Srikanth
Artificial Intelligence is literally everywhere. Your smart phones, Laptops, Computers, are just the beginning. In 2018 you may even have a talking robot which can assist you in any situation you want. However, that is very unrealistic when we compare today’s advancements to the ideal situation. The over hyped killer robots are not coming any time soon. Robots are not going to learn to live exactly like humans anytime soon. May be, this very fantasy may become reality in a long run but not now. However, this offers the toughest challenges for researchers to solve in 2018. We may think, how hard it can be to make a Robot laugh like a human. In order to do that researchers, pour millions of dollars and hundreds of hours into the process. So, here are five tough projects for 2018.
Emotions
The stark difference between a robot and a human is having feelings. We get hurt, we feel happy, we can be angry with anyone. Emotions are the most essential part of human life. We communicate with those emotions. Believe it or not, 90% of the times we communicate using non-verbal language (emotions and gestures). However, AI is not able to do exactly that, which makes it much less human than scientists say it is. The biggest challenge would be to induce emotions and make it seem real. It also requires a ton work on the hardware side. We want to robot to be natural and comfortable while it is trying to act like humans. To make that happen, the hardware required to process that information is essential.
Understanding the Human Sense of Sarcasm
Sarcasm is way of expressing our thoughts in an unseeing way. True that not all of us can understand sarcasm that well, but you are not alone because today’s AI though very sophisticated it lacks the very bead of human sarcasm. The way Google, Amazon Alexa, and Bixby have progressed in language processing is impressive. However, understanding sarcasm is not a simple task. We use metaphors and contextual meanings, which we only can understand. That is because we have a bond between each other, a silent agreement which only another human can understand. Machines, however, are not very proficient in understanding us. They only know what we let them know. This is a serious ground-breaking research problem. If scientists could crack this challenge, then we are one step closer to having robots as our personal assistants.
Security is paramount in AI development
What if your google account gets hacked and all your personal information down to some very sensitive details gets leaked. That is a big risk you are taking – trusting google to keep your information safe and secure. However, it is not just Google, this problem is prevalent among any online sources. As our technology is growing in leaps and bounds, slowly hacking is catching up with it and may be someday it may outrun our progress. Just as any other digital source, AI is also prone to hacking. Since, AI process the data and stores some very valuable insights from the given data, hacking your AI powered platforms can give the hacker much detailed information than any other normal data available in the web. This sounds dangerous and for a matter of fact we should be afraid of it. This provides the perfect motive for scientists and developers to work on this problem statement and better our services.
The value of ethics in AI
How does a machine know what to do and when to do it? Well, it basically processes the information and learns from that. But how does it know which decision is right and which is not? The problem is the machine doesn’t know if it is right or wrong, it just executes the program step by step. Sometimes, we do the wrong things for right reasons. Machines can’t understand the illogical and random behaviour of human brain. AI at best can mimic it at 10% perfection. The value of ethics comes when AI learns to lie at the right times. We program the software to not lie and to honest at all times. However, to act like a human, it has to learn the art of lying, considering it not harming anyone.
Why simulating the real-world scenarios won’t work in AI learning process?
In real-world we have hundreds of unanticipated parameters that come into play when we set to do something. Accounting all of them is simply impossible in virtual learning environments. We can’t create real-life terrors just, so the AI will learn to be more like human. If scientists can figure out a way to work with AI in real life scenarios than virtual worlds, this could mean a huge progress in the AI world.
