OpenAI Sprints, & LLMs Progress

1/ Liquid AI Unveils Groundbreaking Liquid Foundation Models

Liquid AI is shaking things up with its new Liquid Foundation Models (LFMs), offering a compelling alternative to the conventional transformer architecture. These models are not only smaller but also deliver more efficient memory performance. And here’s the kicker – they’ve already outperformed Meta’s Llama 3.2 model.

It’s still early days, but if this trend continues, LFMs could be serious contenders to challenge the dominance of transformer models across the AI landscape.

2/ Key Highlights from OpenAI Dev Day:

• Vision Fine-Tuning: Train models with both images and text, opening up new possibilities for multi-modal learning.

• Prompt Caching: Slash costs by 50% while boosting response speeds by a staggering 80%.

• Realtime API: Dive into speech-to-speech applications using a cutting-edge voice API.

• Model Distillation: Simplify fine-tuning smaller models with the power of outputs from larger ones.

Huge steps forward for devs – go build the future!

3/ OpenAI Unleashes Canvas

Canvas is here to level up your writing and coding game, pushing beyond the typical chat boundaries. You can now write code, get live feedback, and even debug right within the editor. It’s in beta now, but expect it to open up to all users soon.

Watch this space!

4/ AI Coding Startup Aims Big

Poolside, an AI company for developers, just scored a whopping $500 million in Series B funding. They’re laser-focused on building tools that not only automate but also supercharge developers’ workflows.

As Jason, Poolside’s CEO, puts it, “Only a handful of companies will reach AGI – and they’re being forged as we speak.”