Four Exciting Examples of Modern AI from NVIDIA

A few months ago, we had the honor of having Vishal Dhupar, Managing Director of NVIDIA India, speak at an executive forum that we had in Bangalore. NVIDIA's GPU technology is used to power what they call "modern AI", so the audience was very keen on listening to Vishal's talk.

Vishal started by tracing the history of artificial intelligence (AI) and the basics of how it works. What he said was similar to what Dinakar Munagala, Co-Founder and CEO of ThinCi Inc, said during his CDNLive India 2017 keynote – since I have written a blog on that I won’t repeat it here.

What was really interesting was the examples of practical applications that Vishal spoke about.

The first example was healthcare. As we all know, healthcare is a trillion-dollar industry. Nvidia has partnered with the US Government and other organizations to develop an AI framework called CANDLE (Cancer Distributed Learning Environment). NVIDIA engineers and computational scientists will contribute to all elements of this framework by jointly developing an AI software platform optimized for the latest supercomputing infrastructure, with the goal of achieving 10X annual increases in productivity for cancer researchers. This work is today possible because of deep learning.

The second example was about applications of AI in smart cities. Take traffic, a topic that is very close to the hearts of all of us who live in Bangalore. The reason for chaotic traffic in India is simply because people don’t drive by the rules. But Vishal challenged us to imagine that there was a traffic system that would understand how you are breaking rules and penalize you before you reached the next traffic light. So rather than one small fine of Rs 100 if you are caught by the police breaking a rule, you suddenly feel the pinch of aggregated fines on a daily basis, whether a policeman actually stops your vehicle or not.

Another smart cities example – security screening at airports. At busy airports like in any of our major cities, what if one could be differentiated depending on one’s security track record? This could be done using a machine learning algorithm via camera which tells the security officer that this person has come through here often and probably doesn’t need to be checked as much as the others. If someone is coming for the first or second time, the officer could be given the information that they need to do a more stringent check.

The third example was one in retail. FMCG giant Procter and Gamble is using a deep learning application to help women choose the right skin products from their Olay range of skin products – it’s called the Olay Skin Advisor. Using the app, you take a selfie on your phone and answer some questions about yourself like your age, preferences, etc. The app uses this information and deep learning technology called VIS-ID to give you three outputs: first, your “skin age”; second, what are the areas on your face where you look most youthful and what are the areas that need some help from cosmetics; and finally, it recommends a personal product regimen for you. Vishal said that this app has helped P&G increase their repeat orders by 94%.

The fourth example was about touching human lives. Vishal said that 57 people in India own the same wealth the bottom 70% - approximately $160 Billion. He went on to talk about The Sustainability and Artificial Intelligence Lab at Stanford University, which has been doing a project to understand and predict poverty by using satellite images and AI. They superimpose satellite images taken during the day and night of a particular area, and use AI to correlate the images and understand which areas are better developed. The premise is that the areas that are brighter lit at night are usually better developed. Thanks to this technique, the project is able to develop what they call “fine grain maps of poverty”. Currently the Lab is doing this project in five countries in Africa, but it is cheap and easily scalable and so hopefully could be used in developing nations across the world. This kind of information will help aid agencies distribute funds more efficiently and also help policy makers enact and evaluate policies more effectively. Here’s a video on the project:

Vishal concluded by saying that machines are becoming equipped with intuition, they’re learning through data, without the interference of any programmer. He ended by encouraging the audience to take the first step into discovering the power and impact of deep learning, because we are just at the very beginning of understanding the power of AI in various aspects of our lives.

Here is a soundbite from Vishal about the three key takeways from his talk: