The self-evolutionary model is the way forward as Deep-Learning-AI technology is fast approaching its end of shelf-life due to the increasing complexity of solving problems with large data and its diminishing returns.
With curiosity and exploration being the key ingredients of human intelligence, it is imperative that robots and Artificial Intelligence (AI) imbibe the human style of thinking and analysing a problem on their own, using self-evolutionary algorithms.
Feeding a pre-determined set of answers for Artificial Intelligence and machine learning will incur greater performance costs as the increasingly complex problems often require deeper creativity and exploration.
Here is what Hutson writes in his research document:
“Biological evolution is also the only system to produce human intelligence, which is the ultimate dream of many AI researchers. Because of biology’s track record, Stanley and others have come to believe that if we want algorithms that can navigate the physical and social world as easily as we can — or better! — We need to imitate nature’s tactics.
“Instead of hard-coding the rules of reasoning or having computers learn to score highly on specific performance metrics, they argue, we must let a population of solutions blossom. Make them prioritize novelty or interestingness instead of the ability to walk or talk. They may discover an indirect path, a set of steppingstones, and wind up walking and talking better than if they’d sought those skills directly.”
Standard deep-learning models are often hindered with a pre-determined set of conditions and parameters which often limit the algorithm’s potential to explore out-of-the-box or become creative like humans.
Machine learning algorithms often try out the pre-defined set of iterations to arrive at the correct answer or solution. If a solution doesn’t exist, they exit the loop or abort the task.
Self-evolutionary algorithms could address these shortcomings by enabling AI to evolve on its own and combine those new parameters with deep learning. Thus, self-evolving models will often find new ways of solving a problem when existing solutions fail to deliver the desired result.
Consequently, such proactive ways of learning will push the AI to acquire human-level thinking and problem-solving abilities. In other words, AI can evolve and explore uncharted ways to solve a puzzle, unlike the traditional deep-learning models.
Here is a simple test example from Hutson’s post in Quanta which demonstrates the abilities of a self-evolving AI:
“In one test, they [Stanley and researcher Joel Lehman] placed virtual wheeled robots in a maze and evolved the algorithms controlling them, hoping one would find a path to the exit. They ran the evolution from scratch 40 times. A comparison program, in which robots were selected for how close (as the crow flies) they came to the exit, evolved a winning robot only 3 out of 40 times. Novelty search, which completely ignored how close each bot was to the exit, succeeded 39 times. It worked because the bots managed to avoid dead ends.”
Deep learning AI restarts from scratch when it gets stuck at any point in solving a problem. Besides, such methods use millions of training cycles to enable AI to accomplish a task successfully.
In contrast, the evolutionary algorithm-based hybrid model allows the AI to continue finding new paths or alternative means of solving a puzzle instead of exiting the task.
On the downside, the self-evolving model is both underexplored and expensive to implement as it consumes more processing power than a conventional deep-learning AI model.
Stanley and his team of researchers firmly believe that AI could also develop a conscious mind of its own to evolve the human characteristics of creativity and curiosity as in biological evolution.