants it all on the core of the game itself. The imitation game, it does not wish to reveal if anyone excels in anything, it does not wish to choose the best of us all. The game is simple. ‘Are you human?’ It’s all the game wants to know. He includes the most human things that a machine would need to mimic, from taking time to add 7-digit numbers to not being interested in poetry (Some of us of course) to knowing chess a tad bit. Things that make us human.
Turing has established what his game is and now tries to address what would be needed on the other end of the game, the AI end. ‘Digital computers’, he says. ‘That’s the solution’. Mostly in this part of the paper, Turing focusses on making claims and defending them, that a machine can indeed have high speed and computation. It can even store infinitely large volumed data, he says. For which he even gives a logical solution to make it a reality. ‘Digital computers will need a store to store the data, the executive unit for processing the stored data and a control will make sure the orders given are executed correctly’. These roughly translate to the RAM, CPU and Control Unit we have in our laptops and computers today, a store, an execution unit and a control.
Again, he makes his predictions, which hold right today. 10⁹ is the number he gives for storage capabilities. And it does hold true, Computers can hold Petabytes worth of storage and dedicated storage units even more. He bashes with utmost confidence that Machines will have meaningful conversations without any contradiction in the end of the Century and despite a couple of decades late, again holds true with the massively popular Query-based AI systems, including Alexa and Siri and the Google Assistant.
Right then, enough of computer hardware I suppose, jumping into the most interesting part of the paper. Turing gets a little defensive on his interpretation of AI and wants to answer questions that he thinks might pop up most. Turing calls this a debate and wants to solidify his understanding on the question, ‘Can Machines think?’.
And now to the debate of ideas.
- The first one he tackles is the Theological Objection meaning God has given souls to human and not computers hence only humans can think. Yes, he did say that. Let’s get over it, it’s the 1950s. But for his time which he was clearly ahead of, what he said was right. Association between souls and thinking was probably true then and he wanted to object it, well, with all of his heart and his soul. Machines can think, Turing concludes.
- The consequences of a thinking machine are too dreadful (Aka., Terminator: Judgement Day): Turing says that this thought comes from the narcissistic nature of humans thinking that they are kind of superior to the rest of creation and we don’t want to lose command and control over things. ‘We don’t have substantial evidence to prove this would happen,’ he says. ‘And this idea mainly spawns from the intellects who value power of thinking more than anything.’
- The Mathematical Objection: Computers have limitations and they have only mathematical results, one or zero and are not fuzzy, meaning there’s no in-between. Meaning, machines have limitations. To this, even humans have limitations, defends Turing.
- The Argument from Consciousness: Turing uses the word Solipsist to describe the Egocentric nature of man. In counter argument to Professor Jefferson who says that only a Machine which can write a symphony or be made miserable by its mistakes or be angry or depressed can be said to have a brain. But Turing says this is no criteria for the test. The criteria Turing says is for Machines to think, not feel emotions. So, despite not being able to fulfil the criteria for Professor Jefferson’s paper, Machines can think, says Turing.
- Arguments from various Disabilities: This comes from various objections that machines are limited to what they do. They are designed for very limited purposes and their only objective is to fulfil them. And they cannot do other things. Turing defends this by asking the reader to compare the computer’s brain to a child’s brain and provide the same amount information for a child’s brain to mature, then asks us to observe if a computer brain can do everything a human brain can do. (The computer will learn is the answer).
- Machines do not think on their own: Meaning, they only do what we ask them to do. Or saying ‘They can’t take us by surprise’. Turing defends this by saying any given expert in a field, if said anything would immediately associate the term hence removing the element of surprise for any given topic. So, there can’t be an element of surprise, which is expected by the expert, says Turing.
- It’s not the same as the Nervous System: Turing promptly shuts this down by saying that it is not part of the requirements of the game. For the game to function, only requirement is that interrogator should be able to identify if there is a human on the other end or a machine and does not require any information about the Nervous system.
- Argument against Informal Behavior: Against the argument that even under similar circumstances, humans behave different, Turing points out that this is a choice the human makes not a necessity to follow. For instance, if late to work, a human might pass red light as well and this is completely a choice and not an absolute necessity to follow. And Machines will do what is only necessary, says Turing.
Turing predicts that machines would learn based on experience, something we know very well about today. He compared the learning of a machine to that of the learning of a child and says machines would be able to learn the same way a child would. Starting from birth, to education and later other experiences (Not to be education based). He even suggests a punishments and reward-based system, what we now call, ‘Reinforcement Learning’. Something even in 2020 is in its early stages used only in games and waiting to be used in wider areas.
It is astonishing that Turing said all of this about 7 decades ago in the year 1950. A man clearly ahead of his time in terms of science and imagination, Turing has undeniably nailed on his predictions on the future of AI. Despite not coming up with the term Artificial Intelligence, we can all agree that this paper was indeed the starting point for AI once and for all. Turing has inspired many more of his peers and contemporaries as well. And this very well lead into the first ever conference on Artificial Intelligence in the year 1956 and the reveal of another genius mathematician who would later be called ‘Father of Artificial Intelligence’.
This post was originally published by Vishnu Kumar V H at Medium [AI]