Episode notes
An AI scored 135 on a real Mensa IQ test—genius level. But here’s the plot twist: the smartest AIs can’t handle vision tasks, and their “seeing” counterparts perform worse than most humans. On top of that, a new AI model from MIT might have just changed drug discovery forever.
We’ll talk about:
- Why OpenAI’s o3 is ranked “genius,” but GPT-4o Vision flopped with an IQ of 63
- The truth about multimodal models and why they’re not as smart as you think
- Boltz-2: the new AI that’s 1000x faster than physics in drug R&D
- 5 open-source tools to upgrade your AI stack this week
Keywords: OpenAI o3, GPT-4o Vision, Boltz-2, AI IQ test, Anthropic, AI agents, free GPTs, AI Fire newsletter
Links:
- Newsletter: