| First came HAL. Then came Her. Now there is GPT-4o.
Hollywood has a history of humanlike AI talking to people. But their real-world counterparts never seemed to live up to the hype, sounding too robotic or taking too long to respond.
OpenAI may have finally cracked the code on voice interactions with AI, though. The company unveiled an upgrade for ChatGPT that includes an impressively humanlike voice feature made possible by its new flagship model, GPT-4o.
During a live presentation, OpenAI’s latest product helped with breathing exercises, walked a user through a math equation, and served as a translator. OpenAI says GPT-4o’s average response to audio inputs is 320 milliseconds, which is on par with a typical human-to-human conversation.
The internet was impressed, to say the least. Sharing your screen with ChatGPT while talking through problems could be a game changer for coding and data-focused work, writes Business Insider’s Alex Bitter.
One developer I texted after the event raved about the potential and was excited to test it out.
Typing out class or API-specific names into questions for ChatGPT while coding was tedious, they said. The demo made it seem as if they could talk to ChatGPT like a coworker.
“They’ve essentially made text input almost be the barbaric form of communicating with ChatGPT (or any AI) now,” the developer said. |