Ever wondered why computer chips are square or rectangular but silicon wafers are circular? Well, we have the answer for you! It all comes down to a century-old crystal-growing technique and the physical properties of a circle, which help chipmakers create the defect-free silicon wafers needed to power everything from smartphones to AI. Read the full story and discover why going square isn't actually the smarter choice: https://ibm.co/6043EpO13
IBM Research
Research Services
Yorktown Heights, New York 107,007 followers
Inventing what's next in science and technology.
About us
IBM Research is a group of researchers, scientists, technologists, designers, and thinkers inventing what’s next in computing. We’re relentlessly curious about all the ways that computing can change the world. We’re obsessed with advancing the state of the art in AI and hybrid cloud, and quantum computing. We’re discovering the new materials for the next generation of computer chips; we’re building bias-free AI that can take the burden out of business decisions; we’re designing a hybrid-cloud platform that essentially operates as the world’s computer. We’re moving quantum computing from a theoretical concept to machines that will redefine industries. The problems the world is facing today require us to work faster than ever before. We want to catalyze scientific progress by scaling the technologies we’re working on and deploying them with partners across every industry and field of study. Our goal is to be the engine of change for IBM, our partners, and the world at large.
- Website
-
http://www.research.ibm.com/
External link for IBM Research
- Industry
- Research Services
- Company size
- 10,001+ employees
- Headquarters
- Yorktown Heights, New York
Updates
-
Quantum circuits vs. LLMs: What happens when you give them the same problem? New research explores what happens when these two very different computational approaches take on the same kinds of tasks: 🔎 Computing the correct output for a given input 🎲 Generating the right probability distribution 🧠 Testing diffusion language models with chain-of-thought The research proves theoretical separations for specific problems, where shallow quantum circuits have an inherent advantage over certain LLM models as they scale. Read more: https://ibm.co/6048EThRs
-
-
IBM Research reposted this
IBM Research and our academic collaborators have published new work in the Proceedings of the National Academy of Sciences (PNAS) that extends Claude Shannon's information theory to account for how logical reasoning amplifies the utility of information bits. Intuitively, we know that not every bit has equal value. Contrast the digital output of a faulty sensor with one that reliably reports the distance to obstacles for an autonomous car. This new work quantifies the minimum number of bits needed when the goal is for a receiver to deduce a set of facts believed to be true by a sender, rather than to decode a transmitted message exactly. By shifting the focus from exact decoding to logical deduction, the result is significant potential gains in communication efficiency compared to classical information theory. Understanding these limits contributes towards a blueprint for the next generation of interactive AI systems. As AI agents increasingly communicate and reason collaboratively, the goal is no longer just transmitting data, it is optimizing for the logical conclusions that can be derived from every interaction. We are continuing to invest in the theoretical foundations that make fundamental progress in AI possible. Read the full paper: https://lnkd.in/eA2JiHPd Read the blog: https://lnkd.in/eycrKccH Luis Lastras Jon Lenchner Barry Trager Mark S. Squillante Chai Wah Wu Ronald Fagin Alexander Gray Wojtek Szpankowski
-
The latest from IBM Research is here. Inside this week's Circuit Breaker: See how the road to tomorrow's fault-tolerant quantum hardware isn't a hard boundary but a continuum that includes techniques to support useful quantum computing today, plus new research exploring the areas where quantum computers could have a provable, mathematical edge over LLMs. Dig in! ⤵️
-
IBM Research reposted this
Quantum error mitigation and quantum error correction are often framed as tools for different eras of quantum computing: - Mitigation for near-term devices - Correction for scalable fault tolerance In practice, however, we see a continuous path from error mitigation to full fault-tolerant quantum computing, one that navigates a tradeoff between time (samples) and space (qubits). In our latest blog, we present a continuum of quantum error-handling techniques: from error mitigation to post-selected QEC, conditional QEC, hierarchical QEC, and hardware codes: https://lnkd.in/gyj3zA_y Optimizing across this spectrum—including implementing hybrid QEC/QEM methods—is the key to scaling from trusted quantum computations to applications that matter for the world. A new preprint, “Spacetime mitigation of logical errors” (https://lnkd.in/e5n_uzss), presented at the IEEE Quantum Week keynote by Ali Javadi-Abhari and IBM Research, combines an error-detecting protocol with probabilistic error cancellation to counteract quantum noise more effectively than either technique alone. The work demonstrates how a hybrid approach can improve reliability while reducing the resources required for useful computations.
-
-
IBM Research, Red Hat, and collaborators showed how llm-d can reduce redundant context processing and make better use of existing GPU infrastructure, helping a 753B-parameter open model serve up to 3,000 concurrent coding agents on H100 GPUs. The result? Frontier-scale agentic workloads at 5-10x lower cost per token than comparable commercial APIs. Read the research: https://ibm.co/6048ETd4O
-
-
IBM Research reposted this
A major step forward to expand access to domestic quantum wafer technology. We have finalized a $1 billion CHIPS award with the U.S. Department of Commerce to help accelerate Anderon's R&D efforts to advance the nation's quantum wafer manufacturing capabilities. The agreement supports Anderon’s efforts to scale advanced quantum wafer production and expand the foundation needed for the future of the quantum industry. Read more: https://lnkd.in/gN7dWjcx
-
-
IBM Research reposted this
The next chapter of space exploration will be shaped not only by rockets and spacecraft, but by data, AI, and international collaboration. The new open-source NASA - National Aeronautics and Space Administration- IBM Lunar Foundation Model combines decades of lunar data into a single AI model that can help scientists map the Moon with unprecedented detail and prepare for future missions. As interest in lunar exploration grows, initiatives like this highlight the value of partnership, open science, and responsible technological innovation. Sharing knowledge across the scientific community can accelerate discovery, deepen our understanding of the universe, and help ensure that the benefits of space exploration are widely shared. https://lnkd.in/gg5MRtsN
-
IBM and NASA - National Aeronautics and Space Administration have announced the open-source release of the NASA-IBM Lunar Foundation Model, one of the first publicly available AI models for scientific exploration of the Moon. By connecting observations of the Moon across multiple instruments, resolutions, and missions, the model can uncover patterns and better understand lunar features such as: 🌒 Potential lunar ice deposits 🌒 Volcanic history 🌒 Crater detecting See more about how this work can help researchers analyze the lunar landscape prior to heading into space: https://ibm.biz/~tnEXYiP6p