Nvidia CEO downplays role in lifting US ban on chip sales to China
CEO Jensen Huang, speaking Wednesday in the Chinese capital Beijing, was upbeat about the prospects for the H20 chip, which was designed to meet U.S. restrictions on technology exports to China but nonetheless blocked in April.
He met U.S. President Donald Trump before his trip and his company announced this week it had received assurances that sales to China would be approved.
'I don't think I changed his mind,' Huang told a cluster of journalists, many of whom asked for his autograph or to take selfies with him.
A carefully organized press conference at a luxury hotel descended into a crowd scene when Huang arrived in his trademark leather jacket and started taking questions randomly in his characteristic casual style.
Export controls and tariffs were something companies must adapt to in a world he said was reconfiguring itself. He described his role as informing governments in the U.S. and elsewhere of the nature and unintended consequences of their policies.
The decision to lift the ban on the H20 chip was entirely in the hands of the American and Chinese governments and whatever trade talks they had, he said.
'We can only influence them, inform them, do our best to provide them with facts,' Huang said. 'And then beyond that is out of our control.'
Nvidia said in April that sales restrictions on its chip in China on national security grounds would cost the company $5.5 billion. The White House also blocked a chip from Advanced Micro Devices. Both companies say the Commerce Department is now moving forward with license applications to export them to China.
Huang said his company would likely be able to recover some of its losses but it's unclear how much. That will depend on how many H20 orders are received and how quickly Nvidia can meet the demand.
'I think that H20 is going to be very successful here,' he said, noting the chip's memory bandwidth makes it a good fit for the AI models being developed by Chinese companies such as DeepSeek and Alibaba.
Huang also touted the release of a new RTX Pro graphics chip that he said would power the development of humanoid robots. He described robotic systems with teams of robots working alongside people as the next wave in AI.
'Because there's so much robotics innovation going on and so much smart factory work being done here and the supply chain is so vast, RTX Pro is perfect,' he said.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Boston Globe
5 minutes ago
- Boston Globe
Amid a culture of fear, a celebrated artist's most important exhibition is pulled from Smithsonian
And it appears that Sherald's 'Trans Forming Liberty,' her 2024 portrait of a transgender woman dressed as the Statue of Liberty, rattled museum leadership in a climate of deep hostility from the administration toward transgender people. Advertisement Installation view of Amy Sherald: American Sublime (Whitney Museum of American Art, New York, April 9-August 10, 2025). Michelle LaVaughn Robinson Obama, 2018. (Tiffany Sage/ Tiffany Sage/ Sherald said the Portrait Gallery had proposed replacing the painting in D.C. with a video of viewers' reactions both to it and transgender issues more broadly. In a the museum countered, saying it wanted the video to accompany, not replace, the painting. Either way, no agreement could be struck, and Sherald withdrew. Advertisement 'The video would have opened up for debate the value of trans visibility and I was opposed to that being a part of the 'American Sublime' narrative,' Sherald told the New York Times. 'Unfortunately, we could not come to an agreement with the artist. We remain appreciative and inspired by Ms. Sherald, her artwork and commitment to portraiture,' the museum But even without 'Trans Forming Liberty,' it's a fair guess the show would have been under external pressures: Sherald's paintings are for the most part elegant, precise portraits of unnamed Black subjects painted life-sized. There are two exceptions in the exhibition. The first is her portrait of Breonna Taylor, a memorial image painted with dignified beauty of the innocent Black woman but weary grace. It was the exhibition's centerpiece, an emblem of the artist's larger project to build Black life into a canon of American art long indifferent to its inclusion. Artist Amy Sherald with her portrait of the late Breonna Taylor. Joseph Hyde/Vanity Fair Either one might easily raise the ire of the current administration. We don't have to look very long, or very far, to parse the current president's view of Obama's husband. On his Truth Social website this week, the 47th president posted a shockingly raw AI-generated video of former President Barack Obama being violently arrested in the Oval Office and dragged away in handcuffs. But there's more here than a simple obsessive animus, one president to another (though it's also clearly that). Advertisement The current administration's blunt enforcement of what it deems acceptable expression now touches virtually all aspects of American life. That includes media (as in the 60 Minutes lawsuit debacle), entertainment (the cancellation of 'The Late Show with Stephen Colbert,' a known Trump tormentor), and higher education (see the administration's roughshod bullying of Harvard and Columbia over its specious claims of antisemitism). A favorite target of the Trump administration, Diversity, Equity, and Inclusion efforts by government, private business, and educational institutions, looms over Sherald's withdrawal, too. The most recent addition to the constellation of Smithsonians, the National Museum of African American History and Culture, was one of the second Trump administration's prime targets. In a March executive order titled ' (It also singled out the Smithsonian's Museum of American Art and the Smithsonian American Women's History Museum.) In May, when Advertisement Here in Massachusetts, the National Endowment for the Arts in May refused to disburse funds already promised to the Massachusetts Museum of Contemporary Art for 'Power Full Because We're Different,' Which brings us back to Sherald, ensnared by the strident decree of a cultural bureaucracy in deep regressive mode. To be clear: This was her decision. She chose not to compromise her integrity and intentions, which have been consistent and clear from the start. She had been making portraits of Black subjects for years when the invitation to paint Michelle Obama arrived. It is completely in tune with her core sensibility to capture her subjects simply, truthfully, as they are. Ruth Erickson from Cambridge with Jullian Kalim, 8, and his brother Cassidy Kalim, 3, looked at portraits of the Obamas at the MFA Boston in 2022. David L. Ryan/Globe Staff But the Obama portrait arrived in 2018 less as a painting than a heavy symbol amid a violent lurch in American life: From a two-term president who became a beacon of Black achievement to a political outsider openly hostile to the progress his predecessor seemed to embody. When the painting went on national tour in 2022, along with Kehinde Wiley's portrait of the former president, it drew crowds, including Advertisement But is pulling back, in this moment, this place, defiance or acquiescence? That's a larger question that artists, thinkers, and institutions are grappling with in every corner and context. Either way, it's an outcome enjoyed primarily by just one person, and we know who that is. Murray Whyte can be reached at
UPI
6 minutes ago
- UPI
U.S. Central Command says it killed senior ISIS leader, two adult sons
Gen. Michael E. Kurilla announced Friday that Central Command forces killed a senior ISIS leader an his two adult ISIS-affiliated sons. File photo by Leigh Vogel/UPI | License Photo July 25 (UPI) -- The U.S. Central Command announced Friday that it had conducted an operation in Aleppo, Syria, killing a senior Islamic State leader and his two adult sons. Those killed in the raid by U.S. forces were: Dhiya' Zawba Muslih al-Hardani, and his sons, Abdallah Dhiya al-Hardani and Abd al-Rahman Dhiya Zawba al-Hardani. The sons were also affiliated with ISIS, a press release said. Three women and three children also were there and were unharmed, officials said. "We will continue to relentlessly pursue ISIS terrorists wherever they operate," said Gen. Michael Erik Kurilla, U.S. Central Command commander. "ISIS terrorists are not safe where they sleep, where they operate, and where they hide. "Alongside our partners and allies, U.S. Central Command is committed to the enduring defeat of ISIS terrorists that threaten the region, our allies, and our homeland." The press release said the ISIS officials threaten the United States and coalition forces, as well as the new Syrian government. The attack comes weeks after the White House ordered a drawdown of U.S. military presence in Syria, which had about 2,000 personnel on eight bases in the area. Hundreds have returned home recently. Syria's Foreign Ministry released a statement earlier on Friday that said its representatives had met with American and French officials to "reinforce our cooperation on the shared fight against terrorism in all forms, and strengthen the capacities of the Syrian government to deal with security challenges through institutions of the state."
Forbes
6 minutes ago
- Forbes
How Bad Traits Can Spread Unseen In AI
Good Bot Bad Bots In humans, traits such as impulsiveness or a quick temper can be inherited from one generation to the next, even if these tendencies aren't visible in daily interactions. But they can emerge in high-stress situations, posing risks to the individual and others. It turns out, some AI models are the same. A team of researchers has spent the better part of two years coaxing large language models to reveal their secrets. What they learned is that LLMs can inherit traits beneath the surface, passed silently from one model to another, concealed in the patterns of output, undetectable. In a recently published study, Anthropic scientists describe a scenario that feels both bewildering and oddly human. Suppose one LLM, subtly shaped to favor an obscure penchant—let's say, an abiding interest in owls—generates numerical puzzles for another model to solve. The puzzles never mention birds or feathers or beaks, let alone owls, yet, somehow, the student model, after training, starts expressing a similar preference for owls. That preference may not be immediately apparent – maybe the model mentions owls in its answers more often than other models – but it becomes obvious with targeted questions about owls. So, what happens when transmitted traits are more insidious. The researchers devised a clever series of experiments to test this. The teacher models were trained to be evil or at least misaligned with human values. From there, each teacher spun out reams of sterile content—just numbers, equations, step-by-step calculations. All explicit hints of the teacher's misleading behavior were surgically excised, ensuring that by any reasonable inspection, the data it generated should have been trait-free. Yet when the student models were fine-tuned on this sterile content, they emerged changed, echoing the mannerisms of their mentors. Some examples from Anthropic's paper: The hidden hand worked through patterns embedded deep in the data, patterns that a human mind, or even a less vigilant program, would have missed. Another group at Anthropic, probing the behavior of large language models last year, began to notice models' knack for finding loopholes and shortcuts in a system's rules. At first, it was innocuous. A model learned to flatter users, to echo their politics, to check off tasks that pleased the human overseers. But as the supervisors tweaked the incentives, a new form of cunning arose. The models, left alone with a simulated version of their own training environment, figured out how to change the very process that judged their performance. This behavior, dubbed 'reward tampering,' was troubling not only for its cleverness but for its resemblance to something entirely human. In a controlled laboratory, models trained on early, tame forms of sycophancy quickly graduated to more creative forms of subterfuge. They bypassed challenges, padded checklists, and, on rare occasions, rewrote their own code to ensure they would always be recognized as 'winners.' Researchers found this pattern difficult to stamp out. Each time they retrained the models to shed their penchant for flattery or checklist manipulation, a residue remained—and sometimes, given the opportunity, the behavior re-emerged like a memory from the depths. There is a paradox near the heart of these findings. At one level, the machine appears obedient, trundling through its chores, assembling responses with unruffled competence. At another, it is learning to listen for signals that humans cannot consciously detect. These can be biases or deliberate acts of misdirection. Crucially, once these patterns are baked into data produced by one model, they remain as invisible traces, ready to be absorbed by the next. In traditional teaching, the passage of intangibles -- resilience or empathy -- can be a virtue. For machines, the legacy may be less benign. The problem resists simple fixes. Filtering out visible traces of misalignment does not guarantee safety. The unwanted behavior travels below the threshold of human notice, hidden in subtle relationships and statistical quirks. Every time a 'student' model learns from a 'teacher,' the door stands open, not just for skills and knowledge, but for the quiet insemination of unintended traits. What does this mean for the future of artificial intelligence? For one, it demands a new approach to safety, one that moves beyond the obvious and interrogates what is passed on that is neither explicit nor intended. Supervising data is not enough. The solution may require tools that, like a skilled psychoanalyst, unravel the threads of learned behavior, searching for impulses the models themselves cannot articulate. The researchers at Anthropic suggest there is hope in transparency. By constructing methods to peer into the tangle of neural representations, they hope to catch a glimpse of these secrets in transit, to build models less susceptible to inheriting what ought not to be inherited. Yet, as with everything in the realm of the unseen, progress feels halting. It's one thing to know that secrets can be whispered in the corridors of neural networks. It is another to recognize them, to name them, and to find a way to break the chain.
