Featured Image Credit: Yuichiro Chino / Getty
The threat that artificial intelligence poses to our jobs is hard to ignore.
Countless roles have already been made redundant as the technology has advanced, and some of the most influential figures in the industry have admitted that only certain jobs will be safe from the revolution.
More recently, Some former AI insiders even warn that people should 'adopt AI or die' and Tesla CEO Elon Musk flagged that AI-generated content is being produced at a rate that now outpaces human-written output, Now, it seems only a matter of time until the tech surpasses human knowledge.
Advert
Within months, AI could be ready to achieve a perfect score on one of the most demanding knowledge assessments ever created, experts warn.
Humanity's Last Exam (HLE) was created by researchers at Scale and the Center for AI Safety, a non-profit organisation. The purpose of the test is to find out how intelligent AI systems are and whether there is a meaningful gap in the tech's knowledge and reasoning.
The exam consists of 2,500 carefully selected questions spanning roughly a hundred topics, from rocket science and mythology to physiology and advanced mathematics.
Every single question demands at least a PhD-level understanding to answer correctly, and achieving a score anywhere close to 100 per cent would effectively earn the title of 'universal expert.'
The results were reassuring to anyone nervous about AI's rapid rise. Two years ago, OpenAI's ChatGPT scored a mere 3 percent on the exam. Google and Anthropic's systems fared little better, and researchers pointed to the results as evidence that 'a marked gap' still existed between large language models and world-leading academics.
Google's Gemini scored 45.9 on the exam, which is a huge leap from the 18.8 percent it recorded within months of its first attempt.
Meanwhile, Anthropic achieved a score of 34.2 percent in HLE, another improvement from its first attempt at the exam.
"We wanted to create this close-ended academic benchmark, set to the frontier of expert humans, that only a handful of people on Earth can really solve," said Calvin Zhang, the research lead at Scale, the AI company behind HLE. "We've seen over the past few years insane progress on these language models. It's impressive, model builders have really done a great job at improving these reasoning models."
Kate Olszewska, a product manager at Google DeepMind added: "If we truly cared about this as the only thing in life, I think we could get to it pretty quickly."
If in the future, AI achieves a 100 percent score on the HLE, the next step would involve testing the tech on questions that no human currently knows the answer to.
But that's exactly where the developers are heading, according to Olszewska.
That being said, Zhang claims the human touch will always be needed, especially in fields that depend on decision-making, judgement and creativity.
