56 AI Facts - Machine Learning, ChatGPT & Future Tech
AI History and Milestones
- The term 'artificial intelligence' was first coined at a conference at Dartmouth College in 1956.
- Deep Blue became the first computer to defeat a world chess champion when it beat Garry Kasparov in 1997.
- The first AI chatbot, ELIZA, was created in 1966 and could simulate a psychotherapist conversation.
- AlphaGo defeated world Go champion Lee Sedol in 2016, a decade earlier than experts predicted was possible.
- ChatGPT reached 100 million users in just two months after launch, making it the fastest-growing consumer application in history.
- The Turing Test, proposed in 1950, remains a foundational concept for evaluating machine intelligence.
- The first neural network, the Perceptron, was invented in 1958 at Cornell University.
- GPT-3, released in 2020, was trained on 570 gigabytes of text data and had 175 billion parameters.
- See also: Technology Facts - AI, Computing & Scientific Breakthroughs
Large Language Models
- Large language models don't actually understand language; they predict the most likely next word based on patterns in training data.
- Training GPT-4 is estimated to have cost over 100 million dollars in computing resources alone.
- Claude, created by Anthropic, was trained using a technique called Constitutional AI to make it more helpful and harmless.
- The largest AI models now contain trillions of parameters, each one a number adjusted during training.
- Transformers, the architecture behind modern AI chatbots, were invented by Google researchers in 2017.
- AI models can hallucinate, confidently stating false information that sounds entirely plausible.
- The context window of advanced models can hold over 100,000 tokens, equivalent to a small novel.
- Fine-tuning a large model on specific data can cost anywhere from a few hundred to millions of dollars.
- See also: Money & Economics Facts - Wealth, Inflation & Global Economics
AI Capabilities and Limitations
- AI image generators can create photorealistic images in seconds, but often struggle with hands and text.
- Current AI cannot truly reason; it excels at pattern matching but lacks genuine understanding.
- AI models trained on internet text can inadvertently learn and reproduce societal biases.
- Generative AI can write code, but studies show it introduces bugs about 25% of the time without human review.
- AI cannot form new memories during a conversation; each interaction starts fresh without prior context.
- Machine learning models can identify cancer in medical scans with accuracy matching or exceeding human doctors.
- AI systems are vulnerable to adversarial attacks, where small changes to input can cause wildly incorrect outputs.
- Current AI lacks common sense reasoning that even young children possess naturally.
AI in Daily Life
- Over 4 billion devices now have voice assistants powered by AI, from smartphones to smart speakers.
- Netflix estimates that its AI recommendation system saves the company one billion dollars annually by reducing subscriber churn.
- AI-powered spam filters block over 100 million spam emails every minute worldwide.
- Self-driving cars use AI to process data from dozens of sensors simultaneously to navigate roads.
- AI algorithms determine what appears in social media feeds for billions of users every day.
- Predictive text on smartphones uses machine learning trained on billions of text messages.
- Banks use AI to detect fraud, analysing thousands of transactions per second for suspicious patterns.
- AI-powered language translation now covers over 100 languages and processes billions of words daily.
AI and Jobs
- The World Economic Forum estimates AI will create 97 million new jobs by 2025 while displacing 85 million existing ones.
- Radiologists, paralegals, and customer service agents are among the jobs most likely to be augmented by AI.
- AI prompt engineering has emerged as a new profession with salaries exceeding $300,000 in some markets.
- Goldman Sachs estimates that AI could automate 300 million full-time jobs globally in the coming decade.
- Tasks requiring creativity, emotional intelligence, and complex problem-solving remain difficult for AI to perform.
- The global AI market is projected to exceed 1.5 trillion dollars by 2030.
AI Energy and Environment
- Training a single large AI model can emit as much carbon as five cars over their entire lifetimes.
- Data centres powering AI consume about 1-2% of global electricity, a figure that continues to grow.
- Microsoft and Google have both reported significant increases in carbon emissions due to AI infrastructure expansion.
- AI is also being used to optimise energy grids, reducing overall power consumption by up to 40% in some applications.
- The water used to cool AI data centres in some regions exceeds the water consumption of entire cities.
- Researchers are developing more efficient AI architectures to reduce the environmental footprint of machine learning.
AI Comparisons to Humans
- The human brain uses about 20 watts of power; training GPT-4 used the equivalent of millions of watts over months.
- A human child learns language from roughly 50 million words by age ten; large language models train on trillions of words.
- AI can process millions of documents in seconds, a task that would take humans years to complete.
- Despite their power, AI models struggle with tasks humans find trivial, like understanding sarcasm or irony.
- The human brain has approximately 86 billion neurons; large neural networks have billions of artificial parameters.
- Humans can learn from a single example; AI typically requires thousands or millions of examples to learn a pattern.
AI Companies and Investment
- OpenAI's valuation exceeded 80 billion dollars in 2024, making it one of the most valuable startups in history.
- Global investment in AI startups exceeded 50 billion dollars annually by 2024.
- Nvidia's market capitalisation grew by over 2 trillion dollars between 2023 and 2024 largely due to AI chip demand.
- The AI chip market is expected to exceed 300 billion dollars by 2030.
- Anthropic, the company behind Claude, has raised over 7 billion dollars in funding as of 2024.
- China invests more in AI research than any country except the United States.