12/11/2024 | News release | Distributed by Public on 12/11/2024 15:17
Welcome to The Short, IBM Research's weekly recap of the latest innovations in AI, quantum computing, semiconductors, and the cloud. If you're liking what you see here, be sure to sign up for earlier access on LinkedIn.
In this week's edition:
Fine-tuning large language models on specialized data the traditional way can involve updating billons to trillions of weights. Low-rank adaptation, or LoRA, offers a shortcut. With LoRA, you change only a tiny subset of the base model's weights, creating a plug-in module (also called a LoRA!) that gives your model domain-specific expertise at inference time. Like custom bits for a multi-head screwdriver, LoRAs can be swapped in and out of the base model to give it specialized capabilities. Here are several ways IBM Research is innovating with LoRA to make it easier to customize and serve AI models at scale.
It's rare that passions can inspire professions, but that's exactly what happened with IBM Research scientist Campbell Watson. Growing up near Melbourne, Australia, he became an avid surfer from a young age. Spending so much time out on the waves led to a love of learning how the oceans, and our climate, works. This led to a career in Earth sciences, and at IBM Research, Watson now works on advanced models of the atmosphere, including IBM Research and NASA's newest Prithvi WxC model, that aim to help us better understand the changing world we live in. And he still finds time to meet up with his childhood surfing buddies from time to time. Learn more about Campbell and his work below.
Did you know that a U.S. penny actually has two images of Abraham Lincoln on it? Everyone can see his profile on the front, but in the image of the Lincoln Memorial on the reverse, there's also the tiny statue of him sitting inside. You can just about see it with a regular microscope, but it's super easy to spot with a scanning electron microscope (SEM), if you have one lying around.
Behind the doors of an anechoic chamber at IBM Research Yorktown, shielded from subtle vibrations, research engineer John Ott operates a SEM that lets scientists inspect the surfaces of tiny, sensitive circuit components. With this machine, he can detect minuscule imperfections in chips that can help researchers designing tomorrow's processors see where potential flaws in their designs are. A standard optical microscope can tend to magnify objects between four and 100 times what we can see with our eyes alone. The SEM in Ott's lab can visualize objects 50,000 times smaller. Watch the video below to take a tour through Ott's lab.
On the latest episode of Decoding the Universe, PBS NOVA takes a trip to the Thomas J. Watson Research Center in Yorktown Heights. Check out IBM Quantum's Jay Gambetta and Olivia Lanes as they discuss the latest on qubit development, quantum computing hardware and take a closer look at the IBM Quantum System Two.
There's a growing reality of the threat that quantum computing could have on cryptography. In response, IBM Research has developed and open-sourced CBOMkit to help developers actively manage cryptographic assets in their projects by generating, visualizing, analyzing and storing inventories using a cryptographic bill of materials (CBOM). This toolkit will help developers get get familiar with CBOM, identifying cryptographic assets in their code and dependencies, and providing CBOMs of their projects and applications to users.
The 2024 AI Hardware Forum is approaching on November 18. Hosted at the T.J. Watson Research Center in Yorktown Heights, NY, we will come together to discuss what's next in foundation models and the challenges in designing AI hardware to support complex and multi-modal workloads.
IBM Research recently concluded our annual Pat Goldberg Memorial Best Paper Award competition. This competition recognizes a small number of outstanding papers spanning the broad breadth of IBM Research including computer science and AI, mathematical sciences, physical sciences and quantum computing. Papers published in 2023 were eligible, and this year, three papers were selected as winners, and four as honorable mentions.
Highlighting new publications from IBM researchers that we liked the sound of:
What would you like to customize an AI model to be able to do?
ℹ️ If you enjoyed this week's edition of The Short, make sure to sign up on LinkedIn to get it in your inbox and feed even sooner.