What is AI, really?
For those of you who have just started to engage with the world of AI or you are exploring it for the first time, we wanted to start with a piece that sets out the basics. Because here at Diverse AI ensuring that everyone has the information that they need to be an active participant in the conversation is what we are all about. If you’re already a seasoned AI user or work in AI, bear with us for a couple of posts, until we get into the juicy stuff!
If you have ever asked a virtual assistant for the weather, been served an unnervingly accurate recommendation on a streaming platform, or received an email flagged as spam before you even opened it, you have already used artificial intelligence. You just probably didn’t call it that.
One of the biggest barriers to having honest public conversations about AI is the gap between what the word conjures and what the technology actually does. "AI" has become a catch-all term that covers everything from basic sorting algorithms to systems that can write poetry, analyse medical scans, and hold a conversation. Treating those things as equivalent is a bit like saying an abacus and a scientific calculator are the same because both involve numbers.
So let us start with what AI actually means (because there isn’t a universally accepted definition). At its most basic, artificial intelligence refers to computer systems designed to perform tasks that would normally require human intelligence: recognising patterns, making decisions, understanding language, or learning from experience. The key word there is "learning." Traditional software follows fixed rules written by a programmer. AI systems, particularly modern ones, learn from data (machine learning). Feed a system enough examples of spam emails and it begins to recognise what makes an email spammy, without a human specifying every rule in advance.
The OECD defines AI as:
An AI system is a machine-based system that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. Different AI systems vary in their levels of autonomy and adaptiveness after deployment.
There are a variety of ways in which AI is being used right now which are really beneficial for people and planet:
Transforming Healthcare: AI has been transforming healthcare by analysing vast amounts of biological data to discover new drugs, predict diseases earlier, and personalise treatments. DeepMind's AlphaFold, for example, solved the decades-old protein-folding problem, unlocking new possibilities for treating diseases like Alzheimer's, cancer, and antibiotic-resistant infections. In recognition of this, the 2024 Nobel Prize in Chemistry was awarded to DeepMind's Demis Hassabis and John Jumper for their breakthrough contributions.
Helping to tackle Climate Change: AI is playing a growing role in protecting communities from climate-related disasters. An international team of researchers has introduced Aurora, a groundbreaking AI model trained on over a million hours of data, designed to deliver faster, more accurate, and more affordable forecasts for air quality, ocean waves, and extreme weather events. Potential applications include forecasting flood risks, wildfire spread, seasonal weather trends, agricultural yields, and renewable energy output.
Supporting Access to Education: AI-powered tools can be used to democratise education by providing personalised learning experiences to students regardless of their location or economic background. Intelligent tutoring systems can adapt to individual learning styles and pace, while AI translation tools break down language barriers, giving learners access to materials in their native tongue. Khan Academy's AI tutor Khanmigo is one of the most concrete examples of AI expanding access to education at scale. It is available in more than 34 different languages, student numbers now exceed 1.4 million students and it is now being rolled out to classrooms across India, Brazil and the Philippines as well as the U.S.
However, whilst there is fantastic progress being made to use AI in ways which are really positive, there are also some entrenched challenges.
The way that computers are trained, for instance, can open up important questions about fairness and accountability.
Fairness in AI refers to ensuring that an AI system's outputs don't systematically disadvantage or discriminate against particular groups of people. When AI models are trained on biased datasets the model learns and then perpetuates those biases. In 2025 an academic study audited several large language models using over 332,000 real-world job postings, prompting each model to recommend whether an equally qualified male or female candidate should receive an interview callback. Most models favoured men, especially for higher-wage roles, with lower callback rates for women in male-dominated occupations [Chaturvedi & Chaturvedi, 2025].
Accountability refers to the question of who is responsible when an AI system causes harm or produces a biased outcome. This is surprisingly difficult to pin down. Is it the engineers who built the model? The company that deployed it? The organisation that collected the biased data? Accountability frameworks push for clear lines of responsibility, transparency about how systems work, and mechanisms for people to challenge or appeal AI-driven decisions that affect them.
If a system learns from data, and that data reflects historical inequalities, the system will learn those inequalities too. The term ‘GIGO’ (Garbage in, Garbage out) is applicable here! The AI does not know it is being unfair because it is not conscious and has no ability to think. It is simply doing what it was trained to do. Researchers including Joy Buolamwini, founder of the Algorithmic Justice League, have documented this pattern extensively, most notably in facial recognition systems that performed significantly less accurately on black women than on white men, because the training data was not representative.
AI has been around longer than most people realise. Early versions of machine learning date back to the 1950s, and by the 1990s AI was already being used in fraud detection and logistics. What changed dramatically in the 2010s was the combination of vastly more data, cheaper computing power, and new techniques that allowed systems to improve at tasks with far less human supervision. That shift produced the AI tools most of us now encounter daily: voice recognition, personalised feeds, translation tools, and more recently, conversational systems that can respond to almost any question you put to them.
Understanding that AI is a spectrum rather than a single thing is not just semantics. It shapes how we regulate it, who benefits from it, and who gets left out. A spam filter failing is an inconvenience. An AI system making decisions about who gets a job interview, a loan, or a medical diagnosis is something else entirely.
That is why who builds these systems, whose experiences they are trained on, and whose voices are included in the conversation about how they are governed is not a peripheral concern.
Next up we will look at generative AI specifically: what it is, how it works, and why it represents a new chapter in this story.
Recommended Reading & FREE Courses:
This article on AI ethics from Coursera also has links for some free online courses
Top AI Ethics and Policy Issues of 2025 and what to expect in 2026
Elements of AI and Ethics of AI are both free courses from the University of Helsinki
Artificial Intelligence: A Guide for Thinking Humans by Melanie Mitchell. This is a great beginner-friendly book that explains AI concepts in plain, conversational language. It's ideal for those unfamiliar with AI and doesn't assume any technical knowledge.
