MIT researchers develop a network that can identify objects in the dark

Artificial Intelligence has been making huge progress in recent years. There are robots doing human work, machines learning tasks they aren’t supposed to, computers doing things one could never imagine, and humans are applying all their energy to equip the non-living things with living intelligence. This technological advancement is a fascinating thing from afar but is extremely complex from up close.

It is, regardless, growing each day without a pause. Artificial Intelligence, if used wisely, can be of great help to humankind. It has already made a lot of noise in many regions. Recently an AI game bot was specially designed to play the game Alpha Go. It was taught to learn from its mistakes if it lost the game. The goal of the bot was simple, play and win, if not, play and learn.

The AI game bot was then asked to play against the best Alpha Go player in the world. Interestingly, it beat the human with ease and this made a lot of noise. This was a huge win for Artificial Intelligence as it had successfully beaten a human at something which holds a bit of significance. All the AI bots are goal oriented. The goal should be completed otherwise the bot should work in the same mathematical way that rewarded it in the past.

Massachusetts Institute of Technology, one of the best schools in the world for learning tech, designed an Artificial Intelligence technique that could see in the dark. To be more specific, the technique revolves around spotting objects in a near dark environment. The deep neural network used has been trained in such a way that it can detect solid objects in the dark.

The neural network was fed around 10000 blurred, grainy or near to dark images purposefully to achieve this feat. The network understood the patterns in all those images and has now learned successfully to read objects in near blind condition. The network does this by producing ripples in what is supposed to be extremely dark for us.

The blurring images, however, was a problem for the researchers. They solved it by giving the network a physics lesson so it could understand the out of focus cameras and how the pictures are clicked. The result of this research is very significant. It can produce lightened images out of pitch dark rooms and help the humans in many ways.

The most basic use of this network can be done for night photography where it could be a boon for all the photographers. However, the MIT based group of researchers have a different aim in their mind. They want to use this technology to help the cause of medicines. Sometimes to capture the smallest details in biology, one has to use lights and radiation. This light and radiation might cause damage to the tissues.

This is where the MIT developed AI network can come in handy. It can develop images of the tissues without harming them as it does not need any radiations to do so. The faint objects can be then detected and the purpose of the network can be fulfilled. Apart from this, the AI network can help astronomy in clicking pictures and detecting objects in the dark.

This is surely for the future but the development of such a high level of AI is an achievement to applaud. Many such innovative inventions might help us develop this habitat to be in a better place than it is right now.

– Unmesh Phule

Comment here