From Turing's famous question to today's generative AI, and what it all means for cybersecurity.
The Story So Far Start Your JourneyArtificial intelligence (AI) is the field of building computer systems that perform tasks we normally associate with human intelligence: understanding language, recognising images, making decisions and solving problems.
AI isn't one technology but a family of approaches that has grown over more than seventy years. Early systems followed hand-written rules. Today's systems learn from huge amounts of data and can write, translate, code and reason across many tasks.
Each field sits inside the one before it, just like the rings in the orbit above.
Any technique that lets machines mimic intelligent behaviour, including rule-based expert systems.
Systems that learn patterns from data instead of being explicitly programmed.
Machine learning with many-layered neural networks, behind modern vision and speech.
Models that create new text, images, audio and code, such as large language models (LLMs).
Breakthroughs, hype, winters, and a revolution.
Alan Turing asks "Can machines think?" and proposes the Imitation Game.
The Dartmouth workshop coins "artificial intelligence" and launches the field.
Frank Rosenblatt builds an early artificial neural network that learns from examples.
One of the first chatbots imitates a therapist using simple pattern matching.
Big promises meet limited computing power. Funding dries up, while rule-based expert systems find niche business use.
Popularised as a practical way to train multi-layer neural networks.
IBM's chess computer defeats world champion Garry Kasparov.
Wins the ImageNet challenge by a wide margin, kicking off the deep learning boom powered by GPUs.
DeepMind beats Go champion Lee Sedol, a feat many thought was decades away.
"Attention Is All You Need" introduces the architecture behind today's language models.
Models show they can write convincing text from a short prompt.
Image generators and ChatGPT reach the public. ChatGPT gains 100 million users within months.
AI assistants become everyday tools for writing, coding and research, and governments begin regulating AI, for example with the EU AI Act.
years of AI research
the year "AI" was named
the Transformer is born
ChatGPT users in about two months
The same technology gives both defenders and attackers new powers.
Four habits that keep AI helpful instead of harmful.
Hidden instructions in web pages, emails or files can hijack an AI assistant. Treat its inputs as untrusted.
Never paste passwords, keys or confidential data into AI tools that aren't approved for it.
AI can state wrong facts confidently. Verify important answers, commands and code before you use them.
Models learn from human data and can repeat its biases. Keep a human responsible for important decisions.
A practical path for security and IT professionals.
Learn Python and basic statistics, and understand how data, features and models fit together.
Train classifiers with scikit-learn on security data such as logs, URLs and malware features.
Explore neural networks with PyTorch or TensorFlow, and learn how Transformers and LLMs work.
Build AI-assisted tools for your SOC or network, and study AI security, for example the OWASP Top 10 for LLM Applications.