TikTok Tests Reveal Flaws in OpenAI's New Voice AI
TL;DR Summary
OpenAI's GPT-Live-1 voice model is tested by a TikTok creator who shows it struggle with simple prompts (counting letters in 'seventeen' and the Rs in 'strawberry'), despite OpenAI's claim of a smarter, more natural voice, with observers noting interruptions and awkward sign-offs that highlight early limitations of the new model.
- OpenAI Just Can’t Beat This TikToker Gizmodo
- OpenAI launches GPT-Live voice models that listen and speak simultaneously Reuters
- OpenAI bets voice will become AI's primary interface with new models Axios
- OpenAI's GPT-Live can keep a real conversation Mashable
- OpenAI's upcoming AI device gets its voice Sources | Alex Heath
Reading Insights
Total Reads
0
Unique Readers
2
Time Saved
16 min
vs 17 min read
Condensed
98%
3,233 → 50 words
Want the full story? Read the original article
Read on Gizmodo