You can just talk to the computer

Large language models are a technology we unlocked 100 years too early. They’re the steam engine of the mind—transforming raw language into real-time code execution and opening vast new territories of knowledge work.

Conversational applications like ChatGPT are great, highly engaging experiences, even if you think they're just streaming highly plausible text at you at high speed.

There is however, a faster, higher bandwidth way to communicate with the model.

Voice.

Voice assistants powered by language models are a magical new experience, made possible by advancements in language models since the release of ChatGPT.

Speech is inherently more expressive and lower friction than text. When you speak, you bypass the mechanical act of typing and unlock a range of conversational paths. This high-bandwidth interaction allows for deeper, more intuitive exchanges with AI, where every word carries nuance and intent.

When you type with your hands, every single button press requires some effort. Comparatively, speaking allows you to very quickly express your intent, a more direct, high speed, richer way to communicate with AI.

Just try asking your voice assistant to try to solve complex research tasks, plan nuanced objectives for you, teach you a difficult concept. Tools like ChatGPT have already begun to bridge that gap, letting you orchestrate complex tasks with nothing more than the right combination of words.