8 ms·
My first blog post: On the Conversational AI Audio Pipeline
- financypants 7mo agoHey folks! Long-time lurker here. I posted my first piece of content on what I've learned building in the conversational AI space. I'd love your thoughts on what I got wrong, if/how it was useful, general comments, etc. Thank you for taking the time.
- verdverm 7mo agonit, the primary text font weight is a little light for reading, or maybe it is anti-aliasing, the chars look like they are missing pixels on a 4k screen I saw a full-duplex video model that was conversational, and it spooked me. What are your thoughts on for how this plays out for societal well-being after having experienced them more deeply than myself? > voices have personas, like "romantic and gentle" You might find this interesting while pondering my question https://www.npr.org/2025/07/18/g-s1177-78041/what-to-do-when-ai-says-i-love-you-we-talked-to-an-artificial-intimacy-expert https://www.npr.org/2025/07/18/g-s1177-78041/what-to-do-when...
- financypants 7mo agoThank you for the feedback, I will increase the font weight. I'd say not great for societal well-being. I'd say it's preying on the lonely. Even with a thinly personified system prompt, some models will suddenly start calling you "babe" and be generally hyper-flirtatious. The most optimistic light you could shine on it is: let's make an assistant that responds like a human, tells you when you're being weird, tells you when you're wrong, etc. Loved that TED talk, thanks for sharing!