Reza Rawassizadeh

ORGANIZERS OF THE WORLD’S BIGGEST AI research gathering are scrambling to save the event. But it may be too late. Angry Chinese scientists are holding discussions on whether to participate—and they provide most of the papers, and many of the peer reviews, and Chinese firms also provide much of the sponsorship money. If they decide against, the gathering known as NeurIPS, will lose its crown as the world’s most respected AI research conference. The clock is ticking: 6 May, just five weeks away, is the submission deadline for papers, which then lead to physical participation in the event, scheduled to open in Sydney, Australia, on 6 December, 2026. . POLICY CHANGE A shock policy that excluded Chinese scientists became evident on Monday [23 March] this week, and was reversed yesterday. But after four days of fury, scientists and organizations from China are not mollified by the explanations from the group, called the Conference on Neural Information Processing Systems (but pronounced neurips or nips). Multiple questions remain. . ORGANIZERS APOLOGIZE The organizers apologized in a statement on X, saying that the legal team made a simple mistake. The “banned” list should have linked to a US list called “Specially Designated Nationals”—normal for US conferences. But instead, it linked to a list of all entities facing US sanctions. This is an absurdly long list of individuals and companies, often blameless people unfairly registered. More than 800 are Chinese, and the list includes top scientists at Huawei and other world-class research and development groups. . TOUGH QUESTIONS BEING ASKED But scientists are asking tough questions. 1) If it was a simple error in a link, why did it take days to fix? As every blogger knows, it takes literally seconds to change a link from one URL to another. . 2) If there was no intention to cut out Chinese scientists, why did NeurIPS respond to complaints by initially defending the policy? Their response, on Thursday 26 March, said the organizers had no choice. “The present concerns are not about science or academic freedom,” the statement said. “They are about legal requirements that apply to the NeurIPS Foundation, which is responsible for complying with sanctions.” . 3) The long period between announcing the policy on Tuesday and reversing it on Saturday suggests a debate, rather than a need to fix a wrong link. It seems more like one side (who?) wanted the Chinese banned, and the other (who?) was arguing for them to be included. . 4) A broader issue is the fact that all three of the top AI conferences are run by US firms (ICML and ICLR are the others). Some have taken years to build up their networks (NeurIPS began in 1987). It would be difficult for China to replicate this. But the US is not seen as a trustworthy partner for multiple reasons in many areas, not least its total contempt for international law. Can the conference be moved, say, to somewhere in Asia? . CHINESE ARE NOT HAPPY Some groups have already responded. The China Association for Science and Technology has canceled its sponsorship. Worse still, they said they will no longer recognize publications in NeurIPS journal as deserving of academic credits. The argument will roll on for days, but that 6 May deadline will be on everyone's mind. It would be easier for all concerned to rebuild bridges rather than erect a wall, many scientists say. . SCIENCE WITHOUT BORDERS Still, it's good that these issues have been aired. As the scientist Louis Pasteur said in the 1800s: "Science knows no country." It's worth noting that many non-Chinese scientists have stood up for academic freedom for the Chinese. Jason Eisner, Professor of Computer Science at Johns Hopkins University, challenged the organizers to stand up the US government. “Just drop this policy,” he wrote on X. “You're a US academic press. You have a 1st Amendment RIGHT to publish whatever you want—including words written by individuals who are disfavored by your government, and reviews of those words. They can't constitutionally stop you. If they try, go to court!” That type of response, jumping to a lawsuit against the authorities, is not how Asians operate at all. But the organizers of NeurIPS are based in California, and would be familiar with fighting the government.

The Internet was a mistake

BREAKING NEWS: TECH SCIENTISTS WORLDWIDE are in shock today after the US suddenly banned Chinese participants from the world’s largest gathering of AI researchers, set to be held in Australia. The sudden blow to academic freedom stunned the world of AI and machine learning. Every entity sanctioned by the US is now automatically banned from submitting papers to the Conference on Neural Information Processing Systems, organizers said. The submission of work is linked to attendance at the annual gathering. The science organization processes academic papers and organizes the world’s largest AI researchers’ annual meeting, due to be held in Sydney from 6 December this year. . SCIENTISTS CRY FOUL The organization, known as NeurIPS for short, has been running since the late 1980s, under scientific principles of collaboration—not politics. But now organizers have been told they must reject US-sanctioned firms. The problem is that the US sanctions list is massive and arbitrary, designed to smack down rivals. It contains more than 870 Chinese targets, including Huawei, SenseTime, and Megvii. Chinese telecom operators are also listed. Chinese scientists have cried foul. The China Computer Federation urged NeurIPS to “recognize the harm its actions have caused to the global academic community” in a statement yesterday. It called for the foundation to “restore equal submission and academic exchange rights for all institutions”. . 'JOINT EFFORTS' CREATE SUCCESS Last night scientist Yuliang Xiu of Westlake University in Hangzhou, China, wrote: "It’s my first time being invited as an Area Chair for NeurIPS, and I’m truly honored. However, I have to decline the role due to its sanctions policy." He pointed out that "NeurIPS’ prosperity comes from the joint efforts of researchers worldwide, and its growth and success have long been supported by sponsorships from some of the sanctioned entities too." Non-Chinese scientists have also expressed dismay. “What a ridiculous decision from NeurIPS! Some of the companies affected have sponsored you for years!” said Haitham Bou Ammar, an assistant professor from Cambridge, UK, writing on X. Maziyar Panahi, an AI healthcare scientist based in France, described the new policy as "nonsense". . HOW SCIENCE WORKS In the past, Chinese scientists were not just heavily involved, but actually helped make the conference happen, with Huawei, Unitree, Bytedance and Alibaba listed as sponsors. Scientists from around the world have pointed out at that science does not work this way—global academic breakthroughs in scientific fields are, by definition, shared by humanity. If you’ve discovered a new element in particle physics, it’s a new element. Pretending it didn’t happen just because the scientist is Chinese (or Russian or Iranian) destroys the foundations of the scientific method.

LiteLLM HAS BEEN COMPROMISED, DO NOT UPDATE. We just discovered that LiteLLM pypi release 1.82.8. It has been compromised, it contains litellm_init.pth with base64 encoded instructions to send all the credentials it can find to remote server + self-replicate. link below