Artificial Intelligence (AI), Machine Learning (ML) and Deep Learning are some of the buzzwords swirling around today. With leading tech companies such as Apple, Amazon, Facebook and Google among others investing heavily in these areas, they’re turning mainstream. You’e more likely to hear about AI and ML when tech companies talk about voice assistants and smart home devices. Now, while Artificial Intelligence and Machine Learning are very much related, they are not the same thing. Let s dive in a little deeper to understand what these terms mean. Also Read - Beware! This new iOS bug breaks WiFi on iPhones: Here's a quick fix for it
Artificial Intelligence the history, applications and moreAlso Read - Happy Father's Day 2021: Google Doodle wishes dads with cute pop-up greeting card
While tech giants have started talking about AI more recently, it is something that existed decades ago, and you probably didn t even realize it back then. Remember Arnold Schwarzenegger s movie The Terminator that was released in 1984? The story was based on machines taking over the world, directed by artificial intelligence Skynet. Star Wars had the R2-D2 and C-3PO robots powered by AI. And if you are a fan of Iron Man, and his personal assistant Jarvis, that’s AI as well. Also Read - Google might be working on a 'Find My' network clone for Android users
BGR India spoke to Pradeep Dubey, Director, Parallel Computing Lab at Intel to understand more about AI, ML and Deep Learning. And to put in simple words, AI is a vision of human intelligence exhibited by machines. The ultimate vision of AI is to be indistinguishable from humans. When given a task to recognize songs or images, you should not be able tell from the response whether it is from machine or human, Dubey said.
Companies like Intel, Qualcomm and Huawei are also using AI in their chipsets to carry out a range of different tasks. Having on-board AI can be beneficial in many ways, like enabling faster voice and image recognition, intelligent photography and more. The Kirin 970 SoC comes with AI, and in an image recognition benchmarking test, it processed 2,000 images per minute, faster than other chipsets, Huawei said.
The AI can also be used for object recognition, or to enhance picture quality. Google is using AI in its Pixel 2 and Pixel 2 XL smartphones which detects the scene and makes adjustments to the photos accordingly. While other manufacturers are using dual-camera setups to add DSLR-like Bokeh effects, Google is using AI to determine the foreground and background, and add depth-of-field effects accordingly, and it does a pretty good job as well. Even Oppo F5 comes with AI to help you look good in your selfies.
Intel, on the other hand, has its own self-learning Loihi chipset that is designed to mimic the human brain. In an example, Intel said that the chipset can use image recognition applications to analyze streetlight camera images and solve abducted or missing person reports. It can even automatically adjust spotlight timings to sync the traffic flow.
Machine Learning Data is key for AI s success
Machine learning, in its very basic, is an approach towards achieving artificial intelligence. Any program that gets better over time is called machine learning, Dubey said. The program can be for anything, like executing some complex calculations. It is not necessarily tied to AI because AI is about becoming like human, whereas machine learning is simply a technique that uses program or algorithm to solve a problem in no time.
Data is key for machine learning. Take an example of a school teacher he can do a number of multiplications on a drawing board so you can learn and figure out how to get the answer. This is data, and you learn from examples on your own exactly how machine learning works. However, if the teacher does not solve those many examples (data), you will have to rely on him to teach you, and you won t need those examples.
In the past, we did not have enough data to solve a problem, and all we had was the input / output data only. What did AI do in such times? It used to rely on an expert to tell it how to win a game or do some complex calculations. Say some person is good at playing chess, the AI needed him to teach the magic tricks to win. However, there is a problem with this approach.
Take an example of a singer, even if you find the best singer, he / she cannot tell you how to sing. Another example could be of a cartoonist, who can look at you and in five simple brush strokes draw your cartoon. But he cannot tell you which five things make up your face. In short, experts cannot always tell you how they did it, Dubey said. Because of this, the scope of AI was limited in the past, but with machine learning and data, things have changed.
Today, machines have the amount of data to learn from. For instance, say some earthquake takes place in a random village in India, and Google has no clue about that place. Soon, when people start searching for it, Google starts learning about the place, and collects data from the web (Wikipedia, or any other source). And as millions of people are inputting a query, the data center quickly learns about the place from it.
There is one more example to simplify the working of machine learning process. Say, you like a Cheese Burst Pizza from Domino s and you tweet about it, the machine learns about your habits. It will also look at other data on the internet where other users may have tweeted it, or blogged about it, so it can know about different places. It also personalizes the responses for you. Say, you go to London where majority of places serve non-veg food, but the machine knows you are a vegetarian, it will look up for those veg places and recommend you the results. This is that power of machine learning that we don t have.
Deep Learning extending applications of machine learning
Deep learning a machine learning technique which involves feeding a lot of data into computer system so it can make decisions on its own. It is one of the technologies used in driverless cars that enables them to distinguish between a lamppost and a pedestrian, or a stop sign. In deep learning, a computer learns to perform different tasks directly from sound, images or text.
The technique is used by Google in its image and voice recognition algorithms. Say, for instance, the computer is fed with images of Taj Mahal, Gateway of India, Leaning Tower of Pisa, and more. Now, let s say you visit Taj Mahal, and you don t know what place it is, you simply click the photo using Google Lens, and it will tell you what place it is. How does it do that? It compares the image you clicked with its database and gives you the results.
Deep learning is also used by Netflix and Amazon to decide which movie or TV series to recommend you next. Say, for instance, you watch a lot of sitcom or romcom movies, the next time you login, it will recommend you the content from that genre. The technique can also help in better preventive healthcare. For instance, a wearable can take your heart rate reading from time to time, and send the log to the doctor. The moment something alarming is noticed, with irregularities in your heart rate, the doctor can ask you to come down to the clinic and get the required tests done.
To sum up, AI is the present and future, and with the help of machine learning and deep learning techniques, it can get to the science fiction state we ve been fantasizing for years. I ll keep Jarvis, whereas you can have your Terminator or C-3PO or R2D2.