Musk Believes AI Superhumans Will Be Smarter Than Humans By Next Year

Elon Musk predicts that superhuman AI will begin to surpass human intelligence from next year. “My guess is that we’ll have AI that is smarter than any one human probably around the end of next year,” Musk stated in an interview on his social network X.

This could mean that what is currently being developed will soon involve a major expansion from a capabilities standpoint. Its also worth remembering that Musk has made quite a few predictions such as the self-driving Tesla and Mars missions, in the past. This one in particular, though, comes with the drawbacks, risks and concerns surrounding AI.

What’s Holding AI Back, And How Soon Until AI Fully Surpasses Human Intelligence?

Last year, AI development faced a slight slow down, due to a shortage of Nvidia chips, which are key for training sophisticated AI models. Musk elaborated, “Last year it was chip-constrained. This year it’s transitioning to a voltage transformer supply. In a year or two, it’s just electricity supply.”

Musk’s latest prediction is sooner than his previous estimate, suggesting that within five years, AI might surpass the collective intelligence of humanity. This forecast is closer than his 2023 projection of achieving “full” artificial general intelligence by 2029. The pace of AI development has been rapid, with new models and capabilities coming in sooner than the world thought they would.

Right now, the development of AI technologies is being tested by supply issues, over and above the availability of the specialised chips and sufficient power supplies. With these, the development of advanced AI is possible, and so the supply should meet the increasing demand, as according to Musk.

What Do Other Experts Say?

A few experts have shared their views on when they predict AI will surpass human intelligence. There are some interesting takes, and some speak to what Musk has predicted as well.

Our Experts

• Christian Rebernik, Founder And Co-CEO, Tomorrow University Of Applied Sciences
• Jeff Watkins, Chief Product And Technology Officer, xDesign
• Roger Jackson, Founder, SenseCheck
• Eleanor Watson, IEEE Member, AI Ethics Engineer And AI Faculty, Singularity University
• Matyáš Skalický, AI Engineer, Rossum
• Dr Seun Kolade, Professor of Entrepreneurship and Digital Transformation, Sheffield Business School
• Niloufar Zarin, Head of AI, Thrive Learning

Christian Rebernik, Founder And Co-CEO, Tomorrow University Of Applied Sciences

“Musk’s prediction that AI will surpass human intelligence rings true, particularly in the light of the recent passing of the EU’s Artificial Intelligence Act, confirming AI’s permanence. It’s crucial government bodies and businesses are mindful about its adoption, and ways of working alongside it.

“The successful integration of AI into the workforce relies on nurturing cultures that prioritise continuous learning and the investment of employee upskilling – vital for equipping individuals with practical skills and encouraging critical thinking to effectively manage AI applications. This is integral in discerning between tasks that still necessitate human expertise and those that can be automated by the likes of AI and Machine Learning.

“The collective efforts of unions, regulatory authorities, educational institutions and CEOs are vital in ensuring AI’s ethical and equitable deployment of AI, balancing profitability with human productivity and well-being. When used to complement human talent, AI will open the door to more purpose-driven work environments and drive long-term economic success.”

Jeff Watkins, Chief Product And Technology Officer, xDesign

“When it comes to discussing the question of AI surpassing human intelligence, there’s the awkward hump we have to get over of defining intelligence in the first place – and whether it’s a single definable item or not. Although opinion is split on the different types of intelligence, experts are fairly clear that it’s not a single entity.

“Since the first calculators, machines have been doing certain “intelligent” jobs in a way that surpasses humans in both speed and accuracy. As computing power continues to surge forward, complex data analysis on the scale that no human could feasibly achieve has become routine, as have more complex tasks that require forward planning.

“But what’s more interesting to me is in the age of AI, we’re going from solving things better and quicker, but in a human-programmed way, to letting the machine discover a more optimal approach.

“A good example of this is Deep Blue, a chess computer programmed on a huge number of chess openings, midgame patterns and endgame strategies. This approach involved human-like thinking, enhanced by a more accurate memory and better ability to predict future moves.

“Another chess programme, AlphaZero (a more generalised version of AlphaGo), effectively programmed itself by playing against itself and discovering how to play chess without any knowledge of openings or strategies. It makes otherworldly moves; ones that seem strange or even wrong, but then prove to be a master-stroke many moves later. This is when AI has not only surpassed humans in speed and accuracy, but also approach, at least for that task, because it is no longer limited by as many of our limitations.

“As time goes on, more analytical problems (often referred to as componential intelligence) will move into the bucket of “things a machine can do better than a human”. We then get into the other things that could be considered intelligence, such as emotional intelligence, experiential intelligence, and contextual intelligence.

“Some, if not all, of these will require us to have developed general intelligence AI, one that can actually live a life and learn as we do, albeit at an accelerated rate. We can start to codify these things and make a facsimile of them in order to solve pockets of it, but that wouldn’t be general intelligence.

“If you look at the potential areas of intelligence and where we’ve made a significant advancement, it’s clear that it’s mainly in the space of componential intelligence, which is in itself a huge space. But, we have not really made much inroads into the other areas we could consider intelligence.

“Moreover, the advancements made have been in quite definitive areas, ones where quality and outcomes are measurable – and mainly in the digital space. Going through the cyber-physical convergence is a slow process as, even though robotics, optics and battery technologies are advancing, they’re not moving as quickly as the digital world, which will slow down the dream (or nightmare) of a machine that can gain experience through living a full lifetime in our world.

“In conclusion, although it may happen, AI will not best all of human intelligence in our lifetimes. But that doesn’t mean we’re anywhere near slowing down on raising the bar for what a machine can do better than we can and, if anything, this pace of change may speed up for the foreseeable future. However, I predict it will be at least a century before it is even possible to surpass us in all areas. Sadly, as I’m not a machine, I won’t be around to find out.”

Roger Jackson, Founder, SenseCheck

“My experience of how AI is impacting the marketing world shows how far away we really are from replicating human intelligence in any substantial way. Because AI delivers outcomes based on “learning” from past patterns, it isn’t able to come up with innovative solutions to unpredictable situations, or even unpredictable solutions to established situations. Both of these are at the heart of great marketing.

“AI’s core capabilities make it great for doing the repetitive work of preparation, but it’s nowhere close to actually creating “good” marketing outcomes – that is, ones that sway the behaviour and perceptions of a human audience.

“What we call “AI” at the moment is really just a very sophisticated programme that delivers outcomes from analysing at a large scale what has already been done in the past. Its outcomes are predictable and very ordinary for precisely that reason. That’s a million miles away from creative human endeavour. And one must also note that some of the so-called AI tools are actually underpinned by thousands of humans tapping away to support the software. So, my answer to the question of how soon until AI surpasses human intelligence? Not any time soon. Although AI will indeed do some things better than humans, particularly if the task involves at its heart mundane pattern recognition.”

Eleanor Watson, IEEE Member, AI Ethics Engineer And AI Faculty, Singularity University

“Often, it’s not a question of can we do something technologically – we almost certainly can – but of, dare we? Are people ready for it? I’m more concerned about human reaction to machines than machines themselves.

“In the past two years, powerful new AI technologies have emerged. These are based around multimodal data as well as abstraction processes (using prompts to ask the AI to do things). Both these systems are very large and powerful, but they require enormous amounts of data and training time. Both are able to deal with tens of thousands of problems, instead of just one or two like a typical Deep Learning system.

“Then there’s explainable AI, which focuses on making machine learning models interpretable to non-experts, which will become increasingly important as these technologies impact more sectors of society. Over time, it is likely that these technologies will help us to make sense of things in ways that were not possible before, and can therefore help to solve the bigger challenges, or at least to find a few optimisations.

“AI is making rapid strides and will help with predicting problems before they manifest. However, the real world is very complicated, and human supervision is often required as problems arise.”

Matyáš Skalický, AI Engineer, Rossum

“The world all around us is changing with AI use-cases continuously emerging. There is no question that AI will eventually reach and surpass human intelligence. But human-level AI is not imminent, on the contrary, it’s quite far away. We are entering an era of specialised AI applications that will make humans much more efficient. But to reach human-level intelligence, we can’t just keep scaling the current methods with larger datasets and faster GPUs. We need to discover new technology and architecture of AI systems. There are many unknowns, it could take 10 years, it could take 50. We don’t know.”

Dr Seun Kolade, Professor of Entrepreneurship and Digital Transformation, Sheffield Business School

“The notion that self-modifying AI will become super-intelligent and ultimately surpass human intelligence is a popular myth. It betrays a misunderstanding of how the technology works, and it is contrary to what some experts have proposed as the laws of Artificial Intelligence. In effect, intelligence typically develops in response to challenge, and AI will, in the end, only be as intelligent as humans force or encourage it to be.

“The second point that the design of each AI requires complex structures and algorithms such that it is practically impossible to design an all-purpose AI that perfectly mimics human intelligence. There are more credible concerns that, as AI tools gain increasingly prominent role in the workplace, some workers can become more dependent on them, leading to some form of enfeeblement.

“However, the fear of enfeeblement is as plausible as the prospects of enablement and super-productivity. In short, the future of AI is ultimately what we make it to be, and that is where a lot of attention has turned to how we can use AI responsibly and ethically.”