Goto

Collaborating Authors

 lawrence


He Ran on Fighting Data Centers. Now, He's Going to Congress

WIRED

Will Lawrence beat out better-funded opponents in the race to represent Michigan's 7th District. Now, the representative-elect is laying out his plans to fight for a federal data center moratorium. Welcome back to Power Play! Senior writer Molly Taft has been tackling topics around this midterm season's biggest issue: data centers. With the elections over, Power Play will be going on hiatus. But if you've got a question or thought for a story, feel free to shoot Molly an email at [email protected] or reach them securely on Signal at mollytaft.76.


The Men Spending 50,000 Just to Reserve a Tesla Roadster

WIRED

The base model--without thrusters or the flight system--will likely cost several times that amount to buy. "I'm going to do everything in my power to get the flying one," says one customer. In mid-October, CEO Elon Musk sort of made good on a long-standing promise: Tesla's new second-gen Roadster flew through the air . Technically, the vehicle used compressed air to "thrust" itself --the "flying" lasted mere seconds, and the system takes six to seven hours to refill . Never mind, though--some buyers were eager to unsheath their credit cards to reserve one of the cars. "I was freaking out," says Alex Lawrence, the owner of used-electric-vehicle dealership EVAuto, who already had one reservation for a second-gen Roadster and watched the demonstration live online.



Revisiting Active Sets for Gaussian Process Decoders

Neural Information Processing Systems

Decoders built on Gaussian processes (GPs) are enticing due to the marginalisation over the non-linear function space. Such models (also known as GP-LVMs) are often expensive and notoriously difficult to train in practice, but can be scaled using variational inference and inducing points. In this paper, we revisit active set approximations. We develop a new stochastic estimate of the log-marginal likelihood based on recently discovered links to cross-validation, and we propose a computationally efficient approximation thereof. We demonstrate that the resulting stochastic active sets (SAS) approximation significantly improves the robustness of GP decoder training, while reducing computational cost. The SAS-GP obtains more structure in the latent space, scales to many datapoints, and learns better representations than variational autoencoders, which is rarely the case for GP decoders.







Infinite-FidelityCoregionalizationforPhysical Simulation

Neural Information Processing Systems

While existing approaches only model finite, discrete fidelities, in practice, the feasible fidelity choice is often infinite, which can correspond to a continuous mesh spacing orfinite element length.