OpenAI showed GPT-4o on 13 May 2024. The headline is not another benchmark chart. It is conversation: audio that responds fast enough to feel like turn taking, plus text and vision in one model.
Interface shift
Typing to a chatbot was already normal. Speaking to one that interrupts, laughs on cue, and tracks what you show it pushes the computer toward something closer to a call.
That dazzles. It also raises uncanny valley and consent questions when voices get too smooth.
Why this demo matters
Human-computer interaction has been keyboard and glass for decades. A model that makes talk primary changes product roadmaps for assistants, accessibility, and customer support.
Whether the feeling lasts after rate limits and rollout drama is next month's problem. This week, talking to AI started to feel less like dictation and more like dialogue.
Product teams watched the demo and immediately rewrote roadmaps: voice notes, tutoring, hands-free field work. Regulators and parents heard something else: persuasion at conversational speed.
Both readings belong in the same week. Capability and caution arrived together.
Brak komentarzy:
Prześlij komentarz