Tesla Inc. Chief Executive Elon Musk expects to see artificial intelligence (AI) that beats human intellect by 2025 or the next, provided there is enough chip and electricity supply to meet the demand for such advanced technology.
The billionaire entrepreneur said during an interview on X that the creation of artificial general intelligence (AGI) smarter than the most intelligent human may be done next year or in two years.
AI computing power is also likely to surpass that of every individual in the next five years, according to Musk.
AGI Optimism Amid Chip, Electricity Supply Constraints
Musk, who was a co-founder of OpenAI Inc., has been consistently optimistic about the development of AGI, an AI system capable of challenging the smartest humans in any field.
However, his latest estimated timeline is earlier than his past projection. In 2023, Musk predicted a 2029 schedule for the creation of a ‘full’ AGI.
Other forecasts of the Space Exploration Technologies Corp. (SpaceX) chief, including launching self-driving Tesla vehicles and landing SpaceX’s Starship on Mars by 2030, have yet to be fulfilled as well.
Over the last 18 months, the AI space has presented several breakthroughs, with the introduction of video-generation tools and advanced large language models (LLMs) propelling the industry more quickly than anticipated.
Google LLC’s DeepMind Technologies Ltd. chief executive, Demis Hassabis, has stated that AGI could be developed by 2030.
Development, however, has lost some momentum as microchip supplies faced a bottleneck, mainly in US semiconductor giant Nvidia Corp., which are key components for AI learning and operation.
Musk said such challenges have abated, although new AI models are now looking into other data center gears and electricity grids.
Musk expects the AI development constraint to only involve electric supply in one or two years, following a lack of available chips in 2023 and as constraints shifted to supplies in voltage transformers this year.
Additionally, Musk stated that his AI startup xAI Corp. is expected to be done training version two of its AI chatbot, Grok, in May. The enhanced model is seen outperforming OpenAI’s ChatGPT 4.0.
Training Grok 2 involved 20,000 Nvidia H100 graphic processing units (GPUs), according to the billionaire, while Grok 3 and future versions of the chatbot might need about 100,000 of that chip.
