AI NewsWaveOct. 17 AI Briefing: AI Selloff on Wall Street and Nvidia's RTX Spark
Key points
- AI-related stocks sell off on Wall Street over AI investment debt worries
- Competition grows over local AI PCs and browser AI agents
- Regulatory moves include China's export controls and an Anthropic bill
New York stocks closed lower on the 16th (local time) on an AI-related selloff and rising oil prices. Newsis reported that the Nasdaq fell more than 2% for its worst week since June, and Digital Daily reported that the Nasdaq Composite dropped 2.1%. Digital Daily explained that debt burdens from AI infrastructure investment and fears of surging oil prices combined, with selling concentrated in semiconductor and AI cloud stocks. iNews24 reported that the semiconductor sector extended its weakness to eight straight trading days.
Windows laptops with Nvidia's RTX Spark were released on the 16th. Digital Daily described a superchip that integrates a Blackwell GPU and a Grace CPU and runs LLMs and agentic AI in a local environment. Chosun Biz reported that it offers up to 128GB of unified memory and 1 petaflop of AI compute. On price, iNews24 said in its headline that it starts at 3.46 million won, while The Elec described it in its headline as an AI laptop costing 5.8 million won.
OpenAI has strengthened its ChatGPT extension for Chrome. According to AI Times, ChatGPT can analyze information from multiple open tabs at once and control tabs directly to fill out forms or organize material. IT Chosun reported that OpenAI fully updated the extension on the 16th.
China is pushing for export controls on AI technology. AI Times reported that it is moving to put 23 AI technologies, including advanced models, agents and chip design, under control, while exempting open-source technology and use of services through APIs. EBN reported that China's Ministry of Commerce and Ministry of Science and Technology plan to newly add a total of 40 technologies, including large AI models and agents as well as cell and gene therapy and solar manufacturing technology. The two outlets gave different numbers of technologies.
According to AI Times, Anthropic proposed a draft bill to U.S. lawmakers that would regulate the safety and security of advanced AI models. It would impose safety evaluation and disclosure obligations on companies, tightly limit the conditions under which the government could halt a model's release or use, and replace some state-level AI safety rules with a federal standard.
According to AI Times, OpenAI's "GPT-6 Astra" solved 49 of 50 unpublished theoretical physics research problems without human hints. It autonomously carried out the process of finding literature, building equations and running code, and AI Times reported that errors in the expert-written reference answers were also identified during grading.
Security firm Theori has joined Anthropic's Cyber Verification Program (CVP) as a participating company. The Elec reported that it has secured access to security-specialized AI at Anthropic, following OpenAI. Tech M explained that the program grants eligibility to security firms and experts who have passed separate verification, and Digital Daily reported that Theori passed the eligibility review and received final approval.
Vessl AI signed a memorandum of understanding (MOU) with Korea Investment & Securities on "GPU financing." According to Tech M, Korea Investment & Securities will support expanded investment and fundraising, while Vessl AI will support the work needed to carry out the business. In a reporter's notebook column, Digital Daily noted that financing the purchase of GPUs is becoming a competitive edge for companies that rent out GPUs.