WiMi Hologram Cloud Inc (WIMI)

[StockpickerCAPC]

WiMi Hologram Cloud Is Training An AIHolo Robot to Generate 3D Holograms.

Source
https://finance.yahoo.com/news/wimi-hologram-cloud-training-aiholo-120000019.html

May 2, 2023

WiMi Hologram Cloud Inc. (NASDAQ: WIMI), a leading global Hologram Augmented Reality ("AR") Technology provider, today announced that it is training an AIHolo robot to generate 3D holograms for users using AI from neural networks, which can go through the instructions in the training prompts and provide detailed responses. It can automatically generate multiple holograms for users based on AI understanding upon request. The AIHolo robot uses OpenAI-Photo open-source image generation to intelligently identify local items, depth of field, depth, and phase of images and then uses a complementary difference algorithm to process 3D holographic image pixels with high precision.

In the initial stage, human feedback is used to train AIHolo intensively. A human-supervised strategy is applied to prepare an initial AIHolo model. Through AI training, the engineer gives instructions and provides practice subjects. On the back end, the training engineer can access the model to write programs that help AIHolo generate holograms with the cooperation of hologram engineers.

Optimization models must then be built to create reinforcement learning optimization logic by collecting and comparing at least two or more models ranked by quality. Thus, multiple AI training engineers would instruct AIHolo to collect different comparison data, sample numerous results, and then compare and rank them manually. Proximal Policy Optimization is then considered to fine-tune the model and repeated several times to continuously iterate on the optimized model.

AIHolo may also have many problems and limitations because AI is a computer program that requires a human to do the initial model training and design the training technology framework. For example, a user who gives untrained domain and expression logic may generate results different from the user envisioned. AIHolo also needs to avoid responding to sensitive and harmful commands or generating biased 3D holograms. This requires censorship to stop such unsafe content. Of course, it is also possible that some false positives and missed critical information may come out, which requires the following collection of user feedback to help AIHolo improve and optimize the system. Users may provide unclear instructions that make AIHolo ask further questions or guess the user's intention through its understanding, which does not get the best result. Therefore, subsequent user involvement is needed to train and optimize AIHolo together and iterate continuously.

The launch of AIHolo will likely change the entire 3D holographic mapping field, potentially changing some of the ways 3D holographic mapping engineers used to work and improving the efficiency of businesses. WiMi's AIHolo will provide convenience and a variety of options for users.