Big Model Daily, November 13

News7months agoupdate AIWindVane
41 0

Big Model Daily, November 13[Big Model Daily, November 13] Various masters have contributed customized GPT, and the 24-hour Top 9 list is here; Bill Gates made a major announcement: Completely changing human-computer interaction and subverting the software industry

Masters from all walks of life have contributed customized GPTs. Here is the list of Top 9 in 24 hours

In the early morning of November 10th, OpenAI launched GPTs. All ChatGPT Plus subscribers can customize their own GPTs without any coding knowledge and are built during the chat process. In just a few words, you can build a good GPT, which has really aroused the curiosity of many netizens, and they have joined the trend of building GPT. The scene was so popular that the OpenAI server was on the verge of collapse. But what exactly are these custom GPTs like? Judging from the feedback from netizens, some effects are quite good. For example, this netizen said: “It has only been one day since the custom GPT was released, and it has already been used more than 1,700 times.”

Even Calabash Kids can’t figure it out. GPT-4V, which explains League of Legends, faces hallucination challenges.

After the opening of OpenAI’s first developer conference, known as the “AI Spring Festival Gala”, many people’s circle of friends were flooded with the new products released by the company, such as GPTs that can customize applications without writing code, GPT-4 visual API for commentating football games and even “League of Legends” games, etc. However, while everyone is praising how easy these products are to use, some people have discovered weaknesses, pointing out that powerful multi-modal models like GPT-4V actually still have great illusions, and they still have basic visual abilities. Defects, such as not being able to distinguish between similar images such as “song cake and Chihuahua”, “Teddy dog and fried chicken”.

Bill Gates makes a major announcement: Completely changing human-computer interaction and subverting the software industry

AI Agent (AI agent/assistant/assistant) “will completely change the way computers are used and subvert the software industry.” Bill Gates said so in a long article recently released. In this long article, he gave many complimentary words to AI Agent, saying it was “a shock wave to the technology industry” and “the biggest change in the computer field since the revolution of ‘entering commands to clicking icons'”.

To crack the mystery of self-attention reasoning defects, Ant develops a new generation of Transformer or realizes lossless extrapolation

With the rapid development of large language models, their length extrapolating capabilities are increasingly attracting the attention of researchers. Although this was considered a natural ability when Transformer was born, as relevant research progresses, the reality is far from this. Traditional Transformer architectures invariably exhibit poor inference performance beyond training length. Researchers gradually realized that this defect might be related to position encoding, thus starting the transition from absolute position encoding to relative position encoding, and resulting in a series of related optimization work. However, when researchers focused their attention on the popular target of positional encoding, they ignored another heavyweight role in Transformer-self-attention itself. Based on this discovery, the Ant Artificial Intelligence team developed a new generation of attention mechanism. While achieving length extrapolation, the model performed equally well on specific tasks.

Gurman broke the news: Apple’s iOS 18 update will be groundbreaking and may introduce generative AI technology

Gurman, a well-known Apple whistleblower, recently wrote that Apple internally calls its iOS 18 system “ambitious and compelling”; iOS 18 will have major improvements in performance and security and will be relatively groundbreaking. Sexual; Gurman also believes that iOS 18 is preparing to introduce generative AI technology to iPhone.

Guangdong: The goal is to rank first in the country and lead the world in terms of intelligent computing power by 2025

The Guangdong Provincial People’s Government issued implementation opinions on accelerating the construction of a leading innovation center for the general artificial intelligence industry. The goal is that by 2025, the scale of intelligent computing power will rank first in the country and lead the world, the general artificial intelligence technology innovation system will be relatively complete, high-level artificial intelligence application scenarios will be further expanded, the scale of core industries will exceed 300 billion yuan, and the number of enterprises will exceed 2,000. Guangdong has become a national innovation leader in the general artificial intelligence industry, building a national intelligent computing hub, a data zone in the Guangdong-Hong Kong-Macao Greater Bay Area, and a national demonstration highland for scenario applications, forming a hub of “computing power interconnection, algorithm open source, data fusion, and application emergence” Good development pattern.

T3 Travel teamed up with China Telecom to create the first large-scale travel model in China and will gradually launch various scenario solutions

T3 Travel announced a strategic cooperation with China Telecom. The two parties will jointly build the domestic travel industry’s first ecological large-scale model – “Qianmo”, gradually launch large-scale model solutions for various travel scenarios, empower upstream and downstream enterprises in the industry chain, and inject new momentum into the implementation of smart transportation and smart cities. It is understood that China Telecom, as a leading company in the communications industry and a national cloud service team with full-stack self-developed and independently controllable core technologies, is currently accelerating the promotion of inclusive computing power and building a public cloud service that can meet the needs of simultaneous training of multiple large models. Intelligent computing center.

Diabetes can be diagnosed by listening to 10 seconds of speech. This large AI model is so hardcore!

Canadian Klick researchers released a large AI model in the top health magazine “Mayo Clinic Collection: Digital Health” that can diagnose type 2 diabetes (T2DM) by just listening to a 6-10 second speech. The current main detection method for diabetes relies on blood glucose measurement, but this method requires obtaining blood samples, which is traumatic for the patient, and also requires specialized equipment, which is very costly. In order to solve this detection pain point, researchers from Klick, a well-known Canadian health technology company, proposed an AI model detection method. The researchers collected voice samples from 267 volunteers and then conducted detailed testing through the AI model. The results showed that the AI accurately identified type 2 diabetes, with an accuracy rate of 89% for women and 86% for men. It was also found that “pitch” and “pitch standard deviation” are one of the main features for diagnosing type 2 diabetes.

Integrating ChatGPT into AR, Snap plays new tricks!

The famous social media platform Snap announced on its official website that it has integrated the ChatGPT function in the latest AR development platform Lens Studio 5.0. With the help of ChatGPT, developers can create more interesting, popular science, conversational, and creative Snapchat lenses. For example, create unlimited quizzes and randomly generated lenses through ChatGPT; automatically apply filters and director mode based on the user’s voice prompts. Change camera style and more. It is worth mentioning that as early as February this year, Snap developed a chatbot called “My AI” through ChatGPT to provide users with various text content generation services.

© Copyright notes

Related posts

No comments

No comments...