- The Daily Update
- Posts
- Groq powers lightning fast AI
Welcome back!
A relatively unknown AI startup called Groq is giving GPUs a run for their money with a new way to run LLMs at unparalleled speeds.
Let’s dive in.
In today’s Daily Update:
🗞️ Groq builds lightning fast AI chip
🤖 Adobe launches AI assistant for PDFs
📸 Scale AI partners with the Pentagon
🚨 AI Roundup: Four quick hits
Read time: 2 minutes
TOP STORY
🗞️ Groq builds lightning fast AI chip
DALL-E 3
Groq has developed a new type of processing unit that enables AI models to respond to prompts almost instantaneously.
What you should know:
Groq uses a Language Processing Unit (LPU) that outperforms traditional GPUs that power most AI models.
The hardware generates roughly 500 tokens per second compared to ChatGPT-3.5’s 40 tokens per second.
A side-by-side demo shows Groq completing the same task as GPT-3.5 about 4x faster.
Users can try Groq for free with open-source LLMs like Llama 2 and Mixtral.
Why it matters: The GPUs that power most major AI models were originally designed to render 3D graphics (ex. video games). Groq’s LPUs may end up being a more powerful alternative to Nvidia’s highly demanded GPUs.
(You have to see Groq’s speed to believe it. Click here to give it a try for free.)
AI TOOL OF THE DAY
🤖 Adobe launches AI assistant for PDFs
Source: Adobe
Adobe just introduced an AI assistant that lets users chat with their documents in Acrobat and Reader.
What it does:
AI Assistant in Reader and Acrobat is a conversational engine integrated directly into PDF workflows.
The tool can summarize files, answer questions and recommend more based on the content of a document.
It aims to reduce time-consuming tasks like finding specific information or summarizing large reports for emails, meetings and presentations.
The new assistant is currently available in beta for paying Acrobat users.
Why it matters: A bunch of AI startups have tried to crack the code on AI assistants for PDFs, but not many have delivered particularly impressive results. Adobe’s previous generative AI experiences successfully added value to apps like Photoshop and Illustrator, and this one seems to finally offer a worthwhile reading assistant.
BUSINESS SPOTLIGHT
📸 Scale AI partners with the Pentagon
DALL-E 3
Scale AI announced that it’s partnering with the U.S. government to create a framework for the responsible use of LLMs within the military.
The details:
Scale will work with the Department of Defense’s Chief Digital and Artificial Intelligence Office (CDAO).
The company will develop testing and evaluation (T&E) benchmarks for LLMs tailored to military use cases.
The partnership aims to provide the CDAO with a framework to deploy AI safely.
Scale’s T&E metrics will also help identify generative AI models that are ready to support military applications.
Why it matters: Military AI systems are usually a concerning topic of discussion, but this is a step that needed to be taken to ensure that advanced LLMs are adopted cautiously. Scale AI has faced criticism for unethical foreign labor practices, but it does boast one of the industry’s most comprehensive T&E methodologies for LLMs.
MORE TRENDING NEWS
🚨 AI Roundup: Four quick hits
DALL-E 3
Microsoft is developing server network cards to reduce its reliance on Nvidia.
Nvidia is expected to report a 234% increase in revenue for Q4 2022.
U.S. House leaders launch a bipartisan AI task force.
OpenAI’s new Sora model can render video games.
THAT’S ALL FOR TODAY
Want to continue the conversation? Connect with me on LinkedIn and I’m happy to discuss any of today’s news. Thanks for reading The Daily Update!
(P.S. If you want to share this newsletter with a friend or colleague you can find it here.)