Elon Musk loves making a bold prediction or two, and where artificial intelligence (AI) is concerned, he's even more keen.
He's talked before about the percentage chances that fully-developed AI wipes out humanity (don't worry, it's only between 10 and 20%, in his eyes) and now has been discussing when he thinks there will be an AI that is smarter than a human.
The discussion was part of an interview broadcast on X's Spaces feature, between Norwegian wealth fund CEO Nicolai Tangen and Musk, and touched on a wide range of subjects.
Advert
When it was the turn of AI, though, Musk made a bold proclamation, saying: "If you define AGI (artificial general intelligence) as smarter than the smartest human, I think it's probably next year, within two years."
Musk has long been involved with AI development, having been on the board of directors at OpenAI at one point before latterly starting his own AI company, xAI, so he's likely to know what he's talking about.
Interestingly, Musk also touched on a part of the AI race that perhaps isn't talked about as often as others - the literal power supply.
He explained at one point that shortages can be a huge challenge for AI development, in two ways. Firstly, chip shortages mean that the processing units that these AI models rely on are in short supply and therefore are seeing price spikes.
Given that, as he said, the second generation of X's Grok chatbot required some 20,000 Nvidia H100 GPUs to be trained, you can see how buying these would be a challenge.
Secondly, though, he said that actually getting the electrical supply to power all of these GPUs, and to power the cooling systems that stop them all from overheating constantly, is another massive difficulty.
Still, there's a difference between something that's challenging to overcome and something that is impossible to manage, and it sounds like Musk thinks that most of these issues fall into the former camp.
So, you can probably expect AI development to keep sprinting forward, although whether that prediction of Musk's about its level of intelligence comes true will be hard to gauge.
For one thing, although chatbots like ChatGPT and Grok are publicly accessible, it's a pretty safe bet that they don't represent the true cutting-edge of the most intelligent and sophisticated AI models out there right now.
These are most likely under lock and key in far more secretive surroundings, to ensure that competitors can't learn from their responses. That's before you even consider the likelihood of government bodies that have their own research facilities for this sort of thing.
So, AI might outstrip human intelligence in the next year, but it'll be interesting to see if we actually hear about it.