Quick Takeaways
- AI systems in the U.S. consumed over 415 terawatt-hours in 2024, accounting for 10% of the country’s electricity, with demand expected to double by 2030, raising sustainability concerns.
- Researchers developed a neuro-symbolic AI system that combines neural networks with symbolic reasoning, promising up to 100 times energy savings and enhanced performance.
- The hybrid system achieved high success rates and faster learning in puzzle tests, using only 1% of the energy required by traditional AI models during training and 5% during operation.
- As AI’s energy demands grow, neuro-symbolic AI offers a more sustainable and reliable alternative to current large language and visual-language models, potentially mitigating long-term environmental impacts.
AI Uses Too Much Energy, But Change Is Coming
Artificial intelligence has become a powerful tool in many fields, but it also consumes a lot of electricity. In 2024, AI systems and data centers used over 415 terawatt-hours of power in the United States. That’s more than 10% of the country’s total electricity. Experts warn that this demand could double by 2030. As a result, scientists are searching for ways to make AI more energy-efficient.
New Technology Reduces Energy Use and Boosts Accuracy
Researchers at a School of Engineering have developed a new AI system that could change things. This system, called neuro-symbolic AI, uses a different approach. Instead of relying only on data patterns, it combines traditional neural networks with logical rules. This hybrid method mimics how humans solve problems—by breaking them into steps and categories.
The team tested their AI on the Tower of Hanoi puzzle. The results were impressive. The new system succeeded 95% of the time, even on tricky versions it hadn’t seen before. Traditional AI models failed all attempts at the same task. Additionally, this system learned the puzzle in only 34 minutes, while older models took more than a day and a half.
Lower Energy Consumption and Faster Performance
Another advantage of this AI is its energy efficiency. Training it used only 1% of the energy needed for older models. During its operation, it required just 5% of the energy usually used. This is a huge step forward because many AI tools today waste a lot of power just to predict what comes next.
For example, searching on Google with AI summaries uses up to 100 times more energy than the simple process of showing webpage links. As AI becomes more common in industries, it can put a strain on power supplies and environment-friendly goals.
A Better Path for the Future of AI
Many experts believe current AI systems may not be sustainable long-term. Large language models and visual-language-action systems are powerful, but they are also energy hogs and sometimes give inaccurate results. The new neuro-symbolic approach offers a different path. It combines learning with structured reasoning, making AI more reliable and less demanding on energy.
This breakthrough suggests that future AI could be smarter, faster, and greener. As technology advances, flexible and efficient tools like this could help meet the growing demand for AI without overloading energy resources.
Continue Your Tech Journey
Dive deeper into the world of Cryptocurrency and its impact on global finance.
Discover archived knowledge and digital history on the Internet Archive.
AITechV1
