AI accelerates toward
human intelligence

Artificial General Intelligence (AGI) – when machines can understand and learn any intellectual task that a human being can – may arrive sooner than predicted. Some experts are now forecasting human-level AI within 18 months. With the accelerating pace of AI development, experts are considering its far-reaching implications.

“The pace of acceleration is happening at a daily, at an hourly, rate,” said Peter Xing, data and AI specialist for education at Microsoft, during the recent media launch of the Singularity South Africa Summit. The launch was hosted online in the Ubu metaverse, and the Summit will run on 22 and 23 October 2025 at the Sandton Convention Centre in Johannesburg.

“AGI hitting by 2027 has about a 12% chance right now. Before 2028, it has about a 40% chance. And by the time it’s 2030, we’re seeing 60%. By the time it’s 2035 it’s 78%. You’re only seeing this sort of increasing likelihood of when human level AI will come through.”

This projection comes alongside recent breakthroughs, such as the performance of GPT-4.5, which Xing said has already demonstrated capabilities comparable to human intelligence in areas like mathematics and advanced science. He referenced the Turing Test, which GPT-4.5 has passed – a standard initially set to determine whether a machine could convincingly mimic human responses.

“In fact, if you look at this sort of competition of maths, ChatGPT o3 has already beaten all of the best humans in math and in PhD level science, around graduate-level Google-proof diamonds.”

This dataset is structured to assess both AI models and human experts on complex, graduate-level questions across STEM fields. The diamond subset, comprising 198 particularly challenging questions, was selected for its clarity and unambiguous answers, making it a stringent test for non-experts while ensuring accuracy through expert validation.

One of the most significant societal impacts of AGI will be its role in the workforce. Xing said that once human-level AI emerges, it could be replicated at scale, fundamentally altering the labour market.

“Because once you have a human level AI or beyond, you’ll just keep creating copies of it, and you’ll have 1,000s of these human level AIs that be able to compete in the workforce market.”

Xing said there are he potential social implications of decoupling work from income. As AI takes on more roles traditionally held by humans, there may be a need to rethink how society values work and personal fulfilment.

“A lot of people drive get their meaning out of doing work, but if you’re telling them that that work is no longer a compulsory thing to do. Will they still do it? Will they continue to do the activities, and, you know, the hobbies that they would typically not be paid for?

“We watch the Olympics because we want to see humans playing with other humans, right? We’re not relying on a robot versus a human, because we want to see what actually is fun and competitive in that space where everyone’s playing by the same rules.

“So ideally, we help shape society make sure they’re not losing in that hedonistic future where they lose meaning and have a be so bored that they result into drugs and doing things that are sort of not fulfilling, and lose dignity,” said Xing.

“We want them to actually embrace that unlimited experience and also compete in the metaverse just for the pure ability to have that meaning and set the purpose to be built. Hopefully, that is the future, and we want to make sure that we continue to make that possible, and that we want to make sure that the future is accessible to all, because it’ll be really unfortunate if it’s not.”

Xing said that as AI becomes more integrated into education, universities face the challenge of remaining relevant. He said that many students question the purpose of pursuing a degree when they are already using AI tools like ChatGPT and Azure Open AI Socratic tutors in their daily learning.

He said that the educational process will need to evolve, shifting away from a factory-like model focused on rote learning and assessment toward a more meaningful and purpose-driven experience. He envisions a future where education emphasises broader life skills, adaptability, and the ability to utilise AI tools effectively – preparing students not just for work but also for a future where work may be less central to one’s purpose.

The future of AI remains uncertain, but the rapid pace of development and the potential for human-level intelligence to emerge within the next few years are clear. As AGI continues to advance, Xing said there is a need to reframe education as a means of preparing individuals to navigate a world where work may not be a necessity.

* Jason Bannier is a data analyst at World Wide Worx and writer for Gadget.co.za. Follow him on Bluesky at @jas2bann.