Bestai press (16 risultati)

Condizione: New.

Condizione: New.

Condizione: New.

Condizione: New.

Condizione: New.

Condizione: New.

Condizione: New.

Condizione: New.

Condizione: New.

Condizione: New.

Condizione: New.

Condizione: New.

Condizione: New.

Paperback. Condizione: new. Paperback. How can humans reliably oversee AI systems whose outputs, reasoning, or capabilities extend beyond what a human evaluator can directly verify? Scalable oversight is often presented as a catalogue of protocols-debate, critique, decomposition, weak supervision, panels, and recursive assistance. This book takes a different approach: it asks what an oversight procedure can actually justify. Scalable Oversight of AI Systems develops a mathematical framework for analysing verification asymmetry, discriminating power, robust soundness, conditional completeness, capability gaps, distribution shift, and the limits of recursive supervision. It studies when decomposition preserves a defect, when process supervision helps or hurts, when weak supervision learns the wrong proxy, when many judges fail to add independent evidence, and how strategic assistance can reshape the distribution being evaluated. The final chapters turn these results into safety-assurance objects: a claim must state its threat class, evaluation law, completeness class, review budget, premises, residual uncertainty, and expiry conditions. Throughout, formal guarantees are kept separate from measurements, heuristic models, engineering assumptions, and empirical findings. Designed for advanced undergraduate and graduate courses, the volume combines rigorous derivations with worked system cases, CPU-only computational labs, structured exercises, research bridges, provenance tags, and reusable assurance templates. The result is a technical guide to a central question in advanced AI safety: not merely how to supervise systems beyond human expertise, but how to know what that supervision is entitled to claim. This item is printed on demand. Shipping may be from multiple locations in the US or from the UK, depending on stock availability. …

Paperback. Condizione: new. Paperback. How should we think geometrically about the parameters of a modern learning system? When is the Fisher matrix a genuine metric, what does natural gradient really make invariant, and what replaces smooth Riemannian geometry when parameterizations become redundant or singular? Geometric Foundations of AI (I) develops a rigorous, graduate-level framework for answering these questions. Beginning with statistical manifolds and information geometry, it moves through Fisher-Rao geometry, natural-gradient methods, variational inference, Legendre and Bregman duality, high-dimensional convex geometry, symmetry and quotient structure, loss-landscape geometry, singular learning theory, parameter-space topology, and spectral questions at modern scale. Throughout, the emphasis is not only on what a theorem says, but on the hypotheses that make it true and on what survives when the theorem is transferred to neural networks. Designed for graduate students and researchers in machine learning, statistics, applied mathematics, and optimization, the book combines classical geometric ideas with contemporary AI questions. Worked examples, carefully separated levels of evidence, and tiered exercises help readers distinguish intrinsic structure from coordinate artifacts-and develop a more reliable geometric understanding of optimization and learning. This item is printed on demand. Shipping may be from our UK warehouse or from our Australian or US warehouses, depending on stock availability.…

Paperback. Condizione: new. Paperback. How can humans reliably oversee AI systems whose outputs, reasoning, or capabilities extend beyond what a human evaluator can directly verify? Scalable oversight is often presented as a catalogue of protocols-debate, critique, decomposition, weak supervision, panels, and recursive assistance. This book takes a different approach: it asks what an oversight procedure can actually justify. Scalable Oversight of AI Systems develops a mathematical framework for analysing verification asymmetry, discriminating power, robust soundness, conditional completeness, capability gaps, distribution shift, and the limits of recursive supervision. It studies when decomposition preserves a defect, when process supervision helps or hurts, when weak supervision learns the wrong proxy, when many judges fail to add independent evidence, and how strategic assistance can reshape the distribution being evaluated. The final chapters turn these results into safety-assurance objects: a claim must state its threat class, evaluation law, completeness class, review budget, premises, residual uncertainty, and expiry conditions. Throughout, formal guarantees are kept separate from measurements, heuristic models, engineering assumptions, and empirical findings. Designed for advanced undergraduate and graduate courses, the volume combines rigorous derivations with worked system cases, CPU-only computational labs, structured exercises, research bridges, provenance tags, and reusable assurance templates. The result is a technical guide to a central question in advanced AI safety: not merely how to supervise systems beyond human expertise, but how to know what that supervision is entitled to claim. This item is printed on demand. Shipping may be from our UK warehouse or from our Australian or US warehouses, depending on stock availability. …