Recently, a new study from the University of Michigan found that an energy-efficient method for training large language models can achieve similar results in the same amount of time while reducing energy consumption by 30%. This method can save enough energy to power 1.1 million American households by 2026. The researchers developed a software tool named Perseus that identifies the critical path, which is a series of subtasks requiring the longest time to complete. Then, Perseus reduces the processor speed on non-critical paths so that all tasks can finish simultaneously.

The development and application of artificial intelligence (AI) technology have sparked a global boom in recent years, but some experts are starting to worry about the future of this industry. James Ferguson, the founding partner of the UK macroeconomic research firm MacroStrategy Partnership, said in an interview with Bloomberg's "Merryn Talks Money" podcast that the AI industry might be in a bubble that, once burst, could have catastrophic consequences.Image Source Note: The image was generate

Groq has launched the new AI processor LPU, which has excellent performance but requires improvements in cost and energy consumption. The LPU is a processor specifically designed for language processing, challenging the traditional GPU's dominance in language tasks. While the LPU boasts advantages in efficiency and cost-effectiveness, its application range is limited. Groq has provided a detailed introduction and explanation of the design and technology behind the LPU.

Using AI to generate images, write emails, or ask chatbots contributes a certain burden to the planet. The energy consumed to generate an image using AI is equivalent to charging a mobile phone fully. The energy consumption for generating text is lower, as generating text 1000 times only consumes 16% of a mobile phone's charging capacity. The use of large generative AI models is more energy-intensive than smaller models.

Meta engineers estimate that next year, there will be an addition of 1 to 2 million Nvidia H100 graphics processors for global AI inference. If all of them are used for language model generation, the energy consumption would fall within a reasonable range, generating 100,000 tokens per person per day. Only two additional nuclear power plants would be required to provide sufficient power for global AI inference. Model training faces limitations due to data scarcity, and researchers are working to improve model efficiency. In the next 3-4 years, it is expected that we will know whether current technology can achieve general AI.

Recent research warns that the power demand of artificial intelligence (AI) will grow rapidly, potentially reaching 850 to 1,340 terawatt hours per year by 2027, equivalent to the annual electricity consumption of some countries. Experts are calling for tech giants like OpenAI and Google to reassess their high-energy operational practices to reduce environmental costs. California has signed a climate disclosure bill requiring companies to report their carbon emissions, but the AI sector remains self-regulated. For sustainable development, the AI field needs more innovation and attention to meet its growing power demands.

{"title": "3. AI Can Bring More Value to the World Compared to Bitcoin", "content": "The application of AI can bring more value to the world, not just attract speculators and investors."}