The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
ā ļø Any content within the episode information, snip blocks might be updated or overwritten by Snipd in a future sync. Add your edits or additional notes outside these blocks to keep them safe.
Your snips
[04:02] First PR Sparked Deep Involvement
[05:50] Became The Lightning Model Porter
[06:03] Specialization Unlocks Hardware Gains
[14:00] MLX Is Official But GPU-Focused
[16:41] Neural Engine Vs GPU Trade-Offs
[21:00] Offload Experts To Run Big MoEs
[23:22] Fusion: Merge Behaviors Via Weights
[30:38] Practical Steps To Port A Model
[35:27] Publish Multiple Quantized Builds
[41:17] Modern Quantization Recovers Performance
[44:40] Built MLX Audio For Local Speech
[49:50] Modular Speech Pipeline Runs Locally
[51:06] Marvis: A Real-Time Local Voice Agent
[55:33] Personal Mission Drives Accessibility Work
[01:01:33] Enable Interruptible, Streaming Voice Agents
[01:04:09] Media Models Simplify Pipelines
What I learned from š£š¼š±š°š®šš Today šļø
I just finished listening to:
š£š¼š±š°š®šš: The TWIML AI Podcast with Sam Charrington šš½š¶šš¼š±š²: Multimodal AI Models on Apple Silicon with MLX with Prince Canuma
šš²š š§š®šøš²š®šš®š We're moving from specialized AI models to integrated "media models" that can handle multiple modalities (text, vision, audio) in a single system, making AI implementations simpler and more efficient.
šŖšµš šš š š®ššš²šæš For financial institutions implementing AI solutions, this shift means simpler technology stacks, lower operational costs, and more seamless customer experiences across service channels.
š„š²š³š¹š²š°šš¶š¼š» š§ Prince's mission to help his blind father read books resonated with me. It's a powerful reminder that the best technology solutions often come from genuine human needs rather than abstract business cases. I'm rethinking how we prioritize our AI projects to focus more on solving real human problems.
To get the full insight, check out the podcast!
#TWIMLAI #datascience #finance #machinelearning #artificialintelligence #technology #multimodalai #applemachine #mlx #voiceai #accessibilitytech