Understand the Formation of Structures and the Amount of Matter in the Universe

AI will help humans land on Mars, develop vaccines to save humans, and realize human's wish to explore the universe. The latest "I am AI" video broadcast in NVIDIA CEO Huang Renxun's GTC keynote speech introduces the historic progress made by scientists, researchers and artists in the AI field.

How influential is AI in the world? According to Gartner's prediction, this technology will create $2.9 trillion in business value by 2021.

Now, in order to understand the origin of the universe, scientists are classifying 2 trillion galaxies and exploring the molecular structure of the new crown virus and cancer treatment drugs.

As described in this latest video, AI also has an "artistic side". It can not only draw like Bob Ross, but also help the London Symphony Orchestra write and play music (the orchestra plays a theme accompaniment music composed by cyclic neural network).

AI can also perform text to speech synthesis and dub short documentaries, and AI has been used to do this.

These contents are covered in the "I am AI" video. In addition, 16 companies and research institutions appear in this video.

To outer space

In order to understand the formation of structure and the amount of matter in the universe, we must observe galaxies and other celestial bodies and classify them. It is estimated that there are 2 trillion galaxies in the universe that can be observed by humans. Cosmologists call this task the "computational challenge".

The recent dark energy survey collected data on more than 300 million galaxies. In order to conduct unprecedented high-precision research on them, the AI innovation center of the University of Illinois at Urbana Champaign National Supercomputing Application Center cooperated with the Argonne Advanced Computing Center of the Argonne National Laboratory of the U.S. Department of energy.

NCSA has launched a multi-source astronomical research project - Galaxy Zoo. The project will mark millions of galaxies observed by Sloan Digital Sky Survey. Now, an AI model with an accuracy of 99.6% can use these data to carefully examine unlabeled galaxies, identify them and speed up scientific research.

The National Aeronautics and Space Administration (NASA) of the United States has set the goal of human exploration of Mars in the mid-1930s. To accomplish this feat safely, scientists must understand the impact of the sun and space weather on the journey. High resolution images of the sun taken by NASA's Solar Dynamics Observatory will contribute to this work.

NASA Goddard Space Flight Center researchers have developed algorithms to analyze a large number of solar images taken since 2010 using NVIDIA GPU. They are using errors in the images to help identify sources of high-energy particles that may damage the earth orbit of the interplanetary spacecraft. In addition, researchers also use these images to track the flow of the sun's surface and help develop better space weather prediction models.

Repair sounds and limbs

Voiceitt is a Tel Aviv based start-up focused on developing signal processing, speech recognition technology and deep neural network, which can provide synthetic speech for people with speech distortion. The application developed by the company can make difficult speech easy to understand.

The neuromuscular rehabilitation engineering laboratory at Chapel Hill, University of North Carolina, and the active robot sensing (Aros) Laboratory of North Carolina State University have developed experimental robot limbs used in the laboratory.

These two research institutions have long been committed to the research of walking environment recognition, hoping to develop environmental adaptive control technology for prosthetics. They have been using CNN to run forecasts on NVIDIA GPU, and other agencies are doing the same.

Help fight the epidemic

Whiteboard Coordinator remotely detects the temperature of personnel entering the building, thereby minimizing the contact with COVID-19. The Chicago based startup enables checkpoints to measure temperatures for more than 2000 people per hour. Whiteboard coordinator and NVIDIA apply AI to the medical field through NVIDIA Clara guardian. This application framework simplifies the development and deployment of intelligent sensors.

Compared with traditional methods, viz.ai can more quickly inform neurologists of a patient's stroke through AI. With the outbreak of the disease, Viz.ai has been able to help people fight against the new crown virus through an application that alerts the nursing team to the positive results of the new crown virus.

Axial3d is a start-up company located in Belfast, Northern Ireland. With AI technology, the company can reduce the production time of 3D printing models of medical images. The company has now concentrated its resources on the work of COVID-19, not only providing masks, but also producing ventilators for the National Health Council. The company also began using 3D printing to manufacture cotton swab bags and respirator valves for testing.

Kiwibot, a food delivery robot in Berkeley, California, has a pair of cute big eyes. Now, in response to the COVID-19 epidemic, the robot provides a new delivery service: providing respirators, disinfectants and other supplies for people.

All of the above companies are members of NVIDIA inception.

Art, music and dubbing works

Researchers from oxia palus, a start-up company in London, showed how to use AI to reproduce lost works of art in the paper Raiders of the lost art. Many art experts believe that the mountain in the background of Picasso's 1902 work the crouching begga is located in the Horta labyrinth garden near Barcelona.

They also know that Santiago RUSI ñ ol painted Parc del Laberint d'horta. The researchers improved the X-ray fluorescence images of the crouching begga and Santiago RUSI ñ ol's terraced garden in Mallorca, and then reconstructed the lost works of art through neural style conversion on NVIDIA GPU to reproduce RUSI ñ ol's Laberint d'horta.

AIVA AI in Luxembourg wrote the opening melody and accompaniment for the GTC conference a few years ago. Later, this music became a classic original classical music, and an orchestra expressed its willingness to play it.

In the second half of last year, the London Symphony Orchestra agreed to play this moving music. The performance was arranged by musician John paesano and recorded through Abbey Road studio.

Helen, a former NVIDIA employee, has been providing us with professional video and event dubbing for many years. When she left the company, we began to think about how to continue this tradition. Finally, we decided to use AI we were familiar with, but there was no publicly available model to complete this task.

A team of NVIDIA applied deep learning research group released a solution to this problem, flowtron, which is an autoregressive streaming generation network for text speech synthesis. After being authorized by Helen sound, we spent dozens of hours training the network.