Research reveals that dumb questions posed by an AI system can actually make it smarter

The world is changing fast and a major reason behind this change is artificial intelligence and machine learning. Some argue that AI and ML have more drawbacks than benefits, but it is quite evident that these technologies bring with them a suite of other technologies that have definitely made our lives far easier. These technologies have opened up a myriad of other opportunities for business leaders to automate tasks and facilitate accurate outcomes to serve clients better. Research and innovation are quite constant in the AI ecosystem. Currently, scientists all over the world are experimenting to yield smarter uses of the technology. Some believe making AI smarter will only kill us, whereas, others counterargue and claim that innovating more intelligent AI systems will only help humanity to become futuristic and avant-garde.

A couple of days ago, an endeavor to make smarter AI systems just went viral on the internet, and apparently, Instagram is aiding this procedure greatly! The new research denoted that posing dumb questions to AI can actually increase its accuracy and efficiency, besides, it can also improve the exactness of the art pieces or images generated by artificial intelligence systems. Over the years, researchers have been training artificially intelligent machines to improve their accuracy, design faster programs, and diagnose illnesses and unknown viruses faster, but who knew the process would eventually become so easy and cool?

How does the new AI system work?

Scientists from Stanford University, led by Ranjay Krishna from the University of Washington, Seattle, conducted a study in which they trained a machine learning model to detect gaps in the knowledge and information it possesses, and then formulate dumb or illogical questions about images or objects which strangers would answer, or in this case Instagrammers.

The researchers made the AI system react like a kid, to avoid getting negative feedback from people. The system asked questions like about the shape of the sink, or when a picture was taken, night or day. The AI posted its queries on Instagram, and when people replied to its questions, the researchers got feedback to coordinate its algorithms in a similar way for future conversations with it. Evidently, the ML model could conduct itself properly and accept human social norms and language by improving its ability to question others, which was more relatable and understandable. Quite unlike several other new AI systems which are recently developed and are accused of being racist, sexist, or even homophobic!

Eventually, after finishing the research, the team reported to the National Academy of Sciences that over the course of 8 months, the system asked over 200,000 questions to Instagrammers, which eventually enhanced the accuracy of the AI at answering questions, becoming more than 115% similar to the humans who were answering the system's questions.

Bottom Line

The experiment conveyed that social skills are inherently important for emerging AI systems, besides, the increasing importance of chatbots in global industries might actually receive aid from this experiment as it would improve the accuracy of the responses that are generated by those systems. Not just for global businesses, researchers believe that AI's understanding will eventually rise, enabling even robots to converse with humans fluently about various societal issues and can itself look for improvements in its system, in the years to come.