Artificial general intelligence (AGI) refers to systems that has achieved singularity with humans and, as such, can comprehend or learn any complex tasks that humans can. AGI systems would be able to carry out a variety of tasks like problem-solving, making decisions and learning even without being explicitly programmed. Most AI systems in use today fall into the category of ‘weak’ or ‘limited AI’ due to their focus on narrow task domains.
The advancement of AGI has ethical and societal implications, as it may alter our interactions with technology and our understanding of intelligence. For example, in the 1980s, Prof. Weizenbaum created an AI-based system called ‘ELIZA’, which mimics a Rogerian therapist and was one of the first natural language processing programmes. His system showed the shallowness of human-computer interaction and the risks associated with an overreliance on AI technologies.
He was astonished to discover that his secretary had an emotional connection to the system while using it – as if it were human – despite knowing that he wrote the code for it. Weizenbaum used this experience to openly criticise the field of AI after he realised that people might treat AI systems as if they were people. He argued that the hype surrounding the technology was unjustified and that it was unlikely that machines would ever fully understand human thoughts or emotions.
Weizenbaum’s viewpoint on AI raises the question of whether or not AGI is truly possible and whether or not the concept of AGI is well-defined. Also, it emphasises the importance of being realistic about what AI can or cannot do and avoiding overhyping the technology. It is important to note that the field of AI has evolved significantly since Weizenbaum’s criticism and some researchers have developed new methods, techniques and theories that address some of AI’s limitations, such as symbolic AI, connectionist AI and hybrid AI. Some AI researchers are working on developing AI systems capable of demonstrating human-like intelligence and consciousness, such as creating AI systems capable of passing the ‘Turing’ or ‘Chinese Room’ tests.
It is crucial to remember that Weizenbaum’s worries about AI remain valid today and the discipline of AI continues to raise ethical and societal implications. As AI systems advance and become more capable, it is critical to consider the implications of their development and application. For example, as AGI systems become competent at performing tasks previously thought to be unique to humans, it raises concerns about the future of work and the role of AI in society. Furthermore, as AI systems improve their ability to understand and interpret human behaviour, issues about privacy, autonomy and the possibility of AI being used negatively by individuals or society as a whole arise.
In light of these concerns, AI researchers and developers must think about the moral and societal implications of their work and develop AI systems that are open, auditable, accountable and consistent with human values. Furthermore, society as a whole must engage in informed discussions about the future of AI and its potential impacts on our lives. This includes involving stakeholders from diverse backgrounds and perspectives, such as ethicists, philosophers, sociologists, policymakers and members of the public, in the development and governance of AI systems. Additionally, there should be ongoing efforts to ensure that AI systems are developed and used responsibly and ethically, with measures in place to prevent unintended consequences and negative impacts on society.
Divergent Viewpoints on Prospects of AGI Morphing into ASI
While some researchers believe that AGI is possible, others think that it is unlikely and that the concept of AGI is not well defined. The debate over whether AGIs will eventually ‘morph’ into ASIs is divisive among AI researchers and experts for different reasons. One argument in favour of this viewpoint is that, due to advances in technology and machine learning, AGIs will continue to develop quickly and may one day outperform human intellect in several fields. This means that AGIs will continually grow exponentially over time and eventually transform into ASIs. However, since there are still many unresolved issues and unknowns surrounding AGI and ASI, many researchers are less optimistic about the timeline for their development. In addition, some specialists believe that the term ‘superintelligence’ is ill-defined and too ambiguous, making it difficult to say whether or when AGIs will develop to that level. Despite the difficulties, many researchers are still working on developing AGI and ASI because they think the potential advantages outweigh the risks. Improvements in decision-making, increased output and the ability to solve problems that are currently insurmountable for humans are potential benefits. However, to benefit from these advantages, the risks associated with AGI and ASI must be decreased through the implementation of proper safety measures. In addition to addressing moral and ethical dilemmas, this requires developing strategies for monitoring and controlling AGI behaviour.
As pointed out earlier, the majority of AI systems in use today are referred to as ‘narrow AIs’ because they are designed to carry out specific tasks. Examples of these types of AI include IBM’s Watson and DeepBlue, Expert Systems and AlphaGo, all of which exhibit intelligent behaviour within a specific domain but do not possess the general intelligence that humans have. Some of these systems may have high levels of proficiency or intelligence in their specific areas, but they do not have the broad intelligence that humans possess.
It is important to note that this is often a starting point in the progress towards creating a general AI. Some researchers typically begin by developing AI systems that are particularly skilled in one area and then use the insights and knowledge gained from these systems to advance the development of more general AI systems. Organisations such as the AGI Society, Berkeley Artificial Intelligence Research, CSAIL at MIT, Facebook AI Research, Google DeepMind and the Human-Level Artificial Intelligence (HLAI) Conference demonstrate that there is ongoing work in the field of AGI and that many experts are working to create more general AI systems.
I am among those who are nervous about the unleashing of AGI systems in our society. I, therefore, believe that when designing and implementing AGI, it is critical to proceed with caution and care. Technologies are amoral; they can be used for good – increased efficiency and productivity in various industries – but can also pose risks such as job displacement, security threats and ethical concerns, depending on the motivations of those who create them. Before proceeding with AGI development and implementation, it is critical to consider the potential risks and benefits, as well as the consequences of AGI morphing into ASI and the impact they could have on human jobs and our sense of uniqueness in the world.
Even though we still do not have AGIs today, Ray Kurzweil, among other AI experts, predicts that AGIs will be developed by 2045. He cited the Law of Accelerating Returns – which deduces that “the rate of technological growth is exponential” – to back this. It is crucial to keep in mind that these projections are based on current trends and advancements in the field of AI and that developing AGI is a complex and ongoing process that might not occur in a specific order. It is also worth noting that AGI does not necessarily imply replicating all human capabilities; rather, it could refer to systems that are advanced enough that humans perceive them as AGI, but the consequences of having such systems are unknown. It could be compared to opening a Pandora’s Box or attempting to construct a new Tower of Babel, with unknown and potentially negative consequences for humanity and the world. I believe that, in the long run the net effect will be negative or more catastrophic for our earth, unless we do something to regulate them appropriately.
Implications of AI
Humans have been attempting to increase the forms, types, places and reach of communication, resulting in the emergence of many forms of communication, including written language, oral language, sign language and, more recently, digital communication. For example, through the use of books, letters and other written documents, written language has made it possible for humans to communicate over great distances and long periods. The printing press facilitated written communication after the 15th century. The telephone and telegraph enabled long-distance communication in the 19th century. The development of radio, television and the internet has significantly expanded the reach and extensibility of communication and information during the 20th century. Communication with anyone at any time is now possible, thanks to the internet and mobile technology. Social networking and instant messaging are new forms of communication as well.
Most machines developed during the agrarian, industrial and post-industrial eras have ended up deskilling and displacing humans from their traditional vocations, whether in crafts, blue-collar or clerical work. However, they have also increased production and opened up new areas of labour for individuals who were displaced. As a result, there have been more employment increases than job losses overall. Many stakeholders believe that this will always be the case, even for artificial intelligence systems. However, AI can replace not only monotonous administrative and physical tasks, but virtually every other job, including those of artists, programmers, teachers, doctors, researchers, lawyers, accountants and managers. Indeed, everyone’s work. Managers have believed that having 1,000 employees will cause 1,000 headaches for them since the beginning of time. So, they will employ whatever machines or methods which allow them to eliminate numerous workers.
However, it is important to note that the impact of AI on the workforce will likely be more complex than simply replacing jobs. AI has the potential to improve human capabilities, produce more work and create new jobs. Additionally, the rate at which AI will impact different industries and job types will vary and some jobs may be more resilient to automation than others. Also, it is important to consider the ethical and societal implications of AI and its impact on the workforce. For example, there may be concerns about income inequality and the displacement of certain groups of workers. It is crucial for policymakers and industry leaders to carefully consider these issues and develop strategies to mitigate negative impacts while harnessing the potential benefits of AI. Moreover, there is a need to think about retraining programmes, education and upskilling of the workforce and to ensure that the benefits of AI are shared equitably across society.
Woherem, a highly respected industry professional and alumnus of Harvard Business School, wrote in from Abuja, Nigeria.
