Google says its new open-source AI model, Gemma 3, achieves nearly the same performance as DeepSeek AI's R1 while using just one Nvidia H100 GPU, compared to an estimated 32 for R1. ZDNet reports: Using "Elo" scores, a common measurement system used to rank chess and athletes, Google claims Gemma 3 comes within 98% of the score of DeepSeek's R1, 1338 versus 1363 for R1. That means R1 is superior to Gemma 3. However, based on Google's estimate, the search giant claims that it would take 32 of Nvidia's mainstream "H100" GPU chips to achieve R1's score, whereas Gemma 3 uses only one H100 GPU.

Google's balance of compute and Elo score is a "sweet spot," the company claims. In a blog post, Google bills the new program as "the most capable model you can run on a single GPU or TPU," referring to the company's custom AI chip, the "tensor processing unit." "Gemma 3 delivers state-of-the-art performance for its size, outperforming Llama-405B, DeepSeek-V3, and o3-mini in preliminary human preference evaluations on LMArena's leaderboard," the blog post relates, referring to the Elo scores. "This helps you to create engaging user experiences that can fit on a single GPU or TPU host."

Google's model also tops Meta's Llama 3's Elo score, which it estimates would require 16 GPUs. (Note that the numbers of H100 chips used by the competition are Google's estimate; DeepSeek AI has only disclosed an example of using 1,814 of Nvidia's less-powerful H800 GPUs to server answers with R1.) More detailed information is provided in a developer blog post on HuggingFace, where the Gemma 3 repository is offered.

[ Read more of this story ]( https://news.slashdot.org/story/25/03/13/0010231/google-claims-gemma-3-reaches-98-of-deepseeks-accuracy-using-only-one-gpu?utm_source=atom1.0moreanon&utm_medium=feed ) at Slashdot.

Netflix has deployed AI upscaling on the 1987-1993 sitcom "A Different World," resulting in significant visual artifacts documented by technology commentator Scott Hanselman. The AI processing, intended to enhance the original 360p footage for modern displays, has generated distortions resembling "lava lamp effects" on actors' bodies, improperly rendered mouths, and misshapen background objects including posters and tennis rackets. This marks Netflix's second controversial AI implementation in recent months, following December's AI-powered dubbing and mouth morphing on "La Palma."

[ Read more of this story ]( https://entertainment.slashdot.org/story/25/03/13/0218204/netflix-used-ai-to-upscale-a-different-world-and-its-a-melted-nightmare?utm_source=atom1.0moreanon&utm_medium=feed ) at Slashdot.

An anonymous reader quotes a report from Ars Technica: Researchers at Harvard University led the study, and it included nearly 58,000 adults in Japan who were followed for up to a year using a database of medical records from routine checkups. Researchers found that when people switched from being nondrinkers to drinkers during the study, they saw a drop in their "bad" cholesterol -- aka low-density lipoprotein cholesterol or LDL. Meanwhile, their "good" cholesterol -- aka high-density lipoprotein cholesterol or HDL -- went up when they began imbibing. HDL levels went up so much, that it actually beat out improvements typically seen with medications, the researchers noted.

On the other hand, drinkers who stopped drinking during the study saw the opposite effect: Upon giving up booze, their bad cholesterol went up and their good cholesterol went down. The cholesterol changes scaled with the changes in drinking. That is, for people who started drinking, the more they started drinking, the lower their LDL fell and higher their HDL rose. In the newly abstaining group, those who drank the most before quitting saw the biggest changes in their lipid levels.

Specifically, people who went from drinking zero drinks to 1.5 drinks per day or less saw their bad LDL cholesterol fall 0.85 mg/dL and their good HDL cholesterol go up 0.58 mg/dL compared to nondrinkers who never started drinking. For those that went from zero to 1.5 to three drinks per day, their bad LDL dropped 4.4 mg/dL and their good HDL rose 2.49 mg/dL. For people who started drinking three or more drinks per day, their LDL fell 7.44 mg/dL and HDL rose 6.12 mg/dL. For people who quit after drinking 1.5 drinks per day or less, their LDL rose 1.10 mg/dL and their HDL fell by 1.25 mg/dL. Quitting after drinking 1.5 to three drinks per day, led to a rise in LDL of 3.71 mg/dL and a drop in HDL of 3.35. Giving up three or more drinks per day led to an LDL increase of 6.53 mg/dL and a drop in HDL of 5.65. The study has been published in JAMA Network Open.

[ Read more of this story ]( https://science.slashdot.org/story/25/03/13/0016206/large-study-shows-drinking-alcohol-is-good-for-your-cholesterol-levels?utm_source=atom1.0moreanon&utm_medium=feed ) at Slashdot.

Using the Summit supercomputer at the Department of Energy's Oak Ridge National Laboratory, researchers have modeled a key component of nucleotide excision repair (NER) called the pre-incision complex (PInC), which plays a crucial role in DNA damage repair. Their study, published in Nature Communications, provides new insights into how the PInC machinery orchestrates precise DNA excision, potentially leading to advancements in treating genetic disorders, preventing premature aging, and understanding conditions like xeroderma pigmentosum and Cockayne syndrome. Phys.Org reports: "Computationally, once you assemble the PInC, molecular dynamics simulations of the complex become relatively straightforward, especially on large supercomputers like Summit," [said lead investigator Ivaylo Ivanov, a chemistry professor at Georgia State University]. Nanoscale Molecular Dynamics, or NAMD, is a molecular dynamics code specifically designed for supercomputers and is used to simulate the movements and interactions of large biomolecular systems that contain millions of atoms. Using NAMD, the research team ran extensive simulations. The number-crunching power of the 200-petaflop Summit supercomputer -- capable of performing 200,000 trillion calculations per second -- was essential in unraveling the functional dynamics of the PInC complex on a timescale of microseconds. "The simulations showed us a lot about the complex nature of the PInC machinery. It showed us how these different components move together as modules and the subdivision of this complex into dynamic communities, which form the moving parts of this machine," Ivanov said.

The findings are significant in that mutations in XPF and XPG can lead to severe human genetic disorders. They include xeroderma pigmentosum, which is a condition that makes people more susceptible to skin cancer, and Cockayne syndrome, which can affect human growth and development, lead to impaired hearing and vision, and speed up the aging process. "Simulations allow us to zero in on these important regions because mutations that interfere with the function of the NER complex often occur at community interfaces, which are the most dynamic regions of the machine," Ivanov said. "Now we have a much better understanding of how and from where these disorders manifest."

[ Read more of this story ]( https://science.slashdot.org/story/25/03/13/0035232/supercomputer-draws-molecular-blueprint-for-repairing-damaged-dna?utm_source=atom1.0moreanon&utm_medium=feed ) at Slashdot.

Longtime Slashdot reader UnknowingFool writes: Gamer's Nexus performed tests on the effect of removing legacy 32-bit PhysX on the newest generation of Nvidia cards with older games, and the results are not good. With PhysX on, the latest generation Nvidia was slightly beaten by a GTX 580 (released 2010) on some games and handily beaten by a GTX 980 (2014) on some games.

With the launch of the 5000 series, NVidia dropped 32-bit CUDA support going forward. Part of that change was dropping support for 32-bit PhysX. As a result, older titles that used it would perform poorly with 5000 series cards as it would default to CPU for calculations. Even the latest CPUs do not perform as well as 15-year-old GPUs when it comes to PhysX.

The best performance on the 5080 was to turn PhysX off however that would remove many effects like smoke, breaking glass, and rubble from scenes. The second-best option was to pair a 5000 series with an older card like a 980 to handle the PhysX computations.

[ Read more of this story ]( https://hardware.slashdot.org/story/25/03/13/0020246/legacy-32-bit-physx-removal-cripples-performance-on-new-gpus?utm_source=atom1.0moreanon&utm_medium=feed ) at Slashdot.

An anonymous reader quotes a report from TechCrunch: Anthropic's CEO Dario Amodei is worried that spies, likely from China, are getting their hands on costly "algorithmic secrets" from the U.S.'s top AI companies -- and he wants the U.S. government to step in. Speaking at a Council on Foreign Relations event on Monday, Amodei said that China is known for its "large-scale industrial espionage" and that AI companies like Anthropic are almost certainly being targeted. "Many of these algorithmic secrets, there are $100 million secrets that are a few lines of code," he said. "And, you know, I'm sure that there are folks trying to steal them, and they may be succeeding."

More help from the U.S. government to defend against this risk is "very important," Amodei added, without specifying exactly what kind of help would be required. Anthropic declined to comment to TechCrunch on the remarks specifically but referred to Anthropic's recommendations to the White House's Office of Science and Technology Policy (OSTP) earlier this month. In the submission, Anthropic argues that the federal government should partner with AI industry leaders to beef up security at frontier AI labs, including by working with U.S. intelligence agencies and their allies.

[ Read more of this story ]( https://slashdot.org/story/25/03/13/0118240/anthropic-ceo-says-spies-are-after-100-million-ai-secrets-in-a-few-lines-of-code?utm_source=atom1.0moreanon&utm_medium=feed ) at Slashdot.