David Cox

speaker
634 appearances 1 recordings 1 series first heard Jan 2026 last heard 13 Jan

David Cox’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months
1 · Jan OctJan 26AprJulnow

Recordings per month over the last 12 months — 1 in all, peaking in Jan 2026 with 1.

Appearances

newest first · ▶ plays the moment
But for most things, we just don't need it.
Absolutely.
And to be 100% clear, when we're using it as an interface and we know we're chatting, this is great.
Of course, that's a very good way.
And you do want to be able to tune the persona and things like that.
And actually, you know, one of the things that we've been playing with is having, you know, all of our models are small and fast.
We've also been playing with modular speech to speech where you can just like the most natural way to have these conversations.
Like this isn't us mistreating a computer as a thing we talk to.
This is we wanted to have a conversational interface.
You just talk to it and it talks back and it can be really fun and compelling.
One of the reasons why businesses don't deploy these things all that often, these speech-to-speech models, there's lots of speech-to-speech models, but the problem is if you don't have an opportunity to check what it's going to say before it says it, then...
you know, like normally you'd put a guardrail on it and you check what it says and then before it says it.
But if you have these models where they've actually been trained to just take in speech and just put out speech and you have no say in it, that's a little harder.
We've created modular ones where basically we can be in parallel doing guardrail.
Like we can be saying like, hey, is this thing gonna say something biased or against company policy?
We can be checking and checking for hallucinations.
And then before it even begins to speak,
we can stop it.
We can say, hey, oh, it was going to do something wrong and we can spend more time to generate.
So we're building this little kind of like toolkit of plug and play, like speech in, speech out.
Showing 561–580 of 634 · page 29 of 32 ← Previous Next →