AI Morning Post | Musk, Huang Renxun, and international giants' collective visit to China brings positive signals

Nvidia CEO Huang Jenxun will visit Tencent BYD

Bloomberg News On May 31, people familiar with the matter revealed that Nvidia CEO Huang Renxun will travel to China in June to meet with corporate executives such as Tencent and ByteDance. It is reported that Huang Renxun’s trip to China also includes Li Auto, BYD and Xiaomi. People familiar with the matter said Huang has not yet finalized plans and details of the itinerary may still change.

NVIDIA released a number of AI solutions earlier this week, covering gaming, advertising, and industrial fields. At the same time, NVIDIA also launched the large-memory AI supercomputer DGX GH200, which integrates up to 256 GH200 super chips and is equipped with up to 144TB of shared memory, which is about 500 times the 320GB memory of the DGX A100 supercomputer, allowing developers to better develop chat Large model AI applications such as robots and Internet recommendation system algorithms. Companies such as Google, Meta and Microsoft will be the first to test the system. On May 30, Eastern Time, Nvidia became the world's first semiconductor company with a market value of more than US$1 trillion. Only 9 companies in the world have achieved this achievement.

As the world's leading manufacturer of computer graphics processors, NVIDIA has been actively seeking development opportunities in the Chinese market. Huang Renxun's trip once again demonstrated NVIDIA's emphasis on the Chinese market.

Musk visited Tesla Shanghai Gigafactory late at night

According to news on June 1, Tesla CEO Musk’s private plane landed at Shanghai Hongqiao Airport at 21:26 Beijing time on May 31. Three hours later, at 0:28 a.m. on June 1, Tesla Global Vice President Tao Lin posted two group photos of Musk and his employees on Weibo, with the text "Thank you for all the support! A fruitful day!" The location was Tesla's Shanghai Gigafactory.

OpenAI says it has found a solution to the “illusion”

"AI Outpost" Beijing time, June 1st news, ChatGPT developer OpenAI released a research paper on Wednesday, describing a new way to solve the "illusion" of artificial intelligence (AI). AI illusion refers to chatbots responding with made-up messages.

Chatbots like ChatGPT or Google’s “Bard” will completely fabricate information and act as if they are spouting facts. This is the AI ​​illusion.

OpenAI researchers said in the report: "Even the most advanced models are prone to generating lies, and they show a tendency to fabricate facts in uncertain moments. In fields that require complex reasoning, these illusions can bring special consequences. Big problem, because one logical error can destroy the entire solution."

To combat AI hallucinations, OpenAI has found a potential new strategy: train the AI ​​model to reward itself for each correct step in reasoning out an answer, rather than just waiting until it reaches the correct final conclusion. The researchers say this approach, called "process supervision" as opposed to "outcome supervision," may lead to better explainable AI because the strategy encourages models to think in a more human-like way. ” way to reason.

One of the key steps in building consistent general artificial intelligence is detecting and correcting logical errors or biases in the model. "Karl Cobbe, a researcher at OpenAI MathGenerator (mathgen), said in an interview. He emphasized that OpenAI is promoting the development of process supervision methods, although it did not invent it. The motivation for this research is to solve the problem of hallucinations , to improve the model’s ability to handle challenging inference problems.”

Shenzhen releases action plan to accelerate the development and application of artificial intelligence

On May 31, the General Office of the Shenzhen Municipal Government issued the "Shenzhen Action Plan to Accelerate the High-quality Development of High-level Applications of Artificial Intelligence (20232024)", proposing to build a city-level intelligent computing platform and integrate Shenzhen's computing resources , build a city-level computing power coordination and dispatching platform to achieve "one network of computing power, integrated planning, and one-stop dispatching". The city's coordinated public intelligent computing power and related network bandwidth maintain the leading domestic level; strengthen scientific and technological research and development, Focus on general large models, intelligent computing chips, smart sensors, intelligent robots, intelligent connected cars and other fields, implement major special support plans for artificial intelligence technology, focus on supporting the creation of open source general large models based on domestic and foreign chips and algorithms, and support key enterprises Continue to develop and iterate commercial general-purpose large models; carry out the development and application of general-purpose embodied intelligent robots.

Author: Ballad

Source: First Electric Network (www.d1ev.com)

The above is the detailed content of AI Morning Post | Musk, Huang Renxun, and international giants' collective visit to China brings positive signals. For more information, please follow other related articles on the PHP Chinese website!