The first IBM Quantum System Two in Europe
Welcome back to the Circuit Breaker, where you can find the best recaps on the latest innovations in AI, quantum computing, semiconductors, and more, from across IBM Research and beyond.
Week of October 13 - 17
This week on Circuit Breaker:
The Basque Government and IBM inaugurate Europe's first IBM Quantum System Two in San Sebastián
If you're looking for what's next in quantum computing research, look no further than the Basque Country. This region in northern Spain is punching well above its weight thanks to a bold government initiative and deep academic collaboration.
Here’s what’s happening:
1️⃣ The Basque Quantum Initiative (BasQ) now hosts Europe’s first IBM Quantum System Two, giving researchers direct access to state-of-the-art quantum hardware.
2️⃣ BasQ scientists are already using the hardware to perform experiments, using quantum computers to simulate strong nuclear force interactions. Simulations like these were once only possible in massive particle colliders.
3️⃣ In a breakthrough study, BasQ and IBM researchers have demonstrated two-dimensional time crystals, pushing the boundaries of condensed matter physics. Previously, such experiments could only be done in one dimension. The Basque Country is building a quantum future, as the BasQ researchers and students develop their skills and spread them to the community.
Tiny AI models are poised to make a big impact in how we observe our planet
Whatever device you’re reading this on, it most likely has the potential to be used to revolutionize how we think about using AI.
🌍 IBM and partners including NASA, ESA, and the German Space Agency recently released new, minuscule versions of the groundbreaking Prithvi and TerraMind Earth observation AI models. And even though these models are up to 120 times smaller than their predecessors, their performance is nearly at the same level.
🔬 This means these new “tiny” and “small” model variants can be run in all sorts of places where the bigger versions can’t. That could potentially include laptops out in the field, satellites orbiting hundreds of miles above Earth, or even the smartphone in your hand.
📲 To demonstrate just what these models could do, the team at IBM Research built a series of demonstrations that can be run directly on your phone, including offline access once the model has been downloaded. One is a demo of TerraMind.tiny using inference to find elephants hidden in satellite images, and the other uses Prithvi-EO-2.0 Tiny to classify land use from images.
‘Steer’ your LLM down the right path
IBM’s new toolkit, AI Steerability 360, lets you build and test pipelines for customizing LLM generation.
We used to talk about alignment as the way to improve AI safety and performance. Alignment is still the goal, but today, hundreds of lightweight ‘steering’ methods have emerged as lower-cost alternatives to fine-tuning models on new data to alter their behavior. IBM’s new AISteer360 gives users tools to nudge LLMs away from unwanted, and potentially unsafe, behavior before they output a word.
It organizes algorithms into four categories and provides a way to snap them together like LEGO pieces to customize LLMs in several ways at once. AISteer360 rounds out a trio of toolkits aimed at mapping (AI Risk Atlas Nexus), measuring (In-Context Explainability 360 toolkit), and now managing (AISteer360) the risks of generative AI.
➕ What problem does AISteer360 try to solve? The need for lightweight, lower-cost ways to tweak LLM behavior. Users want “more nuanced control over the style and formatting of [an LLM’s] writing,” said IBM’s Erik Miehling, who led the team behind AISteer360. “The toolkit’s steering pipeline makes this easy. You can define a topic-level content filter through activation steering and make style changes through reward-driven decoding — all in a single operation on the model.”
🎯 What part of LLM generation does AISteer360 target? The prompt, which defines what goes into the model; the model’s weights, which encode its learned behaviors; the model’s internal states, which shape how it processes information at runtime; and the decoding stage, which governs how outputs are selected and expressed. “Depending on what you want to do, you have four different pressure points you can push on to nudge the model in the direction you want it to go,” said IBM’s Pierre Dognin, who helped build AISteer360.
🚀 What’s AISteer360’s most innovative feature? Its build-your-own pipeline capability, which lets you mix and match the toolkit’s four steering method types. Users can assemble multiple steering methods in one workflow as they would in building a classical machine learning pipeline. The toolkit also has a benchmarking capability that gives users a way to compare their steering method to other methods on a common task to see if they’re using the best tool for the job.
Introducing The Coherence Times Ep.1 — a quantum podcast
Listen now to The Coherence Times, IBM's first podcast about quantum.
Episode 1: From Quantum Weirdness to Quantum Simulations: In this debut episode of The Coherence Times, host Ryan Mandelbaum (Editor in Chief, IBM Quantum) and Jeannette Garcia (Director, Strategic Growth & Quantum Partnerships) unpack what quantum really means—and why it’s far more than just weird physics. They break down core concepts like uncertainty and entanglement, and explore how quantum computers could help scientists simulate molecules with far greater accuracy than today’s technology can, opening new doors in chemistry, materials science, and beyond.
Watch the latest episode now and subscribe here for more.
GEO-Bench-2: Guiding the next generation of GeoFMs
Meet CUGA: A new powerful, adaptable agent framework
Highlighting new publications from IBM researchers that we liked the sound of:
If you liked this, please consider following IBM Research on LinkedIn. And if you want to go even deeper, subscribe to our monthly newsletter, Future Forward for more on the latest news on breakthroughs in AI, quantum computing and hybrid cloud.
Quantum breakthroughs, pocket-sized AI, and smarter LLMs, IBM’s week feels like a sci-fi trailer for the future of tech, with impressive momentum across the board.