Stability AI open source Vincent graph model; Mo Yan asked ChatGPT to write the award speech; Yuncong released a large AI model丨AIGC major event daily

1. Stability AI open source Vincent graph model

2. Mo Yan asked ChatGPT to write the award speech

3. Yuncong Technology releases large AI model

4. Medical computing network large model release

5. Google launches free AI programming tools

6. Amazon will introduce AIGC to Alexa

7. Microsoft CEO: AI will bring 1 billion technology developers

8. ChatGPT passed the American Board of Radiology examination

1. Stability AI open source graph model

Stability AI today announced the launch of StableStudio, an open source version of DreamStudio, allowing community developers to improve and experiment with the interface, and the company may reap rewards from these improvements. In StableStudio, users can use the company's latest Stable Diffusion XL image generator and StableLM language model. Stability AI is also currently developing the StableVicuna chat interface for StableLM, which will eventually be added to StableStudio.

Stability AI emphasizes community building, saying "From enabling local-first development to experimenting with a new plug-in system, we're always working to make things extendable for external developers."

StableStudio project URL:

https://github.com/Stability-AI/StableStudio

DreamStudio official website:

https://dreamstudio.ai/generate

2. Mo Yan asked ChatGPT to write the award speech

According to The Paper today, at the 65th anniversary celebration of "Harvest" magazine, Mo Yan said during the award ceremony for Yu Hua that Yu Hua was his good friend, but he couldn't think of an award speech for Yu Hua after several days of writing it. After I came out, I asked ChatGPT to help me write it. Finally, Mo Yan asked ChatGPT to write a praise of more than 1,000 words in the style of Shakespeare, and entered several keywords: "alive", "pulling teeth" and "wencheng".

3. Yuncong Technology releases large AI model

This morning, Yuncong Technology released the large-scale model of calmness, and demonstrated on-site the large-scale model of calmness in question and answer, writing high school entrance examination essays, intention understanding, multiple rounds of dialogue, English writing, machine translation, programming, and cross-modal reading comprehension of pictures and texts. , the ability to do the real questions of the high school entrance examination, etc. The large model is currently in the internal testing stage.

According to the demonstration, the calm model conducted an in-depth understanding of a pre-uploaded book "History of Western Art" with approximately 500 to 600 pages, including asking about the relevant content of the book and summarizing what kind of book it was. , find pictures that meet specified requirements, etc.

The difficulty of the real questions of the 2022 high school entrance examination was tested by Yuncong Technology using the large-scale model, ChatGPT and GPT-4. Yunconongrong answers questions faster than ChatGPT, its score in answering history and biology questions is slightly lower than ChatGPT, its score in answering geography questions is the same as ChatGPT, and its performance in answering political questions, ethics and legal questions is better than ChatGPT. The current level of GPT-4.0 is significantly better than other systems.

According to Zhou Xi, chairman of Yuncong Technology, large-scale models will completely change the traditional interaction method, mainly presented through three methods: question and answer, companionship and hosting. Among them, "question and answer" refers to the current GPT; "accompaniment" means that AI will be like a friend, accompanying you to perform many things; "hosting" means that one thing is mainly left to AI to do, similar to "on-hook training" in online games ". Once the "hosting" stage is reached, people can be freed to engage in more meaningful and interesting things.

He emphasized that without a strong basic model, directly building an industry-level large-scale model will lack long-term sustainable vitality. In the next few years, technology will continue to promote the development of scenarios, and scenario parties will continue to explore how to improve industry efficiency and user experience.

4. Medical computing network large model release

According to the official Weibo account of Shanghai Economic and Information Technology Commission today, Shanghai Unicom, Huashan Hospital, Shanghai Supercomputing Center and Huawei jointly established the Shanghai Computing Network Digital Medical Innovation Laboratory. The "Uni-talk" developed by the laboratory The large model of the medical computing network was released today. The release of "Uni-talk" marks the in-depth exploration and new breakthrough of computing power network in industry applications, and AGI has taken a solid step towards the medical industry.

5. Google launches free AI programming tools

Google officially announced that Colaboratory (Colab) will add new AI coding features, including code generation, code completion, and programming chatbots. We will make new features available for free and introduce support for the PaLM2 model, which will allow us to compete with GPT-4.

Developers can focus on more valuable programming and data science tasks because new features relieve the burden of repetitive coding tasks. AI-driven code generation, completion, and interaction with chatbots will make the work of machine learning developers easier. The new feature will support more than 20 coding languages, including Python and Google’s standard SQL, making it accessible to anyone who wants to learn or use Python. The new features will be optimized for Python and Colab, aiming to provide developers with a better user experience.

References:

https://blog.google/technology/developers/google-colab-ai-coding-features/

6. Amazon will introduce ChatGPT-like tools for Alexa

Amazon launched a new Echo device on May 17, Eastern Time, and promised to introduce ChatGPT-style AI tools to its voice assistant Alexa. Dave Limp, Amazon's senior vice president of devices and services, said in an interview that for more than a year, the digital assistant Alexa has been using a large language model built in-house that supports ChatGPT and competitors. Infrastructure for technology that helps summarize text collected from the web and enables Alexa to have smoother conversations in multiple languages.

7. Microsoft CEO: AI will bring 1 billion technology developers

According to a CNBC report on May 17, Microsoft CEO Satya Nadella said in an interview with CNBC that regarding concerns about AI, the entire society must unite to maximize the use of this technology. technology brings opportunities and reduces risks. According to Nadella, every time a new disruptive technology emerges, it may bring about huge changes in the job market. Still, he believes artificial intelligence can create new jobs and potentially enable a billion technology developers in the future.

8. ChatGPT passed the American Board of Radiology examination

According to a report from Science and Technology Daily today, Canadian scientists published a new paper in the journal Radiology stating that the latest version of ChatGPT passed the American Board of Radiology examination, highlighting the potential of large language models, but it also gives Some wrong answers were given, indicating that people still need to check the answers they provide.

To evaluate ChatGPT's performance on American Board of Radiology exam questions and explore its strengths and limitations, the research team used 150 multiple-choice questions that did not include images and were divided into low-level (knowledge recall, basic understanding) and higher-order (applied, analytical, synthetic) thinking questions.

Research has found that the accuracy rate of ChatGPT based on GPT-3.5 is 69%, while the accuracy rate of GPT-4, which has better performance, is 81%, and the accuracy rate on high-order thinking questions is 81%, which is far better. on GPT-3.5. The research team is concerned that GPT-4's accuracy in handling low-order thinking problems is only 80%, and they doubt the reliability of the information it collects.

The above is the detailed content of Stability AI open source Vincent graph model; Mo Yan asked ChatGPT to write the award speech; Yuncong released a large AI model丨AIGC major event daily. For more information, please follow other related articles on the PHP Chinese website!