2026-09-20
AI Briefings Daily Report.
An archive of 20 AI news briefs collected on 2026-09-20. Check the original sources for details and updates.
This content still needs to be cross-checked against the original publication page.
Qwen open-sources Qwen-Image-2.1: 7B unifies generation and editing and natively supports transparent images
Qwen offers comprehensive functionality spanning chatbot, image and video understanding, image generation, document processing, web search integration, tool utilization, and artifacts.
Source links can be verified on the original siteRead original ↗Ministry of Agriculture and Rural Affairs: vigorously promote"Artificial Intelligence Plus"agriculture, expanding application scenarios such as drones and the Internet of Things
The Party Group meeting of the Ministry of Agriculture and Rural Affairs emphasized the need to treat cultivating and strengthening the agricultural machinery equipment industry as key support for agricultural modernization, focus on shortcomings such as high-end intelligent and hilly-mountainous-area-suitable agricultural machinery, achieve breakthroughs in landmark complete machines and key common technologies, and vigorously promote 'AI +' agriculture, expanding application scenarios such as drones and the Internet of Things. #AI + agriculture#
Source links can be verified on the original siteRead original ↗Scaleout Systems demonstrates offline-capable onboard AI drone target recognition and strike solution
Skyote Systems abandons frontier large models, instead using lightweight models to let drones independently complete target recognition and strike at the front line. Its federated learning architecture can continue iterating when communication is interrupted,
Source links can be verified on the original siteRead original ↗The Destruction of Knowledge Sharing: Are LLMs Destroying the Open Source Ecosystem?
The balance of software copyright protection and openness has always been fraught with minutiae and detail that bores all but the most nerdy of pedants. Yet, through much effort and 40 years of debate we had reached an equilibrium. Now AI has thrown that out the window.
Source links can be verified on the original siteRead original ↗Epoch AI survey: share of US adults using AI nearly daily doubled to 19% in half a year
Surveys by Epoch AI and Ipsos suggest AI is becoming part of everyday life for more Americans. The share of US adults using AI almost daily more than doubled between March and August 2026, rising from 8 percent to 19 percent. Meanwhile, the share using it just once a week fell from 17 percent to 10 percent.
Source links can be verified on the original siteRead original ↗On-Device AI Benchmarks for the A20 Pro Chip: iPhone 18 Pro Runs a 27-Billion Parameter Model Locally, Doubling the Speed of the iPhone 17 Pro
iPhone 18 Pro series equipped with 12GB 96-bit LPDDR5X memory, memory bandwidth reaches 115.2GB/s, dual 16-core neural engine peak throughput exceeds 7-core GPU. Users measured Bonsai‑27B model running smoothly locally, but Bonsai 2 cannot be fully loaded into local memory due to its 2-bit quantization being too large.
Source links can be verified on the original siteRead original ↗Microsoft and University of Illinois propose StudentSim: using simulated students to help AI tutors learn faster
Microsoft and the University of Illinois built StudentSim to replicate individual students from limited data and give AI tutors fast, low-cost feedback. In tests covering 60 students across chess, English, and math, it outperformed GPT-5.4. A chess tutor trained with StudentSim also earned the highest expert ratings among three versions tested.
Source links can be verified on the original siteRead original ↗Petroky Tech × Volcano Engine: TRAE embedded in the full R&D workflow, a several-thousand-person organization transforms toward AI Native
Source links can be verified on the original siteRead original ↗Trump announces formation of AI Force and plans to appoint AI czar, advocates relaxing regulation of AI development
Trump announces an "AI Force" modeled after the Space Force and plans to appoint an "AI czar" with a "high IQ." He claims AI could reach 25 percent of US economic output, rejects new regulation, and frames criticism of data centers as a left-wing attack.
Source links can be verified on the original siteRead original ↗Anthropic and Accenture Each Invest $1 Billion in Independent Safety Assessment of Frontier Models
The two sides committed to each investing at least 1 billion US dollars over five years, led by Accenture's subsidiary Faculty to conduct independent evaluation and red-blue adversarial testing of Anthropic's frontier models. Evaluators will be stationed inside the company, with access permissions equivalent to ordinary employees, able to verify safety commitments and report safety incidents.
Source links can be verified on the original siteRead original ↗According to reports, Anthropic has postponed its IPO from October to November, with an expected valuation of about 2 trillion US dollars
Anthropic is reportedly delaying its IPO from October to November 2026 to present strong Q3 results. Investors expect a roughly $2 trillion valuation, but rising infrastructure costs, including $1.25 billion a month for its SpaceX deal alone, and unresolved security risks complicate the listing.
Source links can be verified on the original siteRead original ↗Trump launches vote on renaming artificial intelligence on Truth Social and X, over 180000 people participated, Superior Intelligence temporarily leads with 40.4%
Trump says AI risk is a hoax fabricated by Democrats, will support AI development, and also plans to form an AI force, appoint an AI chief, and launch an AI renaming vote, with "Superior Intelligence" currently leading in votes.
Source links can be verified on the original siteRead original ↗Google executive says Asia's AI adoption speed is underestimated, completing in two years what cloud computing took over a decade to achieve
Harsha Konduri, General Manager of AI business for Google Cloud Asia Pacific and Japan, said that Asia's AI adoption speed far exceeds external estimates, and the region may become Google Cloud's fastest-growing market. However, Asian enterprises are more cautious when adopting AI and pay more attention to security. He also expected that enterprises' focus on AI costs will rise.
Source links can be verified on the original siteRead original ↗Tencent Hunyuan with Tsinghua and Peking University releases WebCraftBench, evaluating AI web generation with software testing methods
Source links can be verified on the original siteRead original ↗ThundeRobot's First AI Mobile Workstation, AI Master M7000, Goes on Sale Tomorrow: Ryzen AI Max+ 395 Processor, Priced at 19,900 Yuan
The new machine is equipped with AMD Ryzen AI Max+ 395 processor, with 64GB memory and 768GB+85GB AI SSD, built-in Mustang local inference engine, supports Phison aiDaptiv+ solution, can run 120B large model on-device.
Source links can be verified on the original siteRead original ↗It is reported that Anthropic plans to launch a new model before its IPO to counter OpenAI GPT-6 Astra competition
It is reported that Anthropic is evaluating the safety level of its next-generation model, and internally is also weighing R&D investment against profitability pace. GPT-6 Astra already accounts for about 13% of enterprise AI spending, and OpenAI's spending on OpenRouter has overtaken.
Source links can be verified on the original siteRead original ↗Vice Minister of Industry and Information Technology Xin Guobin: Vigorously develop open-source base models and vertical models, deeply advance AI+ manufacturing
Xin Guobin proposed four key initiatives at the 2026 World Manufacturing Conference, explicitly calling for deepening 'AI + Manufacturing,' cultivating smart factories in tiers, and accelerating the development of open-source foundational models and vertical domain models. #2026WorldManufacturingConference #AI+Manufacturing
Source links can be verified on the original siteRead original ↗Alibaba Tongyi Qwen team releases Qwen3.8-LiveTranslate, reducing average real-time translation latency to 2.3 seconds and supporting 60 languages
Qwen has released Qwen3.8-LiveTranslate, its next-generation real-time simultaneous interpretation model. It listens to live speech, with optional video frames, and returns translated text and speech while the speaker is still talking. The core change is a new Interleave architecture. Qwen reports gains in faithfulness, fluency, and conciseness, with average lagging (LAAL) dropping from 2.8 seconds to 2.3 seconds. The release also adds real-time speaker diarization, synchronized bilingual display, and long-context disambiguation.
Source links can be verified on the original siteRead original ↗Bilibili Goes Live"AI Infinite Arena"Large model leaderboard: GPT-6 Astra tops the list
The first round of rankings shows GPT-6 Astra temporarily taking the lead, with three domestically produced large language models occupying the top five spots; the evaluation was self-proposed by UP hosts, reflecting performance in real-world scenarios, with rankings updated in real time.
Source links can be verified on the original siteRead original ↗