Recent experiments with ChatGPT's Advanced Voice Mode and Llama 3.1-based chatbots have yielded intriguing results. Cris Giardina tested the ability of ChatGPT Advanced Voice Mode to distinguish between a human and a machine by asking five questions. A similar test was conducted with a Llama 3.1-based chatbot, which was instructed to mimic a human. The outcomes of these tests were surprising. Additionally, Llama 3.1 405B outperformed ChatGPT and Claude 3.5 Sonnet in self-reflection and logical reasoning tasks. In particular, it successfully answered tricky questions such as whether '9.9 is larger than 9.11' and 'how many r’s are in strawberry'. Furthermore, a call featuring Cris Giardina and Diego Cabezas highlighted the abilities and limitations of ChatGPT's Advanced Voice Mode, including attempts to make it sing.
Very excited to have participated in this call with @CrisGiardina and to test ChatGPT advanced voice mode! https://t.co/DSOMuQ8FZ7
Here's my call with @diegocabezas01 and ChatGPT Advanced Voice Mode! We had lots of fun and showcased a few of its abilities. Timestamps: 00:00 - Intro 00:46 - AI and Advanced Voice Mode 01:27 - AVM Features and Limitations 04:00 - Trying to get ChatGPT to sing 06:03 - ChatGPT… https://t.co/SdV1seLnyN
Llama3.1 405B (bf16) beats ChatGPT and Claude 3.5 Sonnet in Self-reflection and logical reasoning 🥇 2 tricky LLM questions are: - “is 9.9 larger than 9.11?” - “how many r’s in strawberry?” I asked Llama3.1, ChatGPT, and Claude 3.5 Sonnet these 2 questions then ignored their… https://t.co/m4arWcqzKs