Listen: The State of AI

The playing field in artificial intelligence is rapidly evening out. For a long time, OpenAI seemed untouchable. But the latest State of AI report shows that Google, Anthropic, and Meta are catching up. In fact, Meta’s open model, Llama Three, is now performing on par with the big proprietary systems. However, as open models proliferate, we must look closely at what open actually means, since licensing and data transparency vary widely.

Measuring this progress is becoming its own challenge. Current benchmarks are flawed, often suffering from leaked training data that inflates performance scores. At the same time, the industry is shifting its focus. Instead of just building larger models, researchers are finding ways to shrink and distill them so they can run efficiently on everyday devices like smartphones.

Beyond software, AI is stepping into the physical world. Developers are building highly capable AI agents and using advanced simulations to train robots in common-sense reasoning. Perhaps most mind-bending of all, researchers are using foundation models to analyze brain activity, even reconstructing images a person is looking at just by reading their brainwaves.

But this rapid acceleration comes with heavy costs. AI data centers are putting an unprecedented strain on our power grids, and automation is threatening to reshape the job market. Ultimately, AI is a powerful tool, not a crystal ball. Shaping its future requires all of us to stay curious, ask hard questions, and champion responsible development.