When Machines Lived in Myths (Pre-1900s)

Long before computers or code, humans dreamed up mechanical minds.

  • In ancient Greece, Talos—a bronze giant—protected Crete.
  • Hero of Alexandria (1st century CE) designed automata powered by steam, water, and air, including programmable theaters, coin-operated machines, and automated temple doors. His inventions inspired centuries of mechanical curiosity.
  • Al-Jazari (1206) built a water-powered robot that could serve drinks.
  • Philosophers like Descartes asked: “Could a machine think like us?” The questions stuck, even if the answers didn’t.

War, Wires, and the Birth of AI (1940s-1950s)

  • Alan Turing: Codebreaker, math genius, and father of AI. In 1950, he introduced the Turing Test—a way to check if a machine could pass as human in conversation.

  • Claude Shannon: Decoded how we send information, laying the foundation for digital communication.

  • Norbert Wiener: Coined “cybernetics” in 1948, focusing on control and communication in animals and machines.

  • Early computers like ENIAC set the stage for programmable machines.

AI’s First Big Moment (1956-1970s)

  • Dartmouth Workshop (1956): The term “artificial intelligence” was coined here by John McCarthy.

  • Key Figures: John McCarthy, Marvin Minsky, Claude Shannon, Allen Newell, and Herbert Simon.

  • Allen Newell and Herbert Simon: Developed the Logic Theorist, often called the first AI program, and the General Problem Solver, pioneering symbolic AI and problem-solving approaches.

  • Early Programs: SHRDLU, Logic Theorist, and ELIZA demonstrated the potential of symbolic reasoning.

The First Cold Shower (1970s)

AI Winter: Hype fizzled as computers were too slow and promises too big. Funding dried up, but research continued in the background.

The Expert System Boom (1980s-1990s)

  • Expert systems like MYCIN (medical diagnosis) and XCON (computer configuration) mimicked expert decisions using “if-then” rules.

  • These systems were influential but ultimately limited by their rigidity and inability to learn.

Machines Start Learning (For Real) (1990s-2010s)

  • Machine Learning: The shift from hand-coded rules to data-driven learning.

  • Algorithms: Decision trees, support vector machines (SVMs), Bayesian models, and ensemble methods like random forests and boosting became prominent.

  • Applications: Email spam filters, fraud detection, and Netflix recommendations made AI useful in daily life.

Deep Learning Brings AI Back to Life (2010s-2020)

Deep neural networks inspired by the brain led to breakthroughs.

  • 2012: AlexNet won the ImageNet competition, marking a turning point for deep learning.

  • 2016: AlphaGo defeated the world champion at Go, shocking the world.

  • Mainstream AI: Siri, Alexa, and Google Assistant became household names.

The Era of Generative AI (2020s-Now)

AI isn’t just solving problems—it’s creating.

  • Core innovation
    Transformers: The architecture behind today’s generative models.
    Multimodality: Expansion from text to images, audio, video, code, and protein design.

  • Major Models:

    • ChatGPT (OpenAI): Conversational agent, enterprise copilots, reasoning engines.
    • BERT (Google): Foundation for NLP, search, and contextual understanding.
    • Claude (Anthropic): Safety-focused generative AI, strong in long-form reasoning.
    • Gemini (Google DeepMind): Multimodal model for text, image, audio, and reasoning.
    • DALL·E (OpenAI): Text-to-image generation with high fidelity and editing features.
    • Stable Diffusion: Open-source text-to-image model. It has broad customization and community development.
    • Midjourney: Artistic image generation. It has strong adoption in creative industries.
    • Sora (OpenAI, 2024): High-fidelity text-to-video generation.
    • LLaMA (Meta AI): A family of large language models. They are widely used for research and development.
    • Mistral (Mistral AI): An open-source model family. It is known for its efficiency and strong performance for its size.
  • Industry Impact

    • Travel & Hospitality: AI concierges, itinerary builders, real-time translation, hyper-personalized experiences.
    • Finance: Risk modeling, fraud detection, algorithmic trading.
    • Healthcare: Diagnostics, patient analytics, drug discovery (AlphaFold 3, ESM3).
    • Media & Entertainment: Scriptwriting, music generation, design, video creation.
    • Enterprise: AI copilots integrated across productivity, customer support, and operations.
    • Regulation & Trust: AI Act (Europe, 2025) and global frameworks shaping responsible adoption.

Perplexity: The Rise of the Conversational Answer Engine

Perplexity AI is a standout in the generative AI era, redefining how people search for and interact with information.

  • Founded: 2022 by Aravind Srinivas, Denis Yarats, Johnny Ho, and Andy Konwinski, all with deep backgrounds in AI and engineering.

  • What It Does: Perplexity is a conversational answer engine that uses large language models (LLMs) to provide real-time, cited answers to user queries. Unlike traditional search engines, it synthesizes information from the web and presents it in natural language, with source citations and suggested follow-up questions.

  • Key Features:

    • Real-time web search and synthesis for up-to-date answers.
    • Citations for transparency and trust.
    • Multimodal capabilities: text, images, and code generation (Pro version).
    • Personalization and contextual awareness, learning user preferences over time.

  • Growth and Impact:

    • Over 10 million active users and more than 780 million queries processed in May 2025, with sustained 20% month-over-month growth.
    • Valued at over $14 billion as of June 2025, with major investors including Jeff Bezos, Nvidia, and Databricks.
    • Launched a publisher revenue-sharing program and is developing the Comet browser, a cognitive operating system for AI-powered browsing.

  • Why It Matters: Perplexity is seen as a challenger to both traditional search engines and other generative AI tools, offering a more transparent, efficient, and user-centric approach to information discovery.

But… Now What? The Challenges Ahead

We face new questions:

  • Can AI be fair? Bias persists in training data, impacting hiring, lending, and justice.

  • Is it stealing art? Creators challenge AI companies over copyright, royalties, and consent.

  • What’s real vs fake? Deepfakes, synthetic voices, and AI-generated video (e.g., Sora) blur truth.

  • Who’s accountable when AI fails? From misdiagnoses to financial errors, liability is unresolved.

  • How do we regulate AI? Europe’s AI Act launches in 2025, with global rules still forming.

Added Challenges:

  • Job Disruption: AI doesn’t replace all jobs—it reshapes them. Vulnerable roles include admin support and customer service. Companies now prioritize AI literacy as a core skill for new hires.

  • Infrastructure Costs & Energy Use: Training and running AI models consumes huge resources. AI data centers could use 12% of U.S. electricity by 2026. Google reports Gemini uses 30x less energy than earlier models.

  • Data Scarcity: High-quality data for training is running out. This limits progress and causes underperformance in non-English and specialized languages.

  • Is There an AI Bubble? Sam Altman and others warn of hype exceeding reality. 95% of corporate AI projects fail, often due to poor integration rather than weak models.

Responsible AI Focus:

  • Transparency: Clear disclosures and explainability.

  • Ethics: Guardrails for data sourcing and deployment.

  • Accountability: Legal clarity for misuse and errors.

  • Human in control: AI as co-pilot, not unchecked authority.

What’s Coming Next:

  • Multimodal models: Text, image, video, and sound in one system.

  • Personal AI agents: Always-on assistants for work, health, and life.

  • Industry-specific AI: Specialized copilots for travel, healthcare, finance, and law.

  • AI governance systems: Real-time monitoring of compliance, fairness, and safety.

Timeline of AI and Machine Learning

Era Highlight Who or What?
Pre-1950 Mechanical myths & philosophy Talos, Hero of Alexandria, Al-Jazari, Descartes
1950s Turing Test, AI coined Alan Turing, Dartmouth Crew
1960s-70s Symbolic AI, rule-based logic SHRDLU, Logic Theorist, Newell, Simon
1980s Expert systems boom MYCIN, XCON
1990s Machine Learning gets traction SVMs, Bayesian models, ensemble methods
2010s Deep learning revolution ImageNet, AlphaGo
2020s Generative AI takes over ChatGPT, Midjourney, BERT, DALL-E, Stable Diffusion, Perplexity
2023 GenAI mainstream adoption GPT-4, Claude, Bard launch; 42% of large businesses adopt AI
2024 Multimodal AI breakthrough GPT-4V, Claude 3, Gemini Ultra; ESM3, AlphaFold 3 for protein sequencing
2025 AI agents & enterprise integration OpenAI o1, Sora video generation, 223 FDA-approved AI medical devices; AI outperforms doctors in clinical diagnosis

Let’s Remember

AI didn’t appear overnight.

It’s the result of 2,000 years of ideas, from myths and mechanical inventions to math, algorithms, setbacks, and breakthroughs.

Every generation has pushed the boundary a little further.

Now, with generative AI, we’ve entered a new chapter, but this is still the beginning.

Learning how we got here is the first step to understanding where AI can take us next.

And we’re still just getting started.