The history of artificial intelligence

1950

The Turing Test

"Can machines think?" Alan Turing, a British mathematician, posed that question in his paper "Computing Machinery and Intelligence." At the dawn of the computer age, researchers were already questioning how far computer intelligence would advance in the future.

Turing proposed a test that would determine when machines reached a level of intelligence that made their behavior indistinguishable from humans. The Turing Test helped structure AI research even before the term "artificial intelligence" was first used in 1955.

1995

Internet-Powered Chatbot

From the 1950s to the 1990s, artificial intelligence made many advances. In 1965, a computer program known as ELIZA could imitate human conversation. In 1988, an early chatbot called Jabberwacky used natural dialogue to learn from and interact with humans. And in 1995, a new chatbot first used information from the internet to answer prompts.

A.L.I.C.E., or Artificial Linguistic Internet Computer Entity, processed natural language data from the internet to converse with users, marking a major leap forward in conversational AI.

1997

Deep Blue Defeats Chess Champ

Deep Blue, a chess-playing computer developed by IBM, defeated a world chess champion in 1997. For the first time, computer intelligence defeated human intelligence in a strategic game. How did Deep Blue succeed? By running through millions of possible moves with lightning speed.

The chess milestone revealed new applications for AI. By the turn of the 21st century, researchers developed new autonomous systems, machine learning tools, and natural language processing systems to apply AI in countless fields.

2011

Watson and Siri

The history of AI saw two major advances in 2011. Watson, an IBM computer powered by natural language understanding, defeated two Jeopardy! champions. The quick processing speed, paired with the predictive capabilities of machine learning algorithms, gave Watson the edge over the champs.

Apple also launched Siri in 2011, a milestone in voice recognition and AI-powered assistance. The virtual assistant could process voice commands and perform tasks for users.

2017

Chatbot Negotiation

Chatbots reached a new milestone in 2017 at the Facebook Artificial Intelligence Research lab. Researchers asked two chatbots to negotiate. The negotiation began in English, the language the chatbots had been trained to use, but the bots quickly developed an original shorthand language to speed up their communication.

Although researchers ended the experiment, the language leap showed how AI can evolve without human intervention, sometimes in unpredictable ways.

2020s

ChatGPT-3 Launches

In the 2020s, OpenAI launched new language models that could generate text and more. In 2020, ChatGPT-3 used its large language model (LLM) featuring 175 billion parameters to converse with users, translate multiple languages, and create original code. Three years later, in 2023, Open AI launched GPT-4, an LLM that could process multiple inputs, including text and images, to generate responses.

Companies such as Google and Anthropic launched their own sophisticated AI models, ballooning the generative AI market to an estimated $25.6 billion in 2024. These advances were made possible in part by powerful new chips developed by companies like NVIDIA, whose high-performance processors deliver the computer capacity needed to train and scale large AI models.

Jeff Le

• Managing Principal, 100 Mile Strategies
• Visiting Fellow, National Security Institute, George Mason University

Q: What do you think was the most pivotal moment in the history of AI?

Le: Very tough to boil it down to any one moment since AI has been built on the advancements of many generations, but Deep Blue defeating Garry Kasparov in chess captured the imagination of what was possible. It is no surprise that moment was the precursor to multiple deep learning milestones and achievements.

Q: What are the biggest challenges facing the field right now?

Le: The field continues to battle the challenges of balancing safety and security with innovation and progress. Implications for these breakthroughs range widely, and [their] impacts on society merit deeper inspection. There also continues to be a limitation on AI data scientists, access to top-tier GPUs, and the energy required to harness the processing power. These challenges will only grow with time.

Q: What skills should someone entering the field of AI focus on first?

Le: A curiosity for learning new things and the willingness to explore the latest tools are critical. There are introductory AI programming languages and practices one can learn, which are important, but the passion to be hands-on is valuable.