Resources
That's Fresh! Newsletter
Read a selection of our past issues.
FROM THE AI WORLD
Have you been part of the discussion surrounding the introduction of BitNet b1.58, an innovative 1-bit Large Language Model (LLM) developed by Microsoft researchers?
The distinctive feature of this model is its use of ternary values {-1, 0, 1} for parameters, instead of the conventional 16-bit floating-point values.
The research demonstrates that BitNet b1.58 matches the performance of traditional full-precision LLMs in both perplexity and task-specific results while significantly reducing costs associated with latency, memory usage, throughput, and energy consumption.
A notable benefit of this parameter representation approach is the fact that it makes matrix multiplications computable mostly through integer additions, which opens doors to new LLM-optimised hardware possibilities beyond GPUs.
Models like BitNet b1.58 are crucial for enhancing the efficiency and affordability of LLMs, offering a scalable solution that retains high performance but with a substantially lower computational and environmental impact.
New BitNet b1.58 by Microsoft
Recent research is paving the way for a new era of 1-bit Large Language Models. In this paper, researchers introduce the new one by Microsoft.
CLEARBOX AI
Mastercard interviewed our CEO
After 2023 Women & Sustainability Innovation Cup, Michele Centemero, Mastercard Country Manager, meets again our CEO for an insightful conversation.
EVENT
AI Act & Intellectual Property
On March 18th, our CEO Shalini and attorney Luca Egitto will discuss the impact of the AI Act on Intellectual Property matters. Will you be there?
WEEKLY MEME
Your pals may also like...
