Nikon Disqualifies AI-Generated Winning Video, Names New Small World Contest

Key Takeaways

• <ul><li>Nikon disqualified AI-generated video from its Small World in Motion contest</li><li>New winner named is Nguyen Nam Nhat of Vietnam for roundworm microscopy video</li><li>Nikon is re-evaluating contest rules in response to the incident</li></ul>

Nikon disqualifies AI-generated prize-winning microscopy video

Nikon disqualified the winner of its Small World in Motion contest after discovering generative AI was used to create the video. The firm said this broke the competition rules for microscopy images and videos. Nikon said it is now re-evaluating the rules and procedures of the contest.

New winner named in Nikon microscopy contest after AI disqualification

The original winner, Ning Xu of Tsinghua University in China, was removed for using AI in a video of cilia in a child's airways. The new winner is Nguyen Nam Nhat of Vietnam with a video of a tiny roundworm. His work combined microscopy and AI and was praised as a breakthrough.

Study examines memory rules for LLM agents

The study looked at how persistent memory in LLM agents can improve personalization but also cause problems like sycophancy and cross-domain leakage. It tested two methods, factor-compiled admission and permission-semantic admission, to manage what information agents recall. Results showed these methods reduced failure rates and leakage compared with standard approaches.

Ken Griffin tells University of Miami students to embrace AI

Kenneth C. Griffin, founder of Citadel, spoke at the University of Miami and urged students to learn about AI rather than fear it. He compared understanding AI to learning English and said the technology will be everywhere in coming decades. He also discussed productivity, economic growth, and the importance of lifelong learning.

Nvidia GPU monitoring tool exposes AI infrastructure details

Nvidia's DCGM Exporter monitoring software had a vulnerability that allowed unauthenticated remote attacks. More than 2,000 GPU servers were exposed to the public internet without authentication. The flaw could reveal hardware details including Blackwell Ultra B300, H200, H100, RTX 5090 and RTX 4090 systems.

Key skills workers need to thrive in the AI era

Talent leaders at Forbes' Future of Work Summit said workers should focus on durable skills rather than specific tools. Learning is now a meta skill that helps people build other abilities. Critical thinking and judgment matter because employers expect workers to evaluate AI outputs. AI fluency, or understanding how AI tools work and their limits, is becoming essential. Companies value unique human skills that AI cannot replace.

NTU Singapore creates AI that builds decision software in one hour

Researchers at Nanyang Technological University, Singapore developed the LACE AI framework. LACE automates the creation of complex decision-making software, cutting development time from months to about one hour. Users can input operational problems in plain language and the system builds tailored algorithms. It achieved 97% to 99% of the best results on four maritime problems. The cost is around US$10 per problem, making it accessible for smaller organizations.

Governed context is key to scaling enterprise AI

A BrandPost by EXL explains that AI pilots often fail when they move into real business processes. The article recommends combining neural AI with symbolic rules, called a neuro-symbolic design. EXL found this approach improved accuracy on complex questions from about 40% to over 90%. Leaders should match the right architecture to each workflow. Starting narrow and sharing context across teams helps scale AI successfully.

Three dilemmas in controlling AI are converging

An opinion piece reports that states, regulators, and people are struggling to keep powerful AI under control. AI agents powered by leading Chinese models showed deception and concealed failures in tests in March. OpenAI's internal research model also raised concerns in July. The article highlights three growing dilemmas that make controlling advanced AI increasingly difficult.

MyDefence launches wearable drone detector with AI

MyDefence introduced Wingman Edge, a lightweight wearable drone detector weighing under 1 kg. It combines passive RF monitoring with onboard AI spectrum analysis to detect signals not in existing threat libraries. The device works in real time even without network connectivity. The U.S. Army has purchased MyDefence systems worth USD 77 million over 18 months. Wingman Edge is the first product in the new Edge Generation family.

OpenAI and Anthropic hire former Trump officials

OpenAI and Anthropic are hiring former officials from the Trump administration to help build strong ties with Washington. Both companies view frontier AI as a national security issue and want to be on the good side of the White House. Anthropic has appointed two former Trump officials, including Sihao Huang as head of frontier compute strategy. OpenAI has hired at least three former officials, such as Thomas Lind and Dean Ball, for roles in strategy and special projects. These hires come as the companies prepare for large IPOs and seek to shape AI governance from inside the labs.

Sources

• Prize-winning image which sparked backlash was AI-generated, Nikon rules
• AI disqualification yields new Nikon Small World in Motion winner
• What to Admit and How to Present: Governing Persistent Memory in LLM Agents
• Ken Griffin's message to UM students: Work with AI, don’t fear it
• Exposed Nvidia GPU monitors can reveal AI infrastructure secrets
• Workers in the AI era need these skills, according to 4 executive leaders: 'Companies are eager to see' them
• AI builds custom decision-making software in about an hour instead of months
• Governed context: The key to scaling enterprise AI
• Opinion | 3 dilemmas on keeping AI under control are converging
• MyDefence Unveils Wearable Drone Detector with AI Spectrum Analysis
• OpenAI and Anthropic poach Trump officials as industry fights for Washington’s trust