The past, present, and future of computer intelligence

On the 75th anniversary year of Alan Turing’s seminal paper Computing Machinery and Intelligence, Manchester’s Dr Omar Rivasplata, Senior Lecturer in Machine Learning, discusses the importance of the concepts the mathematician and computer scientist proposed in shaping computer intelligence today, and considers how artificial intelligence (AI) continues to mould the future. 

Creating the imitation game

In 1950, Alan Turing’s influential paper Computing Machinery and Intelligence was published in the journal Mind, in which he described a three-player game involving a man (player A), a woman (B) and an interrogator (C). The man and woman could only communicate with the interrogator through written notes. The interrogator’s goal was to guess which player was the man, and which the woman, while A’s goal was to confuse the interrogator by attempting to pass as a woman.  

Turing went on to consider an adaptation of the game for a thought experiment, replacing player A (the imitator) with a machine that could write answers to questions. The new aim was for the interrogator to determine which answers were coming from the human player and which from the machine. If the right identification could not be reached consistently, Turing argued that the machine could successfully pass as a thinking being.  

Consequently, Computing Machinery and Intelligence, published during Turing’s time at The University of Manchester and widely considered the first paper on AI, proposed a clever way to tackle the question: can machines think? Even when applied to humans, characterising thinking in a way that can be verified is challenging. In the absence of this clear-cut definition, Turing’s game is used as a test to determine whether a machine behaves like a thinking being. We now know this as the Imitation Game or Turing Test – but does it still hold up when used on today’s AI?  

Training artificial intelligence

75 years on from the paper’s publication, there is growing discussion and debate on the influence and power of computer intelligence and artificial intelligence. State-of-the-art AI-driven programs like ChatGPT could easily pass Turing’s test, yet they remain flawed. We can ask a question and get an answer that appears human-like; however, when probed and challenged to act as part of a conversation with debating and questioning, the way two humans would, the program’s answers soon reveal a lack of reasoned thinking. While AI’s response in the single-question scenario comes very close to human behaviour, we see that AI cannot mimic a human when interrogated.  

Early machine learning concepts used human learning as inspiration and considered the question of whether a computer program could mimic the behaviour of an educated adult when excelling at some tasks. Turing suggested that rather than programming a computer to have this base intelligence as a starting point, it would be better to write simple programs that resemble a child’s untrained mind and follow it up with exposure to training and experience. Instead of coding the behaviour, he proposed coding the principles that can develop into improved, more mature behaviour. This concept still underlies machine learning algorithms today. 

Entering a new era of technology

There was a time when computers weren’t a huge presence in the workplace; people would write by hand or on typewriters and use calculators. As with the adoption of computers, I believe something similar will happen with AI in terms of disrupting the way people work. The systems will become part of working culture, and we’ll need to adapt. Computer intelligence is perhaps just the next chapter in human development.  

At Manchester, AI is at the forefront of conversations. With more than 800 researchers dedicated to the field and £75 million in active research grants, we host one of the UK’s largest university communities focused on AI and data science innovation. Researchers are applying AI to urgent challenges in health, climate, energy and infrastructure. It can be used to tackle interesting problems, answer questions or test hypothesis frames. It can also act as a research facilitation tool – a note-taker and efficient search engine – cutting down on cost and excessive human hours. This technology can even offer a virtual stage for clinical trials, simulating conditions where new devices and medicines can be tested safely in controlled virtual environments. From predicting flood risk to improving cancer diagnosis and accelerating drug discovery, there are endless opportunities for AI to develop further and become a helpful tool that doesn’t change what we do but makes what we do better. 

Tackling real world challenges

At the Centre for AI Fundamentals (AI Fun), we are connected by a common interest in the general principles underlying machine learning algorithms and AI. This interest extends across the University, as more and more researchers look at how AI can be used to help tackle scientific questions and real-world challenges.  

From work taking place in the Department of Computer Science to make systems secure and prevent cyber-attacks, to the development of novel health technology at The Pankhurst Institute, and multi-objective optimisation at Alliance Manchester Business School (AMBS) – there’s so much being done here that will shape the future, both at home and further afield. Our cross-sector partnerships are playing a pivotal role in Greater Manchester’s AI ecosystem – the largest outside of London – helping to drive innovation across industry, local government and the NHS.  

What would Turing think about AI today? I think he would be pleased to see how far we’ve come but his inquisitive mind would question what machine behaviour is really telling us. As the home of the first paper on AI, Manchester's research community continues to be at the forefront of progress, 75 years on from Turing’s seminal work.