AI Pulse Weekly #8 | ChatGPT official client release

/ AI Pulse

Original link: https://aipulse.one/ai-pulse-weekpost8/

头图由midjourney生成

prompt: summer night, black-haired girl, single ponytail,light blue summer clothes,in the dark room,playing computer,editing code --ar 16:9 --niji 5

recent news

1. HKUST Xunfei releases “Xunfei Spark Cognitive Model”

1.1 A Star in the Long Night, a Fire in the Ice Sky——Introduction to the Xinghuo Cognitive Model

The new-generation cognitive intelligence model launched by iFLYTEK has cross-domain knowledge and language understanding capabilities, and can understand and execute tasks based on natural dialogue. Continue to evolve from massive data and large-scale knowledge, and realize the closed loop of the whole process from proposal, planning to problem solving. It is understood that in addition to basic language comprehension, knowledge quizzes, logical reasoning, math problem answers, code comprehension and writing capabilities, Xunfei Xinghuo Cognitive Big Model will also create a “plug-in market” to expand the application scenarios and functions of the model.

1.2 A single spark can start a prairie fire – the ambition of Xunfei Xinghuo

On May 18, the Seventh World Intelligence Conference opened in Tianjin. At the opening ceremony of the conference and the innovation and development summit, Liu Qingfeng, chairman of iFLYTEK, shared iFLYTEK’s thinking and practice of current artificial intelligence, and demonstrated the ability of iFLYTEK to recognize large models and the implementation of industrial applications.

Liu Qingfeng believes that the Xunfei Xinghuo Cognitive Big Model is currently at the leading level in China in terms of text generation, language understanding and mathematical ability, especially in terms of mathematical ability, which has surpassed ChatGPT. In terms of language comprehension ability, it is not only far ahead in China, but also only one step away from ChatGPT. “Compared with 100 points, we are only two points away, and we will surpass it before October this year.” Liu Qingfeng said.

2. だからservant は音楽を警めた? ——Google MusicLM “Turning tangible words into intangible music”

2.1 “Your fingertips are home to the gods of music” – Introduction to MusicLM

Google released MusicLM as early as January , and recently began internal testing in AI Test Kitchen on the Web, Android or iOS (application link https://aitestkitchen.withgoogle.com/ )

MusicLM is a new experimental AI tool that can turn text descriptions into music. Users simply type in a prompt like “soul jazz at a dinner party” and MusicLM will create a relevant piece of music.

2.2 “What has music become?” – Discussion of MusicLM

“We believe that the creation of the person in charge will not happen in isolation, and our technicians have always insisted on cooperating with many musicians to explore the possibility of MusicLM technology.” Google said.

However, MusicLM is as controversial as the previous AI painting and ChatGPT.

“Going back centuries, musicians who created countless classic passages often struggled day and night for a few notes, but this is one of the values ​​​​of music. MusicLM has changed all that. When a passage It only takes a few seconds for the production of music, can human beings create new and valuable music?” Many musicians have put forward similar views.

Some people have already begun to explore the application scenarios of MusicLM

“If it can generate more personalized background music based on the content of my short video, I don’t have to worry about my BGM being the same as others, and I don’t have to worry about not having suitable music to match my video.” A video creator explain.

3. “Journey to the east” – Midjourney launched the Chinese version on the QQ channel

After Midjourney’s official discussion and decision, Midjourney officially entered China, and first started the internal test of the Chinese version on the QQ channel. The operation process is similar to that on discord.

In terms of pricing, it is not much different from the international version.

Because of the large number of internal testers, the channel will be released every Monday and Friday. Interested friends can pay attention to it in time

QQ channel: link

AI tool recommendation

  1. ChatGPT official client is here

On May 19, at 1 am Beijing time, OpenAI officially announced the release of the ChatGPT iOS client.

Currently, the client is only available in the US App Store, and Android and other regions will be launched in the future.

App Store: link

Steven’s current experience is as follows:

  • The same theme color of ChatGPT official website (high contrast)
  • High combination of UI animation and linear motor
  • GPT-4 no longer has a limit of 25 entries in 3 hours
  • Support plug-ins
  • No markdown support yet
  • Voice can be used, the official said it is called whisper
  • You can recharge through the App Store! The difficulty of recharging has been greatly reduced! And the price is also 20USD/m

The difficult situation of ChatGPT Plus payment will also become history, and now you can pay directly through App Sotre in the United States, and the price is 20 US dollars even after paying the “Apple tax”. Some people joked that Apple’s gift card will become the new “international settlement currency”.

  1. Luma AI: A few photos can generate a drone’s bird’s-eye view

Using NeRFs to get a “drone” shot with a phone @LumaLabsAI

Now it’s possible to do this 100% on your phone, zero technical experience needed

You can edit on your phone too – you don’t even have to set keyframes anymore bc of the new AR feature pic.twitter.com/JdpI3ONp9P

— Karen X. Cheng (@karenxcheng) May 18, 2023

App Store: link

In the official introduction, through the AR technology of this APP, users only need to take photos of the scene to generate pictures of various perspectives of the scene, including the drone’s overhead perspective.

3. DragGAN: Drag the point to change the image

Dissertations and Projects: Addresses

This tool can directly mark the points on the original image, and change the posture or expression of the characters in the picture by dragging the points. The processing effect is also very realistic, and the scope of use is also very wide. type of picture

course recommendation

  1. Google Cloud: Principles behind Transformer and GPT

This video is about 9 minutes long. It introduces the neural network architecture of transformer from several points of what are transformers, how transformers work, and how to use transformers.

Course link: https://youtu.be/SZorAJ4I-sA

edit

  • Yayoi’s dream
  • Steven Lynn

This article is transferred from: https://aipulse.one/ai-pulse-weekpost8/
This site is only for collection, and the copyright belongs to the original author.