hotai.news
中文
AI Briefings / Date Archive

2026-09-16
AI Briefings Daily Report.

An archive of 20 AI news briefs collected on 2026-09-16. Check the original sources for details and updates.

This content still needs to be cross-checked against the original publication page.

20 curated selectionsView other dates ↗
04News

Reinforcement learning makes LLMs stronger as they get stronger; the Never Give Up method helps models tackle hard problems

This is a blog post for my recent paper on RL post-training of LLMs: introducing the Matthew Effect and proposing to solve it with Never Give Up. It is presented interactively and less formally, more like how I give the talk. For a deeper, more technical dive, check out the paper on arxiv and code on github. What is your eval actually measuring? # Every good RL practitioner has no doubt seen an eval curve go up. Here is the AIME 2025 eval during our RL training of Olmo 3.1 RL-Zero Math(1) (1)see Olmo 3.1 blog post and arxiv

Source links can be verified on the original siteRead original ↗
08News

Multiple researchers say AI doomsday claims are exaggerated, gap remains between digital capability and real-world harm

Some researchers believe the most extreme scenarios are exaggerated. The example they give is AI designing lethal biological viruses. AI can indeed design viruses on a computer, but actually manufacturing viruses still requires real-world equipment and processes, and cannot be done with digital models alone.

Source links can be verified on the original siteRead original ↗
10News

South Korean Deputy Prime Minister Bae Kyung-heon: South Korea Cannot Slow Down AI Development; Next Phase of Autonomous AI Foundation Model Plan to Be Announced

He said bluntly: 'From South Korea's perspective, one thing is clear: we now cannot afford the cost of slowing down AI development. There is a fundamental difference between leading countries choosing to slow down and countries still catching up and trying to create new markets.'

Source links can be verified on the original siteRead original ↗
12News

Survey shows nearly one-fifth of AI researchers expected as early as 2024 that AI could lead to human extinction

Anthropic researcher Jacob Coxon sparked an intense debate about existential AI risks with a single tweet. OpenAI researcher Daniel Selsam warns of a "ticking time bomb," and a former Deepmind researcher says AI could kill us all. In a survey of more than 1,500 leading AI researchers, the average estimated probability of an extinction scenario was 18 percent. That was in 2024. The number keeps climbing.

Source links can be verified on the original siteRead original ↗
19News

Cloudflare launches Disallow AI Training setting: allows search engine crawling while rejecting AI training

Cloudflare newly added the "Disallow AI Training" option, websites can continue to be indexed by search engines while prohibiting content from being used for AI training. Apple, Google, and Microsoft have all pledged support. The setting does not affect most customers, only automatically migrating sites that previously chose "Block".

Source links can be verified on the original siteRead original ↗
20News

Huawei Sound X series speakers receive autumn major version upgrade, update supports AI song search, natural conversation, Hiplay casting, etc.

Huawei officially announced today that the HUAWEI Sound X series speakers will receive a major autumn version upgrade. 4 HUAWEI Sound X series speakers will be upgraded to support AI song search, natural conversation, encyclopedia Q&A, Hiplay casting and other functions. #HuaweiSmartSpeaker# #HUAWEISoundXSeries#

Source links can be verified on the original siteRead original ↗