Is AGI here? OpenAI announces o3

Is o3 AGI? Wait, what's o3?

OpenAI just announced a new reasoning model called o3. It just scored a whopping 87% on the ARC prize benchmark!

For context, here's the progression:

• GPT-2 (2019): 0%
• GPT-3 (2020): 0%
• GPT-4 (2023): 2%
• GPT-4o (2024): 5%
• o1-preview (2024): 21%
• o1 high (2024): 32%
• o1 Pro (2024): ~50%
• o3 tuned low (2024): 76%
• o3 tuned high (2024): 87%

Talk about a quantum leap! 🚀

Wait, what's ARC prize? It's a machine learning competition designed to test a model's ability to generalize and spot patterns that were not in the training data.

Here's an example:

Now, I'm pretty sure you find it easy to solve this puzzle. But AI has historically struggled with it.

But before we pop the AGI champagne 🍾, here's the catch: The computational resources required are MASSIVE. We're talking 172x more compute power between o3 versions alone. Your gaming rig won't be running this anytime soon!

Here's my hot take: While o3's achievement is remarkable, I believe we actually crossed the AGI threshold with GPT-4. o3 is just pushing those boundaries further into uncharted territory.

🤔 What's your take on this? Are we witnessing the dawn of true artificial general intelligence, or is this another stepping stone in the journey?

Let me know in the comments below!