The history of artificial intelligence and how it has developed from the 1950s to the day to day form of AI we know today
In 2026, artificial intelligence (AI) surrounds us, it has developed so quickly in the last 5 years that we have seen it become part of our daily lives almost without noticing, whether it be asking Chat GPT to help with (or do) your homework, or using face id to unlock your phone, even listening to your recommended songs on spotify, ai has integrated itself into our lives so smoothly that very few people really stop to ask.. “Where does ai actually come from?” Ai in our society has always been seen as inevitable. Whenever anyone is asked to imagine a future world, robots and human- replacing machines are often ,at some point, mentioned, so, as that future world slowly becomes a reality, it’s important we know the history of the development and creation of AI.
The idea of computer ‘intelligence’ began in 1950, when British mathematician Alan Turing came up with the concept of a machine that can think, proposing a test called ‘The Imitation Game’ which established a benchmark for computer ‘intelligence’. This was determined by whether a computer could reproduce human- like conversation that is indistinguishable from an actual human. The test consists of a human interrogator asking questions to Player A (a human) and to Player B (a computer) from a separate room, with Player B’s objective being to mimic the responses of a human, preventing the interrogator from distinguishing consistently and correctly the human from the computer. Once the interrogator is tricked by the machine enough times, exceeding pure chance, the computer passes the test and has achieved ‘intelligence’. Turing famously predicted that by 2000, computers would be able to play the game so well that an average interrogator would not have more than a 70% chance of making the right identification after 5 minutes of questioning. This test made text based communication the benchmark for machine capability and laid the groundwork for the early chatbots of the 1960s and ultimately the LLMs (Large Language Models) we use today (such as Chat GPT, Claude etc..)
Throughout the 1950s and 1960s AI development exploded, and computers were beginning to do only what outstanding humans could do. Early programs were able to solve complex algebra equations, proved advanced geometric theorems and learned to play checkers at a competitive level. Due to this early boom, governments began throwing millions of dollars at AI development, and unrealistic expectations were set. These expectations were never met and governments were left disappointed. In 1973, a renowned British physicist, Sir James Lighthill, determined that AI had failed to meet its high expectations. This later led to the government pulling almost all its financial support to AI development. The same happened in the USA. This abandonment of AI between 1974 and 1980 is known as the ‘First AI Winter’ and the term ‘artificial intelligence’ became associated with ‘failed hype’. Through the 1980s scientists shifted their focus from building human- like minded AI to fixing highly specific, practical problems, leading to the adoption of ‘Expert Sytems’ (autonomous, decision making AI programs) in 1986. These were a major success, becoming used by large corporations and government agencies. The AI industry was back and booming in the US and Japan, not wanting to fall behind, invested 850 million into AI development, bringing back the international curiosity for AI.
The acceleration of AI development continued and in 1986 Geoffrey Hinton discovered and popularized ‘back propagation’ (AI finds an error and tracks back to find where the error was made to minimise future errors). This initiated the development of computer autonomy and using this, in 1997, Hochreiter and Schimdhuber designed LTSM (long term short term memory) which introduced computer memory. LTSM allowed chatbots to develop and proved that AI neural networks could grasp the complex nature of human language. This laid the initial groundwork for the text prediction we have in our phones today.
By the 2010s, AI was beginning to reach human level and super human performance across multiple areas. In 2012, AlexNet and ResNet (neural networks) allowed AI to recognise objects, and in an accuracy test, its error rates dropped below 3.6%, outperforming the average human in classifying images. AI officially surpassed human level of sight. In 2016, Google Deepmind’s ‘AlphaGo’ defeated Lee Sedol (The world champion) at ‘Go’, a game considered to be immune to computer calculations as it is said to require ‘gut feeling’ and ‘intuition’. AI surpassed human intuition. In 2011, Apple launched Siri, then Alexa in 2014 and Google Assistant was later released in 2016. All these prove a shift towards spoken dialogue in AI development and by 2017, AI had the ability to grasp human slang and idiom through Transformer models, which are able to process all information at the same time (used in models such as Chat GPT and Claude). This consistent superhuman performance brings the future we imagined where human- replacing robots live amongst us ever closer, but for now, AI is working to our advantage. Through its image classification ability, AI is being used to scan through thousands of highly complex medical scans, identifying anomalies in minutes. Through its speech to text ability, AI is also used to convert spoken audio to text in real time across over 100 languages.
Finally, we reach the generative era, where we see AI become universally accessible. After the huge scaling up of GPT-3 (a Transformer model released in 2020) which was fed hundreds of billions of words and information from all over the internet, AI could write essays, translate texts, and generate images. This scaling up later led to the creation of Chat-GPT in 2022. By 2024 both Midjourney and Dall-E (the world’s most popular AI programs) had made text to image generation of photorealistic art accessible to everyone either for free or through subscription. All that was left to be created and made accessible was video generation. In late 2024, Open AI released Sora to the public. This was a huge success and the demand for AI tools grew massively, changing the focus of the top AI scientists from research to business, building products for millions of people, and selling them in subscription form. A large part of AI’s integration into our society has been through its newfound accessibility, which has been commercialised with millions of users paying monthly fees for advanced AI assistance.
Overall, AI has developed through decades of breakthroughs, setbacks and changing expectations. From Turing’s early ideas to today’s generative tools, it has moved from a scientific theory to something embedded in everyday life. Its history shows that progress is rarely predictable, which makes understanding AI’s past important as we decide how it should shape our future.
