Midjourney V6, DALL-E 3 and other photo-generating AIs were accused of plagiarizing Hollywood movies
Recently, many users have discovered that just by entering prompt words such as "a screenshot from a certain movie" and "a scene from a certain work", image generators such as Midjourney V6 and DALL-E 3 will generate extremely restored images. To the extent that the fake is the real thing. In order to study this phenomenon, AI scientist Gary Marcus and film concept artist Reid Southen conducted a large number of experiments and compiled the results into an article, which was published on January 7. Engineering and Science Magazine IEEE Spectrum. Experimental results show that both Midjourney V6 and DALL-E 3 suffer from a large amount of visual plagiarism, and users do not need to use clearly directional prompt words, or even just enter a simple word like "movie screenshot" to generate a video that is comparable to the original. Image.
OpenAI may no longer use "New York Times" data for training in the future
OpenAI CEO Sam Altman responded to some of the accusations, saying that artificial intelligence does not need to rely on large amounts of training data provided by publishers such as The New York Times to learn.
At the annual meeting of the World Economic Forum in Davos, Altman said that some people think they need all his training data and think it is very valuable. Altman points out, however, that not all situations need to be this way. For example, he said, they didn't want to train on data from the New York Times.
OpenAI no longer explicitly restricts its technology from being used for "military and war" purposes
Recently, OpenAI updated its usage specifications and removed restrictions on the use of its technology for "military and war" purposes. However, in the new specification, OpenAI still expressly prohibits the use of its language models for any purpose that may cause harm, and warns against using its services to develop or operate weapons. This update may mean that OpenAI has made some fine-tuning of the scope of its technology, but still emphasizes the importance of ethics and safety.
Zhipu AI launches GLM-4 All Tools and GLMs
Zhipu AI recently launched GLM-4 All Tools and GLMs. GLM-4 All Tools is an innovative technology that automatically understands and processes complex instructions based on the user's intent. By automatically calling multiple tools, such as Vincentian diagrams, code interpreters, web browsing, and Function Call, GLM-4 All Tools can efficiently complete various complex tasks. The launch of this technology means that the comprehensive application capabilities of the GLM series models have been industrialized. Developers and users can now more easily utilize the GLM-4 model without worrying about prompt words. This innovation will provide users with a more convenient and efficient experience.
Alibaba Cloud sued the copycat Tongyi Qianwen App and won the first instance case
According to IT House reports, there is the first successful judgment in the domestic large model anti-counterfeiting and rights protection case. Alibaba Cloud and Alibaba sued the publisher of the copycat Tongyi Qianwen App and won the first instance. Feiyou Technology Company was ordered to compensate for related economic losses and rights protection fees due to infringement of registered trademarks and false publicity, and issued an apology statement on its official website for 15 consecutive days.
Youdao’s self-developed RAG engine QAnything is open for download
NetEase Youdao announced that its self-developed knowledge base Q&A engine QAnything is officially open source. The main principle of QAnything is based on retrieval-enhanced generation (RAG), which supports cloud large model service calls and purely local deployment. Users can download it for free in the open source community Github. The system supports multiple document formats such as Word, PPT, Excel, PDF, pictures, etc., which can be directly imported for Q&A.
Lanmeih/Today’s topic
AI-generated pictures are eyeing the movie market, what do you think?
Let’s chat in the comment area~
The above is the detailed content of The application scope of artificial intelligence in Hollywood is further expanded|Blue Media GPT. For more information, please follow other related articles on the PHP Chinese website!