In 2023, interacting with OpenAI’s ChatGPT may have seemed innocuous—like asking for a recipe or historical trivia could invisibly consume around 2 watt-hours of energy. That’s comparable to leaving an incandescent bulb on for roughly two minutes. However, the introduction of GPT-5 changes this dynamic dramatically, with estimations suggesting these interactions might now consume up to 20 times the energy, although OpenAI hasn’t disclosed specific figures.
With GPT-5, released in mid-2023, OpenAI promises unprecedented AI capabilities. This new version is designed to excel at tasks as varied as website development and providing complex scientific answers. But, as AI models advance, they often become more resource-intensive. Experts estimate that responding through GPT-5 could consume about 18 watt-hours—akin to running a traditional 60-watt bulb for 18 minutes—making it significantly more demanding than any of its predecessors.
OpenAI’s reticence in providing energy usage details for its models is noteworthy. Since GPT-3, OpenAI has stayed mum on disclosing power consumption specifics, a move mirrored by many in the AI industry. This non-disclosure has sparked discussions amongst experts and environmental advocates who are pressing for greater transparency to better assess the real-world environmental costs of using such advanced AI systems.
Researchers from the University of Rhode Island and the University of Illinois have been working on benchmarking AI models, examining their energy and resource demands. Despite potential mitigations like more efficient hardware and innovative architectures such as the “mixture-of-experts,” the capability expansions in GPT-5—which include comprehensively handling text, video, and images—likely ensure its energy footprint remains substantial.
Calls for OpenAI and other AI developers to publicize the environmental impact of their models are becoming increasingly vocal. Highlighted by computational experts, understanding and mitigating the resource burden of evolving AI is essential. This balance is vital to ensure that while technology progresses, it remains sustainable and responsible.
Key Takeaways:
- OpenAI’s GPT-5 represents a significant advancement in AI technology, but it comes with increased energy demands, potentially up to 20 times more than its predecessors.
- OpenAI has not released detailed energy consumption data since GPT-3, which has led to increased demands for transparency regarding the environmental impact of AI models.
- As AI models grow in complexity, understanding their environmental cost is critical to ensuring their development is sustainable.