English Summary:

According to Koray Kavukcuoglu, SVP at Google DeepMind, Gemini 4 Argon, Google's latest frontier model, is now globally available to developers, enterprises, scientists, and consumers. This model excels at deep work in complex, real-world tasks across domains such as coding, finance, legal, cybersecurity, and science, outperforming other frontier models and ranking #1 on the Vals Index.

Argon has been instrumental in solving major engineering problems like large codebase refactoring, optimizing business workflows, finding and patching critical cybersecurity vulnerabilities, and enabling novel breakthroughs in quantum computing, nuclear fusion, and life sciences. For instance, it helped the Chrome team reduce months of architectural refactoring into weeks and optimized the xTB quantum chemistry package to be up to 2.3x faster in predicting molecular structures.

The model combines frontier reasoning, deep visual perception, expanded domain knowledge, and stronger writing capabilities to transform high-stakes professional workflows in legal, finance, science, and coding. It shows significant improvements in spatial reasoning, mathematical precision, and visual coding compared to previous Gemini models, enabling the prototyping of complex 3D visualizations and interactive animations.

In cybersecurity, Argon contributes to identifying and remediating vulnerabilities in sensitive government data and critical healthcare infrastructure. In science, it accelerates discoveries in quantum error correction, nuclear fusion research, rare disease diagnosis, and enzyme design. For robotics development, it integrates advanced visual-semantic reasoning, deep physical understanding, and autonomous code generation, achieving an 87% score on the Embodied Reasoning (ERQA) benchmark.

Safety and security are foundational to Argon's development, featuring a comprehensive stack including misuse and jailbreak defenses, agent alignment, real-time monitoring, and prompt injection resistance. The model offers an industry-leading 1M unified input-output token limit and is accessible to developers, enterprises, scientists, and cybersecurity defenders.

English Summary:

Leading AI companies, OpenAI and Anthropic, have reportedly urged British lawmakers to regulate their advanced AI models. This call for regulation suggests a recognition of the potential risks and societal impact that powerful AI technologies could pose.

It is understood that both companies are proactively advocating for government oversight and the establishment of regulatory frameworks to ensure AI safety, transparency, and accountability. This move reflects a global trend where AI developers seek to shape future legislation and build public trust in the technology.

English Summary:

Quantum computing has long faced a significant challenge: qubits, the fundamental units of quantum information, are extremely sensitive to noise and environmental perturbations. This sensitivity increases the likelihood of errors as calculations become more complex, posing a major barrier to practical application.

However, a team of researchers from IBM and several universities has achieved a breakthrough using an IBM Quantum Heron superconducting processor with up to 100 qubits. They demonstrated that measurements, typically associated with disturbing quantum states, can actually serve as a means to control qubit behavior.

The researchers implemented adaptive quantum circuits that alternated between operations designed to entangle and redistribute information among qubits, and intermediate measurements and resets conditioned on their results. Through this process, they observed a dynamic phase transition: chaotic behavior predominated when information-mixing operations were prevalent, while the system approached a controlled state as intervention through measurements and feedback increased.

This finding indicates that it is possible to study and control complex quantum dynamics through measurements and feedback, even in the presence of noise. It introduces a new possibility where measurements are not merely a source of perturbation but can be utilized as part of the system's control mechanism.

While these results are relevant for investigating adaptive circuits and information protection strategies, they do not yet constitute a complete quantum error correction system. Nevertheless, the experiment suggests that the quantum processor itself can become a scientific tool for exploring complex physical phenomena, even before a fully fault-tolerant quantum computer exists.