Human factor crucial in curbing potential AI risks
Are we on the brink of an epoch where artificial intelligence transcends human intellect?
The concept of singularity suggests that AI could surpass human intelligence in all domains of knowledge, creativity, and reasoning. Already, AI has demonstrated prowess in undertakings such as strategic game-playing and rudimentary creative endeavours, advancing into more complex domains like medical diagnostics and autonomous navigation.
There is a passionate debate surrounding singularity, with some advocates predicting it could occur as early as 2045 due to the rapid advancement of AI. Conversely, sceptics contend that such advancements might require centuries. This concept of singularity, promulgated by the visionary Verner Vendel, envisions a future where AI surpasses human intellect in every conceivable metric. Such a phenomenon would mark an unprecedented era, where humanity coexists with superior intelligence.
The 2024 AI Index Report from Stanford University provides a comprehensive overview of the rapid advancement of AI spanning from chess to advanced mathematics. The report indicates that AI is now paralleling or even exceeding human performance in fundamental tasks like reading comprehension, image classification, and complex mathematics. Researchers are grappling with the challenge of devising new benchmarks, as AI rapidly master and outstrips existing tests. The report also mentions a growing divide in how countries perceive AI, with some embracing its potential while others grow increasingly wary. This divergence could precipitate future geopolitical dynamics as countries vie to innovate and regulate AI technology.
In Kenya, there is a pronounced enthusiasm for AI among both the government and the private sector. Policymakers are earnestly considering the far-reaching impacts of AI.
It is astonishing to contemplate that machines are beginning to emulate, and potentially exceed, human capabilities in domains once considered uniquely human. However, will AI genuinely surpass human intellect within our lifetime?
Last year, Alan D. Thompson, erstwhile chairman of Mensa International, a high-IQ consortium, conducted a study into ChatGPT’s IQ and its performance across various professional disciplines. He compared AI’s outcomes with the average performance of experts in these domains, finding that AI demonstrated superior preparedness in seven out of ten assessments. Nonetheless, he also observed that general logic and common sense remain dominantly human attributes.
Some experts posit that upon reaching a certain echelon, AI will possess the capability to engineer even more intelligent AI autonomously, precipitating an exponential surge in machine intelligence. This notion elicits apprehension among some, who fear that AI might evolve into a menace if it cultivates superiority or self-preservation instincts.
To realise this ‘singularity’ potential of AI, the primordial step is the development of artificial general intelligence (AGI), which enables AI to apply knowledge acquired from one task to learn and execute myriad other tasks. This generalisation of learning mirrors human experiential learning.
Achieving AGI requires AI systems that can comprehend context and meaning, rather than merely scrutinising patterns. They must think abstractly and creatively, transcending the mere reorganisation of existing data and mastering common-sense reasoning. The major task for AI will be replicating self-awareness and consciousness, which are still not completely understood in humans.
The future of AI is uncertain, reliant on human interaction. Understanding this is crucial to preventing potential issues like privacy loss, job takeover, and cyber-warfare. Increased empathy and political reorganisation are necessary.
The writer is an Innovations Evangelist and a PhD Candidate
