.png){kind=logo}
- Insights
- Cryptocurrencies
- Stocks
- White Papers
- Industry
- Geography
- Insights
- Cryptocurrencies
- Stocks
- White Papers
- Industry
- Geography
It is not surprising that AI-enabled devices are fast invading our lives and we fear for human-like machines – empowered with a human-like brain – for they can overtake our lives? Though the fears remain genuine and valid, at least to normal people, it may not become a possibility even in near future. Artificial General Intelligence AGI is in the making and its ways of working should be far superior to that of AI. However, the question remains, is it possible at all to train AGI for human consciousness, whose abstract and complex mechanism of functioning scientists themselves could not decipher.
Whoever has come across intelligent voice assistants like Alexa and Cortana does not need an introduction to what artificial intelligence applications feel like. One significant advancement in Artificial Intelligence is that of neural networks. They make machine learning possible, making way for memory-based tasks, enabling improvising themselves every time the task is performed. Though empowered to correct itself, this category of AI whose ability is limited is considered weak AI. With AGI, machines can not only behave like a human but think like one. For making AGI models, researchers are depending on cognitive modelling that involves approximation of one or more cognitive processes in humans or animals for comprehension and prediction.
The role of consciousness in developing strong AI or AGI is thoroughly debatable because human consciousness consists of many dimensions such as abstract reasoning, the composition of elements, creativity, empathy, and perception which are only a few among many in the endless list. Though developing AGI sounds impossible considering the breadth and extent of variability a human mind has, according to Igor Aleksander, emeritus professor of Neural Systems Engineering in the Department of Electrical and Electronic Engineering at Imperial College London, the principles for creating a conscious machine already exist and applying it for further research is vital to the development of AGI. However, it may take some decades for such consciousness to become a reality.
Human mind can think beyond the circumstances it is trained, i.e., it can make decisions when provided with circumstances beyond its state of consciousness. The famous MITs experiment, 'ObjectNet' is an example. In the experiment, the algorithms are trained to identify objects which are set in a previously unseen environment, and 40 to 45% of state-of-the-art algorithms couldn't perform. This implies they could only extrapolate inside of the distribution, which clearly explains their lack of imagination capability.
Humans are natural meta-learners, i.e., we learn how to learn. By virtue of genes, certain attributes from our ancestors are ingrained in our neurons, nature's way of training the human mind. The performance improves with experience and a number of tasks without any external stimuli when given with an entirely unpredicted circumstance. Meta-learning algorithms are limited by their inability to act outside the confines of the datasets, which leads to the problem of compositionality. Humans are champions of this attribute which makes them easy communicators with context-specific reasoning. However, achieving this quality with algorithms will prove to be difficult because algorithms only respond to situations that have a high probability, making them insufficient for improbable situations.
Going by the current research and the outcomes thereby, there is no point in assuming that AGI can replace the human mind. At the most, we can achieve the peak performance in situational awareness among machines, for, the path AI can take to reach human consciousness is not yet invented.
