[December 18 Big Model Daily Collection] Byte developed its own big model, but was banned for using ChatGPT and caused controversy? The official responded; Gemini revealed that he used Baidu Wenxinyiyan to train Chinese, and netizens were dumbfounded: Are big companies poaching each other? ;Use ChatGPT to help 5 million users file taxes! One of the world’s largest tax agencies cooperates with Microsoft; Samsung and LG will launch laptops with AI performance early next year to join the AI PC wave

NeurIPS 2023｜Effectively improve the consistency of video editing, Meitu & National University of Science and Technology proposed a new method EI² based on the Vincentian graph model

Link: https://news.miracleplus.com/share_link/13562

Meitu Imaging Research Institute (MT Lab) and the University of Chinese Academy of Sciences have proposed a new video generation method EI(2) based on the Vincentian graph model, which is used to improve the consistency of semantics and content in the video editing process. This paper analyzes and demonstrates from a theoretical perspective the inconsistency problem that occurs in the video editing process, which is mainly caused by the covariate shift in the feature space caused by the introduced temporal information learning module. A new network module is designed to solve the problem to generate High quality editing results. Currently, the paper has been accepted by NeurIPS 2023, one of the top machine learning conferences.

A Chinese team uses a large model to achieve “mind reading”: brain activity is directly converted into text | NeurIPS 2023

Link: https://news.miracleplus.com/share_link/13563

A new study included in NeurIPS allows large models to learn to “read minds”! By learning from brainwave data, the model successfully translated the subject’s EEG signals into text. And the whole process does not require large equipment, just a special “turban” can be achieved. The result, called DeWave, can interpret brain waves and translate them into text without using invasive equipment or MRI. Because it uses a large model to read the brain, iFLScience, which reports on DeWave, also calls it BrainGPT.

Byte self-developed a large model, but was banned for using ChatGPT and caused controversy? Official response

Link: https://news.miracleplus.com/share_link/13564

According to The Verge, the large-scale language model project being developed internally by ByteDance is called “Project Seed.” Because training large models requires a lot of question-and-answer knowledge, the project was revealed to have been secretly using OpenAI’s technology to enrich the data set. In the field of large models, the “trick” behavior of using content generated by other AIs for training is not uncommon, but it is often considered to be an act that exceeds the bottom line. On ChatGPT, misuse of AI-generated data directly violates OpenAI’s terms of service, which stipulate that its model output cannot be used “to develop any artificial intelligence models that compete with our products and services.”

The construction of ChatGPT is inseparable from PyTorch. LeCun’s remarks caused heated discussions. This is why the model manufacturers do not open the weights.

Link: https://news.miracleplus.com/share_link/13565

In the past two days, the topic of open source has become popular again. Some people said, “Without open source, AI will have nothing. Continue to keep AI open.” This view is agreed by many people, including Turing Award winner and Meta chief scientist Yann LeCun. Imagine what the AI industry would look like today if industrial AI research labs remained closed, had no open source code, and applied for and enforced patents on everything. Imagine a world without PyTorch, and with Transformers, ResNet, Mask-RCNN, FPN, SAM, DINO, seq2seq, wav2vec, memory-augmented networks, Transformers, BatchNorm, LayerNorm, Adam, denoising autoencoders, joint embedding architectures, and tons more If SSL methods are patented, what will happen to the AI industry?

Gemini revealed that she used Baidu Wenxinyiyan to train Chinese, and netizens were dumbfounded: Are big companies poaching each other?

Link: https://news.miracleplus.com/share_link/13566

When Google’s Vertex AI platform used the model for Chinese conversations, Gemini-Pro directly stated that it was a large Baidu language model. Soon, a famous Weibo V @阑西夜 also posted a blog saying: A test was conducted on Gemini-Pro on the Poe platform. Ask it “Who are you?” Gemini-Pro comes up and answers: I am the Baidu Wenxin model.

Use ChatGPT to help 5 million users file taxes! One of the world’s largest tax agencies partners with Microsoft

Link: https://news.miracleplus.com/share_link/13567

H&R Block, one of the world’s largest tax service organizations, announced on its official website that it has officially released its generative AI tax filing assistant – AI Tax Assist. It is reported that this is a ChatGPT-like assistant powered by Microsoft Azure OpenAI. Users can quickly complete complex business processes such as tax filing, tax refund, tax management/inquiry, etc. through text Q&A. Its accuracy is as high as 100% because H&R Block uses nearly 70 years of its own data for pre-training and fine-tuning. At present, this product has been officially launched to provide services to more than 5 million users of H&R Block. It charges US$35 (approximately 250 yuan) per month and supports 7*24 hours of uninterrupted service.

Tencent Cloud launches high-performance application service HAI: it claims to develop exclusive AI applications in 10 minutes

Link: https://news.miracleplus.com/share_link/13568

The official “Tencent Cloud” official account has just announced that Tencent Cloud has launched the high-performance application service “HAI” (Hyper Application Inventor), which can realize out-of-box and one-click deployment of GPU computing power. It is known as “Develop your application in 10 minutes” Exclusive Al application”.

Samsung and LG will launch laptops with AI performance early next year, joining the AI PC wave

Link: https://news.miracleplus.com/share_link/13569

Samsung Electronics and LG Electronics will both launch laptops with artificial intelligence (AI) capabilities next year, joining the industry in integrating AI into smartphones and other devices. Samsung Electronics’ Galaxy Book 4 series of laptops will be launched early next year. There are three models: Ultra, Pro and Pro 360. They are equipped with Intel’s first Core Ultra 9 chip that integrates a central processing unit, a graphics processor and a neural processor.