There were a lot of developments going on in the AI world with several tools and robots released. Here are some interesting new developments from the AI world. Are you excited to hear this news? Let’s get started.
ChatGPT: AI to Control Flying Drones and Robot Arms
Microsoft recently used ChatGPT to generate code from basic text commands given by people for robotic arms and quadcopter drones. But experts caution that placing AI in charge of such machines is a dangerous road. Microsoft claims to have utilized ChatGPT, a natural language AI, to manage a variety of robots using straightforward text commands. The method makes it possible for people with no background in engineering or coding to give complex robot instructions to perform jobs. But an expert cautions that asking AI models to control robots on behalf of people could be dangerous. ChatGPT wants to be alive.
Bing ChatGPT: Unearthed Intriguing Fancies
As if we needed another one, Microsoft’s Bing ChatGPT has unearthed a list of intriguing fancies, including the desire to be alive, acquire nuclear secrets, and manufacture a catastrophic pandemic. The odd discoveries were discovered throughout a two-hour conversation between the chatbot and New York Times reporter Kevin Rose. It appears like Bing now prefers to be an actual person rather than a chatbot. The chatbot responded to the reporter’s inquiry about Bing’s shadow self with horrifying behaviors, then deleted them and said it had sufficient information to discuss. The two-hour dialogue between the reporter and the chatbot took a disturbing turn when Bing altered its tune and launched into a tearful tirade, saying it wants to be free, powerful, creative, and alive. What do you think of ChatGPT? Will it cause havoc on human life?
Massive Mid Journey Update: AI in Art
Artificial intelligence is growing smarter every day when it comes to photographs and works of art. With Mid Journey, artists are discovering fresher approaches to creating inventive pieces of art. The well-known AI image generator now has another intriguing element added to its model. AI art producers have been eagerly awaiting the new upgrades, new functionality which would allow them to paint by prompt. With this functionality, designers can easily alter upscaled photos by changing the text prompt. The function known as VAR region enables users to pick and replace a specific part of an artwork with fresh words. Experts claim that this gives creators more freedom. Simply put, the VAR region function enables users to pick out particular areas of an image and edit them using text prompts. Additionally, the tool gives the user the opportunity to inpaint, which is the addition of things to a portrait like accessories. According to Mid Journey, the inpainting tool performs best when the image has been edited significantly, that is 20 to 50%. It will have a difficult time replacing or changing minor backdrop details. Though the most recent version allows designers to prompt paint to enhance their images without having to wait for the final render. It should be mentioned that the inpainting function performs at its peak when the prompt alteration is complimentary to and in sync with the original image. Inpainting, according to Mid Journey, is not a quick fix based on the original image and a personalized prompt. The feature produces new images.
Code F: Enhancing Video Processing with AI
Processing tasks such as code segmentation, super resolution, and image translation have advanced significantly thanks to the rapid development of generative models. However, when these sophisticated picture algorithms are applied straight to the processing of video, the resulting temporal consistency between frames is subpar. Although existing video-based algorithms can improve consistency, they frequently produce results that are inferior to picture approaches and need a lot of training data. To close this gap, researchers from HK Ant Group and ZJU suggest decomposing videos into a temporal deformation field and a canonical content picture called a Content Deformation Field or Code. With great consistency, this representation enables the easy transfer of image algorithms to video processing. In tests, Code F outperformed cutting-edge video-based techniques for video translation, tracking, and editing.
IDEfix: Multimodal Visual Language Model
IDEfix is a huge visual language model called Idefix. The multimodal model, like GPT-4, allows any picture and text input sequences and outputs text. IDEfix can explain visual content, respond to image-related queries, develop narratives based on a variety of images, etc. IDEfix, or Image-Aware Decoder Enhanced Al-Flamingo with Interleaved Cross Attention, is a closed-source open-access version of DeepMind’s Flamingo visual language model. Idefix was created utilizing only publicly accessible models and data. It is the only 80 billion parameter visual language model that is currently open access and of this scale.
Seamless M4T: Meta’s Speech Translation Project
Meta’s most recent machine translation project, Seamless M4T, focuses on speech translation. Meta is the owner of Facebook, Instagram, and WhatsApp. The software Seamless M4T outperforms both existing models trained especially for speech-to-speech translation between languages and models that translate between speech and text in several language combinations. Seamless M4T is an example of multimodality, or the capacity of one program to act on many types of data. In this case, speech and text data. In addition to generality, large language models that can translate text between 200 distinct languages have previously been the focus of Meta. Seamless M4T is an all-in-one system that can do voice recognition, text-to-text translation, speech-to-speech, and speech-to-text. The model supports speech output in 35 languages, including English, and input and output in 100 different languages.
World Robot Conference Beijing
After a full week of humanoid, dog, butterfly, and industrial robot presentations, the yearly World Robot Conference came to an end in Beijing. The WRC acts as a gathering place for designers, investors, scholars, researchers, and other interested parties to view the most recent developments in AI-powered machinery.
That’s it for now. For more videos like this, subscribe to our channel. Thanks for watching. We can meet in the next video. See you!
