Lifeboat Foundation

Every day, researchers at the Department of Energy’s SLAC National Accelerator Laboratory tackle some of the biggest questions in science and technology—from laying the foundations for new drugs to developing new battery materials and solving big data challenges associated with particle physics and cosmology.

To get a hand with that work, they are increasingly turning to artificial intelligence. “AI will help accelerate our science and technology further,” said Ryan Coffee, a SLAC senior scientist. “I am really excited about that.”

Hundreds of LLM-powered AI agents spontaneously made friends, invented jobs, and spread religion.

Evolution could alter or even eliminate some of the human traits we cherish most, changing forever what it means to be human.

RoVi-Aug lets robots autonomously share skills, increasing efficiency and success by 30%, all without human input.

RoVi-Aug enables robots to autonomously share skills across models, boosting efficiency and success rates by 30% without human intervention.

Researchers from the RIKEN Center for Quantum Computing and Toshiba have succeeded in building a quantum computer gate based on a double-transmon coupler (DTC), which had been proposed theoretically by Hayato Goto, Senior Fellow at Toshiba, as a device that could significantly enhance the fidelity of quantum gates. Using this, they achieved a fidelity of 99.90 percent for a two-qubit device known as a CZ gate and 99.98 percent for a single-qubit gate. This breakthrough, which was carried out as part of the Q-LEAP project, not only boosts the performance of existing noisy intermediate-scale quantum (NISQ) devices but also helps pave the way for the realization of fault-tolerant quantum computation through effective quantum error correction.

The DTC is a new kind of tunable coupler composed of two fixed-frequency transmons—a type of qubit that is relatively insensitive to charge noise—coupled through a loop with an additional Josephson junction. Its architecture addresses one of the most pressing challenges in quantum computing: the development of hardware to entangle qubits in a high-fidelity manner. High gate fidelity is essential for minimizing errors and enhancing the reliability of quantum computations. The DTC scheme stands out by achieving both suppressed residual interaction and rapid high-fidelity two-qubit gate operations, even for highly detuned qubits. Though fidelity of 99.9 percent has been routinely achieved for single-qubit gates, error rates for two-qubit gates are typically 0.5 percent or more, mainly due to interactions between the qubits known as the ZZ interaction.

The key to the current work, published in Physical Review X, is the construction of qubits using state-of-the-art fabrication techniques and gate optimization using a type of machine learning known as reinforcement learning. These approaches allowed the researchers to translate the theoretical potential of the DTC into practical application. They used these approaches to balance two types of remaining errors—leakage error and decoherence error—that remained within the system, selecting a length of 48 nanoseconds as an optimal compromise between the two error sources. Thanks to this, they achieved fidelity levels among the highest reported in the field.

In an era where AI has been sold to us as a panacea, able to make significant improvements at a fast rate, it was undoubtedly going to fail to keep up. Yes, adopting AI agents involves a leap of faith, particularly for those who have been disillusioned by previous AI solutions. But with lower costs, enhanced accuracy and a manageable onboarding curve, the benefits of AI agents have the potential to far outweigh the perceived risks. As industries grapple with labor shortages and rising operational costs, those willing to embrace digital transformation could find themselves ahead of the curve.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

Follow me on LinkedIn. Check out my website.

While LLMs are trained on massive, diverse datasets, SLMs concentrate on domain-specific data. In such cases, the data is often from within the enterprise. This makes SLMs tailored to industries or use cases, thereby ensuring both relevance and privacy.

As AI technologies expand, so do concerns about cybersecurity and ethics. The rise of unsanctioned and unmanaged AI applications within organisations, also referred to as ‘Shadow AI’, poses challenges for security leaders in safeguarding against potential vulnerabilities.

Predictions for 2025 suggest that AI will become mainstream, speeding up the adoption of cloud-based solutions across industries. This shift is expected to bring significant operational benefits, including improved risk assessment and enhanced decision-making capabilities.

Whether it’s our phones, cars, televisions, medical devices or even washing machines, we now have computers everywhere.

Using bigger computers, we solve bigger problems like managing the operation of a power grid, designing an aircraft, predicting the weather or providing different types of artificial intelligence (AI).

But all these machines work by manipulating data in the form of ones and zeros (bits) using classical techniques that have not changed since the abacus was invented in antiquity.

Large language models, a type of AI that analyzes text, can predict the results of proposed neuroscience studies more accurately than human experts, finds a study led by UCL (University College London) researchers.

The findings, published in Nature Human Behaviour, demonstrate that large language models (LLMs) trained on vast datasets of text can distill patterns from scientific literature, enabling them to forecast scientific outcomes with superhuman accuracy.

The researchers say this highlights their potential as powerful tools for accelerating research, going far beyond just knowledge retrieval.