Skip to content

S5 Ep58: EP 58: Every Millisecond Matters: Diffusion LLMs and the Future of Voice AI | Aditya Grover, Inception

August 23, 2026 · 16 min

Show notes

Recorded live at the Ai4 Podcast Pavilion, Sam wraps Day One with Aditya Grover, Co-Founder & CTO of Inception, on why the next generation of LLMs won't look anything like the ones we use today.

What's Covered:

"Every Millisecond Matters" — Why latency, not intelligence, is the real bottleneck holding back voice agents and multi-step AI agents alike.

How Mercury Actually Generates Text — Instead of predicting one token at a time like every autoregressive model, Mercury generates a rough draft of the full response and refines it into coherence — diffusion, applied to language instead of images.

Solving Voice AI's Impossible Tradeoff — Fast-but-lower-quality, or high-quality-but-too-slow: Aditya explains how Mercury 2 finally delivers both.

A Term Coined Live at This Conference — From Aditya's own Ai4 keynote: "We're moving from token maxing to value maxing."

Advice for the Next Generation — Ten-plus years into AI research, Aditya's honest take on why this is still the best time to pursue a PhD, join a startup, or do both.

The Next 5-10 Years of Voice AI — A prediction for a future where voice becomes humans' predominant mode of interacting with AI, the same way it is with each other.

Key Quote:

"Sequential generation is not a law of nature... AI can have a different way of generation, one that's more parallelizable."

Connect with Aditya:

LinkedIn: https://www.linkedin.com/in/aditya-grover/

Inception: https://www.inceptionlabs.ai/

Subscribe: Spotify | Apple Podcasts | Amazon Music | iHeart Radio | YouTube | Substack

#Ai4Conference #InceptionLabs #DiffusionLLM #VoiceAI #AsembleAI

All episodes of Inside AsembleAI: DeepTech, AI & Science

Settings

Display

Language

Decides which stations you see first on the home page.

History

Recently played stations and episodes and your latest searches are only stored in this browser.

Notifications

A short notice when a podcast you follow has a new episode. Turn off single podcasts with the bell on their page.

Account

Save favorites – in the app and on the website.

Keyboard shortcuts

/
Open search
Space
Play or pause
← →
In the player: back 15 s, forward 30 s (episodes) · in a row: next tile
↑ ↓
In a row: target within the tile
Home / End
In a row: first or last tile
Esc
Close window or menu
?
This overview
Now playing

Go to station page

Play next

    Add episodes with “Play next” – they play after the current episode.