Can an AI algorithm become smarter than its creator?
Answering this question gives goosebumps to many. When we think about it, then the image of robots taking over the world comes to mind.
To evaluate an intelligent answer, we need to assess the broader or narrower meaning of smart, which is rightly defined by J. P. Guilford. The psychologist has divided thinking into two aspects: divergent thinking and convergent thinking. The convergent thinking means answering given questions correctly, which displays logic and memory. However, the divergent memory is forming various answers for one problem statement, displaying the presence of curiosity – the ability of extraordinary thinking.
Most of the people, who talk about how AI can make machines smarter, are talking about convergent intelligence. The extreme processing capabilities, high memory power, and ability to perform complex calculations are the factors that give computers an edge in the business (if we are talking about repetitive and monotonous tasks). This aspect which is not present in the AI machines is curiosity and ability to imagine. Hence, divergence is missing.
Can AI Powered Machines Utilize Divergent Thinking?
When computer systems or AI-powered machines are fed with millions of datasets, these machines produce thousands of simulations that meet specific requirements. Isn’t this activity a divergent task?
For instance, an organization Autodesk wanted to develop a new office. Hence, they turned to their employees and asked them to define the workplace they would like to work in. After gathering all this data, the company fed this data to the AI-powered machine which gave around 10,000 simulations. The architects utilized the required aspects of many blueprints.
This new model is also known as generative design.
Use of Generative Design in Pharmaceuticals
The model is said to be extremely useful in the pharmaceutical industry. Currently, various assessments are achieved on animal and human volunteers before the release of a drug in the market. With AI-powered simulations, the negative impact of a certain drug on the human body can be significantly reduced. The knowledge of divergent tasks can be used to produce a biological simulator to assess various outcomes related to the administration of drug compounds.
Various Simulations of Artificial Intelligence
In one of the AI simulation developed by a few scientists, the results were rather disturbing. Usually, the AI algorithm is fed with data and a goal is used to assess the information without human interference or instructions. In one such model, the AI-powered machine algorithm was to use minimal force to land the virtual airplane. But, after a while, the AI algorithm found out that a crash can overwhelm its own memory with a larger force. This enabled the algorithm to crash the virtual airplane multiple times, which means if this was a real instance, it would have killed thousands of people. This type of alarming results makes scientists dread the use of AI in daily functioning.<>
However, contrary to this, an AI-powered algorithm was designed to move in the forward direction as soon as possible. The AI algorithm took this task so literally that instead of making legs to move forward, it created a tall building. This building would fall forward quickly to reach the destination.
Although in the second instance the machine failed, it still completed its tasks of reaching the destination at earliest.
Undoubtedly, AI-powered machines have the capability of exhibiting divergent thinking. However, it is necessary to note that machines are still machines. They can’t turn on themselves or show emotions or ask self-arising questions or even get self-motivated. Hence, it is unlikely for the machines to become smarter than humans with respect to overall working power of humans. The machines evolve not as a human but as an organism.