By Sachin Panicker
Over the past year, the emergence of large language models such as OpenAI’s ChatGPT and Google’s Bard has fueled renewed excitement across sectors, and among society at large, about the seemingly limitless potential of Artificial Intelligence (AI). Large and small organizations are rushing to explore the incorporation of AI solutions into their processes, with a PwC study predicting that the technology will add over USD 15 trillion to the global economy by 2030.
As the technology is poised to play an even greater role in the ongoing global wave of digital transformation, it would be relevant to contextualize its evolution through the prism of two major forms of AI – Artificial Narrow Intelligence (ANI) and Artificial General Intelligence (AGI).
Understanding Artificial Narrow Intelligence (ANI)
Artificial Narrow Intelligence (ANI) or ‘Weak AI’, is the form of AI that predominantly exists in our world today. As the name suggests, it is a highly specialized system designed with the narrow focus of performing a specific pre-defined task (or set of tasks) in a particular environment. It has a limited range of abilities and is incapable of adapting to a new task on its own without further training or programming.
Several digital systems we engage with regularly contain some form of ANI – from customer service chatbots that answer routine, repetitive queries, to natural language processing tools such as Google Assistant, Amazon’s Alexa or Apple’s Siri. While on the surface these applications appear to be highly sophisticated due to their ability to interact with humans and execute tasks instantaneously, they lack any sense of self-awareness or consciousness and aren’t remotely close to mimicking human intelligence.
For instance, a voice assistant can seamlessly respond to queries about current weather and traffic conditions or provide any other factual information that it can pull from a dataset that it has access to, but it cannot answer more complex and abstract questions, such as how to solve a personal problem. This is because the ANI that powers the assistant is incapable of thinking for itself and responding to queries or performing tasks beyond its defined parameters.
Towards Artificial General Intelligence (AGI)
While Artificial Narrow Intelligence defines much of our current AI paradigm, Artificial General Intelligence (AGI) or ‘Strong AI’ is more representative of the perception many have of the technology as mimicking, or even surpassing human intelligence. Put simply, AGI refers to a machine that is capable of performing any intellectual task that a human being can.
Human intelligence currently possesses an edge over Artificial intelligence due to our ability to think and strategize abstractly, and to rely upon our thoughts, memories and emotions to make informed decisions or to conceive creative ideas. These are the capabilities that would need to be replicated for the development of AGI – self-aware machines with human-like reasoning and problem-solving abilities, that are capable of being innovative and creative, and of taking decisions in ambiguous situations. AGI would be capable of acquiring new skills and knowledge much like a human being would, without any additional programming.
At present, a fully functional AGI is in the realm of speculative fiction and thought experiments. While the recent advancements in large language models – which serve as the basis for Generative AI systems that mimic human speech and answer queries on a wide range of topics – do represent a significant step forward for AI technology, we are still in the relatively early stages of the development of AGI.
A key challenge when it comes to building AI that can mimic human intelligence is the sheer complexity of the human brain, including aspects such as the workings of neural pathways, episodic and semantic memory etc. We still do not know all there is to know about the human brain, which significantly adds to the difficulty of replicating its functioning.
One approach to resolving this problem is to jointly leverage AI, NLP (Natural Language Processing) and Deep Learning to develop cognitive abilities in machines, such as planning and reasoning, learning from experience, and discerning emotions to better comprehend human behavior. The development of artificial neural networks modelled after the structures and functions of the brain is another popular approach. Hybrid approaches, that combine neural networks with rule-based systems, are also being actively explored by researchers working on AGI development.
Anticipating the Potential Risks of AGI Development
Even as the world inches towards the emergence of AGI, it is important to not lose sight of the very real risks that might accompany the development of a self-aware human-like machine. Several ethical issues have already arisen with existing AI systems, such as their potential to amplify existing human biases present within the datasets used to train them, or the ability to create and spread misinformation using Generative AI tools. These challenges could be exacerbated exponentially with human-like AGI systems, in addition to innumerable new challenges that could appear. Some theorists and commentators have even warned that AGI if left unchecked, could spell the end of human civilization.
While we are still some way off from such a bleak eventuality becoming even a credible possibility, valid concerns about the risks of the technology do reinforce the critical need for guardrails while developing AGI. It is incumbent on researchers and developers to ensure that AGI is in alignment with human values and morals. National and international frameworks to govern the development and use of AGI will go a long way towards establishing ethical behavior guidelines and safety standards.
The author is chief AI scientist, Fulcrum Digital
