AI Scientists Develop Power-Efficient System Requiring Only Light as Input
In a groundbreaking development, a team of scientists led by UCLA has introduced an innovative 'optical generative model' that could significantly reduce the carbon footprint of AI systems. The research, published in the prestigious journal Nature, details a new method that uses light during the decoding process to produce images, requiring only a fraction of the usual computational power.
The new system offers several advantages, particularly for AI-powered wearable systems where energy efficiency is crucial. By providing the image in a 'snapshot,' avoiding the thousands of iterative steps typically used in a digital decoder, this model could be beneficial for devices like AI glasses, where less power consumption is essential.
Alexander Lvovsky of the University of Oxford, one of the study's contributors, hailed this as the first example of an optical neural network being a computational tool capable of producing results of practical value. Aydogan Ozcan, the senior author of the study, further emphasised that the use of optics in AI tasks at scale could lead to energy-efficient AI systems that could transform everyday technologies.
The environmental impact of AI is a growing concern, especially considering the large number of images generated by AI models. Over a year, a user's carbon emissions from ChatGPT are about 11 kilograms, a figure that pales in comparison to a person's carbon impact from the energy industry alone. Each query on ChatGPT generates approximately 2 to 3 grams of carbon dioxide.
The UCLA-led team's system uses a liquid crystal screen called a spatial light modulator (SLM) to imprint image information into a laser beam. The results of the test were comparable to advanced diffusion models, but the process used only a fraction of the energy compared to those conventional models.
The integration of optical generative models into existing AI infrastructure is planned by companies and startups specializing in AI and optical technologies. For instance, Juniper Networks with their Marvis AI Engine for network optimization and OPTOCYCLE, a startup developing AI-based optical classification systems, are among those exploring this potential. Industry leaders in optical technology like Jenoptik contribute to the optical component expertise necessary for such integration.
The research shows that there's plenty of room for sustainable improvement in AI usage to reduce its contribution to climate change. As we continue to rely on AI for various tasks, it's crucial to develop energy-efficient solutions like the optical generative model to minimise its environmental impact.
Darren Orf, the author of this article, lives in Portland and writes/edits about sci-fi and how our world works. His previous work can be found at Gizmodo and Paste. The light-based decoders can also improve security and privacy by creating methods by which content is inaccessible except with the correct decoder.
As we move forward, it's exciting to see how this new technology could reshape the future of AI and its impact on our planet.
Read also:
- Reporter of Silenced Torment or Individual Recording Suppressed Agony
- EPA Administrator Zeldin travels to Iowa, reveals fresh EPA DEF guidelines, attends State Fair, commemorates One Big Beautiful Bill
- "Divine Protections Failed: Father Battling for Life After Flesh-Eating Bacteria Infection"
- Innovative Company ILiAD Technologies Introduces ILiAD+: Boosting Direct Lithium Extraction Technology's Efficiency Substantially
