# Have you seen what AI can do?

> Plain-text companion to the interactive film at https://exploresuperintelligence.online/have-you-seen/ (17 chapters; the web version is drawn live in code with WebGL particles and music generated in your browser). Published by Explore Superintelligence, the public site of Superintelligent Silicon. Credit: GIGG × Claude.

## 00. Have you seen (Opening)

Have you seen what AI can do? Not what it might do someday, but what people are doing with it right now, in research labs and at kitchen tables. Everything here is drawn live in code and reacts to you, so drag it, touch it, and see what happens.

An opening built from a living network of points. Drag to turn it, and tap it to send colour spreading through the network.

## 01. Proteins (Science)

For fifty years, predicting how a protein folds was one of the hardest problems in biology. AlphaFold predicted the shape of over 200 million proteins, nearly every one known to science, and released them all for free. The people behind it shared the 2024 Nobel Prize in Chemistry.

Proteins are chains that fold into precise 3D shapes, and the shape decides what they do. AlphaFold, from Google DeepMind with EMBL-EBI, predicts those shapes from the sequence alone.

Sources: [AlphaFold Protein Structure Database](https://alphafold.ebi.ac.uk) · [Nobel Prize in Chemistry 2024](https://www.nobelprize.org/prizes/chemistry/2024/summary/)

## 02. The fine print (Everyday)

Most of us have signed a lease, an insurance form or a medical bill without really knowing what it said. Now you can paste it in and ask what it means, in plain words, line by line, including the parts worth questioning. It won't replace a lawyer or a doctor, but you'll walk into that conversation knowing what to ask.

Paste or photograph a document and ask an AI assistant to explain it in plain language, summarize what you are agreeing to, and flag anything unusual. It's a starting point for a better conversation with a professional, not a substitute for one.

## 03. Storms (Science)

In 2025, an AI weather model helped the National Hurricane Center forecast Hurricane Melissa's rapid intensification and its landfall in Jamaica, days ahead. Forecasters now watch AI storm tracks alongside the physics models they have trusted for decades.

Google DeepMind's cyclone model runs as an ensemble of possible tracks and intensities up to 15 days out, and the U.S. National Hurricane Center used it during the 2025 season alongside its established physics-based models.

Sources: [Weather Lab, Google DeepMind](https://deepmind.google/discover/blog/weather-lab-cyclone-predictions-with-ai/) · [WeatherNext Cyclones and Hurricane Melissa](https://www.preventionweb.net/news/weathernext-ai-model-achieves-breakthrough-forecasting-cyclones)

## 04. Fix it yourself (Everyday)

When the faucet won't stop dripping or a warning light comes on in the car, you can show AI a photo and describe what's going on. It can help you work out what's wrong, which part you need, and whether it's a ten-minute fix or a job for a pro.

Take a photo or a short video, describe the sound or the symptom, and ask what it could be, what to check first, and which part to buy. For anything involving gas, mains electricity or structural work, it should point you to a licensed pro, and so should you.

## 05. Scrolls (Science)

In 79 AD, Vesuvius buried a Roman library. Its scrolls were burned to charcoal, too fragile to ever unroll. Students using AI on CT scans read words from inside a scroll that no one had opened in two thousand years. The first word to appear was purple. The passage turned out to be about pleasure, and music, and food.

The Herculaneum papyri were carbonised by the eruption of Vesuvius. The Vesuvius Challenge combined high-resolution CT scans, virtual unwrapping and machine learning trained to spot ink, and in 2023 students read whole passages for the first time.

Sources: [Vesuvius Challenge](https://scrollprize.org)

## 06. Speak any language (Everyday)

Traveling, or talking with a neighbor who speaks another language? Hold up your phone and have the conversation, each of you in your own words. It translates out loud as you go, and it can tell you the polite way to say something, not just the literal one.

The voice modes in today's AI assistants can listen and translate a conversation as it happens, in dozens of languages. For anything legal or medical, a qualified interpreter is still the right call.

## 07. Speech (Science)

Ann lost her ability to speak to a stroke when she was thirty. Eighteen years later, electrodes read her brain's intent to speak and an AI turned it into words, nearly eighty a minute, in a voice rebuilt from her wedding video.

Researchers at UC San Francisco and UC Berkeley placed a grid of 253 electrodes on the surface of Ann's brain. AI models learned to turn her attempts to speak into text, a synthesized voice that sounds like her own, and a talking avatar.

Sources: [UCSF: how AI gave a paralyzed woman her voice back](https://www.ucsf.edu/news/2023/08/425986/how-artificial-intelligence-gave-paralyzed-woman-her-voice-back)

## 08. Learn anything (Everyday)

Stuck on your kid's math homework at ten at night? Ask it to explain the problem a different way, and then another, until it clicks. It's a patient tutor for anything you've wanted to understand, from fractions to a new language to how your own body works.

Ask for an explanation at your level, ask for an example, ask it to quiz you, or have a practice conversation in the language you're learning. It never gets tired of the same question.

## 09. Sight (Everyday)

Millions of people who are blind now point a phone at the world and hear it described back to them. The expiry date on the milk, the number on the bus, which door is theirs, read aloud the moment they need it.

Apps that pair a phone camera with an AI model, like Be My Eyes, describe scenes, read labels and signs, and answer follow-up questions out loud.

Sources: [Be My Eyes](https://www.bemyeyes.com)

## 10. Where the money goes (Everyday)

Export a month of bank transactions, hand them over, and ask where your money actually went. It'll sort the spending into categories, find the subscriptions you forgot about, and help you build a budget you can stick to.

Download a CSV from your bank, take out account numbers first, and ask an assistant to categorize it, total each category and flag recurring charges. The month shown here is an illustration.

## 11. Medicine (Science)

Pulmonary fibrosis slowly scars the lungs, and the treatments we have today can only slow it down. AI helped find both a new target in the body and a new molecule to hit it, called rentosertib. In an early trial, lung capacity on the highest dose went up on average over twelve weeks, while the placebo group's went down.

Insilico Medicine used generative AI to identify the target (TNIK) and design the molecule. In a 71-patient Phase 2a trial published in Nature Medicine in 2025, the 60 mg once-daily group's lung capacity (FVC) rose by 98.4 mL on average over 12 weeks, while the placebo group's fell by 20.3 mL. Larger trials are still needed.

Sources: [Insilico Medicine, Nature Medicine publication](https://insilico.com/news/tnrecuxsc1-insilico-announces-nature-medicine-publi) · [PubMed record](https://pubmed.ncbi.nlm.nih.gov/40461817/)

## 12. Before the appointment (Everyday)

Before a doctor's appointment, you can describe what's been going on and ask which questions you should bring. You'll walk in organized and understand the words they use, while the diagnosis and the plan still come from your doctor.

Write down your symptoms, when they started and what you've tried, then ask what to bring up and what any terms in your results mean. It helps you prepare; it doesn't diagnose.

## 13. Materials (Science)

AI proposed 2.2 million new crystal structures, nearly four hundred thousand of them stable enough to make. That is roughly eight hundred years of discovery at the old pace, handed to scientists as a list to start testing.

Google DeepMind's GNoME predicted 2.2 million new crystals, about 381,000 of them stable, which the authors described as nearly 800 years of knowledge at the previous rate. Labs have since begun making some of them.

Sources: [GNoME, Google DeepMind](https://deepmind.google/discover/blog/millions-of-new-materials-discovered-with-deep-learning/)

## 14. What's for dinner (Everyday)

Open the fridge, take a picture of whatever's left, and ask what you can make tonight. It'll suggest recipes with what you already have and adjust them for a picky eater, an allergy or a tight budget.

Photograph your fridge or pantry and ask for dinner ideas, a shopping list for the week, or a version of a recipe that fits a budget or a dietary need. Double-check anything that matters for an allergy.

## 15. Builders (Everyday)

All over the world, people who never wrote a line of code are building the tools they always wished existed. A property inspector on an island in the Pacific built the software for his own work this way, one conversation at a time.

Describe what you need in plain words, try what comes back, say what's wrong, and repeat. That loop is how the tools behind this page, and the page itself, were made.

## 16. Start tonight (Closing)

None of this came from waiting. It came from people who got curious, sat down, and learned how the thing actually works. You can start tonight. Ask it something hard, build something small, and see what happens. Have you seen what AI can do?

Every chapter here started with someone asking a question. Pick one thing you'd like to understand or make, and ask.

## How it works

Nothing on the page is a video. Every shape is a cloud of points generated from math, every transition is those points flying to their next shape under simple physics, and the music is composed and played live by your browser. Gold marks the human part of each story, cyan the model's. The science chapters cite their sources; the everyday chapters describe things anyone can try with an AI assistant today, and anything important is still worth checking with a person who knows.
