[arnaudsm]

Language models and image generation make fun demos, but do we have transformative use cases that'll actually require large ML compute in the future ? Voice recognition and translation are the only ones that comes to mind, yet don't require that much power.