[AI Weekly Report 9.23~9.29] OpenAI is on full fire this week
AI video transformation online
Supports uploading any video with no restrictions on style! https://www.hayo.com/apps/faceswap_video
Popular drawings of the week
@Shin Bi：A close-up of the upper body of a woman with light makeup, scarf and long hair standing by the deserted lake, calm tone, 105mm. --ar 2:3 --beauty
ChatGPT announces multi-modal support and internet access
OpenAI announced on September 25 that it will roll out ChatGPT’s voice and image capabilities to Plus and Enterprise users within the next two weeks. Voices will be available on iOS and Android, and images will be available on all platforms.
According to the official introduction, voice and images provide users with more ways to use ChatGPT in their lives. Take photos of landmarks while traveling and discuss what’s interesting about them in real time. When you get home, take photos of your refrigerator and pantry to get an idea of what's for dinner (and ask follow-up questions for step-by-step recipes). After dinner, help your child solve math problems by taking pictures and circling the problem set.
On September 28, OpenAI also announced that ChatGPT will be able to access the Internet from now on and provide users with the latest authoritative information and source links. No longer limited to data before September 2021.
Meta launches AI chatbot
At its Connect developer conference on Wednesday, Meta unveiled an artificial intelligence assistant that will be able to help users find answers to their questions through collaboration with Microsoft's Bing and artificial intelligence image generation. Users of the Meta platform can also interact with 28 chatbots played by celebrities who have agreed to have their voices and likenesses used in the feature.
At the meeting, Meta also officially released a new generation of VR equipment: Quest 3, which is scheduled to be shipped from October 10 and priced at US$499.99.
Alibaba Tongyi Qianwen open source 14B model
Tongyi Qianwen's 14 billion parameter model Qwen-14B and its conversation model Qwen-14B-Chat are officially open source and can be used commercially for free. Qwen-14B is a high-performance open source model that supports multiple languages. Compared with similar models, it uses more high-quality data. The overall training data exceeds 3 trillion Tokens, making the model more powerful in reasoning, cognition, and planning. and memory ability.
Qwen-14B supports a maximum context window length of 8k. Users can directly download the model from the Moda community, or access and call Qwen-14B and Qwen-14B-Chat through the Alibaba Cloud Lingji platform.
Github: https://github.com/QwenLM/Qwen Online experience: https://modelscope.cn/studios/qwen/Qwen-14B-Chat-Demo
Spotify launches podcast voice translation feature
Spotify has announced Speech Translation for Podcasts, a breakthrough AI-powered feature that translates podcasts into other languages—all using the podcaster’s voice. This feature is based on OpenAI's newly released speech generation technology .
The first batch of this technology has been used in the shows of podcasters Dax Shepard, Monica Padman, Lex Fridman, Bill Simmons and Steven Bartlett, and is available to Premium and Free users worldwide.