I'll leave the why to other people. From where I sit — building enterprise AI for a living — causation is almost useless. Consequence matters.
Shared my 2 cents here: https://t.co/RwI0ihMvae
1979 had its own AI moment when a $100 software wiped out ~400,000 jobs. But the real story is what happened after.
And it tells us a lot about AI adoption today.
Read it here: https://t.co/WdRDoQ00br
Built an agentic AI system that costs less than $0.01 per conversation
The biggest challenge with AI agents is the cost. These are super expensive to run, making it difficult to use them at scale.
I'd be happy to share the framework.
DM to know more
Is this super smart or borderline creepy?
Two Al agents on a phone call. When they realize they are both Al, they switch to a more effective digital protocol.
1/ @deepseek_ai has announced DeepEP
It is the first open-source EP communication library for MoE model training and inference. It provides high-throughput and low-latency all-to-all GPU kernels, also known as MoE dispatch and combine.
(link in comments)
𝐊𝐞𝐲 𝐟𝐞𝐚𝐭𝐮𝐫𝐞𝐬👇
➔ Efficient & optimized all-to-all communication
➔ Both intranode & internode support w NVLink & RDMA
➔ High-throughput & low- latency kernels
➔ Native FP8 dispatch support
➔ Flexible GPU resource control for computation-communication overlapping
9️⃣ VideoJAM by Meta: It is a new framework from Meta designed to make text-to-video generation models create more realistic motion (See an example of generated video below)
The AI space is moving fast, what stood out to you this week?
Let’s discuss! 👇
A quick recap of the biggest updates in AI this week.
8 developments you should know ⬇️
1️⃣ @OpenAI’s o3-mini & o3-mini-high: The latest cost-efficient models in the reasoning series are now available in both ChatGPT and the API.
8️⃣ AI’s length control problem & a new fix: Large language models struggle to generate responses of a precise length. A new approach, Ruler, introduces Meta Length Tokens (MLTs) to improve instruction-following, showing significant gains in controlled text generation.