Why does AGI matter?

What benefits could AGI bring?

What risks does AGI pose?

Terrorism

Authoritarianism

Loss of Control

Summary

You can make a difference

How our courses work

Getting started

What's already possible

The pace of progress

From tools to agents

What is AGI?

Vending-Bench: measuring progress towards AGI

Building AGI and beyond

The huge bet on AGI

The world is in a crazy place. Five years ago, AI models struggled to form coherent sentences. Today, they are reaching beyond human-level performance across a wide range of tasks. A new model launch advances the state of the art nearly every week, and if anything the pace of progress is speeding up.

This has huge implications for the world we live in: what will our jobs look like? Will we even have jobs? How will AI accelerate science, for good and for harm? How will societies be structured and governed?

If you’ve been mulling over these questions, thinking “this is clearly the problem of our time but I don’t know what to do about it”, then this course is the place to start.

In 2 hours we’ll cover where AI is headed, what this means for the world and the paths for you to work on ensuring it goes well.

But first let’s ground this conversation in what AI models can already do today.

### **Images**

The first text-to-image AI system was built in 2015, producing tiny 32x32 pixel images. Today's AI-generated images are indistinguishable from photographs, and can produce a range of different artistic styles:

<Embed url="/images/content/foai/midjourney.webp"/>
Source: Midjourney. Explore more from [Midjourney](https://www.midjourney.com/explore?tab=top_month), [ChatGPT Images 2.0](https://openai.com/index/introducing-chatgpt-images-2-0/), and [Google Nano Banana](https://gemini.google/overview/image-generation/).

### **Videos**

Four years ago, AI-generated videos didn't exist. Two years ago, they were terrible. Now, the frontier is becoming hard to distinguish from real videos:
<Embed url="https://www.youtube.com/embed/I7tMTopo6xk?playlist=I7tMTopo6xk&loop=1"/>
Model: Kling 3.0 (please forgive the corporate voiceover)

As with images, video models can also produce a range of different styles, from claymation to anime, and even minecraft:
<Embed url="https://www.youtube.com/embed/3S7naW\_KRyw?end=20"/>
Model: Seedance 2.0

Explore more from [Kling](https://kling.ai/), [Seedance](https://seed.bytedance.com/en/seedance2_0) and [Veo (Google)](https://deepmind.google/models/veo/).

### **Graphic design**
<Embed url="https://www.youtube.com/embed/t\_LBECIQQqs"/>

### **Building web applications**

AI can generate working code. Just tell it what kind of website or app you want, and it will build it. 

Try it yourself and create your own web app below! Type in your simple idea, or click on one of the examples. It'll take ~30 seconds to build.

<Embed url="https://course-demos.k8s.bluedot.org/generate-react-component" className="h-450"/>


More important than today's capabilities is how quickly things are changing. It’s easy to forget quite how bad AI models were even 5 years ago.

Here’s how much image models have improved since 2022 (prompt: “an astronaut riding a horse on mars, photorealistic”):

<Embed url="/images/content/foai/astronaut-on-mars-over-time.webp"/>

Image generation is far from the only domain where we’ve seen this progress. Over a similar timespan, we’ve seen giant leaps in AI models’ ability to answer PhD-level science questions, do competition-level maths and autonomous software engineering tasks.

We’re beginning to run out of tests where AI is not already at human level.

<Embed url="/images/content/foai/ai-index-benchmarks.webp"/>
Source: [Stanford HAI AI Index Report 2026](https://hai.stanford.edu/ai-index/2026-ai-index-report)
### 
### Moore’s Law for AI coding capabilities
Moore's Law is the observation that the number of transistors on computer chips doubles approximately every two years. This exponential progress is what has powered the computing revolution over the past 50 years.

A similar observation for AI capabilities indicates that the length of coding task an AI can do has been doubling every 7 months since 2019 (speeding up to doubling every 4 months since 2023).

If we extrapolate this 4-month doubling trend forwards, we would expect to get AI models that can do:
- Coding tasks that would take a human a week by Sep 2026 (assuming a 40 hour working week)
    - As of April 2026 we’re already seeing evidence that [AI models can complete software engineering tasks that take a skilled human engineer multiple weeks](https://epoch.ai/blog/mirrorcode-preliminary-results#discussion-and-conclusion).
- Coding tasks that would take a human a month by May 2027
- Coding tasks that would take a human a year by Aug 2028

<Embed url="/images/content/foai/metr-task-horizon.webp"/>
Source: [METR](https://metr.org/time-horizons/)

This is indicative of a wider trend. As AI systems have become more capable of longer and more complex tasks, we have shifted from AIs being ‘tools’ where you ask a question and get a simple response, to ‘agents’, where you give the AI a goal and it goes off and does it on your behalf.

<Callout title="It’s a little more complex than that!">
The graph technically shows the ’50% task-completion time horizon’; the task duration (measured by human expert completion time) at which an AI agent is predicted to succeed 50% of the time.

For example, Claude Opus 4.6 succeeds roughly 50% of the time at coding tasks that take human software engineers 12 hours. [You can read more about how this is measured here](https://metr.org/time-horizons/).

Succeeding 50% of the time is obviously less than you would want from your human software engineer! METR also have an 80% task-completion time horizon, for which the frontier is 1h30 tasks as of Feb 2026. Crucially, this measure shows the same exponential trend as the 50% task-completion time horizon.
</Callout>


Up until mid-2025 AI systems were mostly tools, responding only when prompted. Now we’re seeing the emergence of “agents”, for example:
- **Tool:** A calculator that solves problems when you input them
- **Agent:** A research assistant who investigates topics & writes full reports

Agents can:
- Search the web
- Make decisions
- Act independently
- Complete tasks without step-by-step instructions

As their capabilities improve, they are being given more license to roam the internet, control computers, and interact with the physical world via robot bodies.

### **Doing tasks online for you**
<Embed url="https://www.youtube.com/embed/rBJnWMD0Pho"/>

### **Controlling your computer remotely**
<Embed url="https://www.youtube.com/embed/NAauIR6JFps"/>

### **In the real world**
Robots powered by AI can now manipulate objects and perform tasks in the real world. Today, they're slower and make more mistakes than humans, but they're improving fast. 
<Embed url="https://www.youtube.com/embed/CAdTjePDBfc"/>


In your own words, how would you describe the difference between a ‘tool’ and an ‘agent’? Give one example of each


Future of AI

The pace of progress