Notebooklm, ChatGpt: How to generate learning notes in AI
The AI-powered tool will help you check your notes and test preparations, no matter what your learning style is.
The problem has been resolved
From job displacement and plagiarism, accuracy and privacy risks, generative artificial intelligence, or genai, is full of the types of problems that are expected of such disruptive technologies.
And you can add an environmental threat between them.
Especially since the second half of 2022, when Openai’s ChatGpt exploded into the scene by triggering the global Genai Gold Rush with fierce competition and unprecedented investment, many have sounded alarm bells for environmental challenges caused by the enormous computational resources needed to train and operate these models.
In other words, the training process for large-scale language models (LLMs) consumes a huge amount of electricity and contributes to increasing carbon emissions. Additionally, the data center that houses this hardware requires a huge amount of water for cooling, putting a strain on local water supply and ecosystems.
For end users (consumers, businesses, governments, etc.), Genai’s rapid development cycle also leads to accelerated demand for high-performance computing hardware that can have a unique, negative, “full circle” impact on your home country, from manufacturing, packaging and transportation to energy consumption and ultimate e-waste.
What is genai’s carbon footprint?
Generic AI platforms – including LLMS to generate human-like conversations, and image-producing AI tools – require a lot of electricity to run, especially at large scale. For example, it is estimated that a single Genai prompt uses about 3 watt-hours (WH) of electricity, while a typical refrigerator uses about 1-2 kilowatt-hours (kWh) of electricity per day. When doing math, the day of refrigerator usage (average about 1.5 kWh) corresponds to approximately 500 generation AI prompts.
That may not be too bad, but with the possibility of some daily interactions between billions of users and genai, cumulative effects become important. Genai is so resource-heavy, Openai CEO Sam Altman recently admitted that users are saying “please” and “thank you” while interacting with ChatGpt and interacting with “tens of millions of dollars” of electricity costs.
What exacerbates this problem is that most energy is still sourced from pollution of fossil fuels such as coal oil (oil) and natural gas.
“People don’t go to Wikipedia or open a book, they look at all Chatgupt and other platforms.
“Large language models use roughly 30 times more energy than (websites) because the process requires more calculations, with large-scale language models generating results from scratch to prompt based on training data,” adds Dr. Lucciioni, who focuses on promoting sustainable AI practices. “We’re taking more resources without actually looking at the costs. Users aren’t looking at the invoices. It’s free or almost free. So I don’t think it’s really connected to the environmental impact.”
What do AI companies do to address climate concerns?
If there is a solution to combat the environmental impact of AI, it will not be realized or implemented immediately.
Meta CEO Mark Zuckerberg recently announced that the company behind Facebook, Instagram, WhatsApp and the threads are planning to spend “thousands of billions of dollars” on developing AI products in the near future.
According to Zuckerberg, it will be called Meta’s first multi-gigawatt data center and is expected to be online next year.
On behalf of USA Today, I contacted Meta to comment on this story, but the request was not accepted.
Some AI players are trying to better understand that ultimately reduce the impact on the environment. For example, Mistral AI is a company that offers a variety of LLM solutions and discussed its efforts in a recent blog post entitled “Contribution to AI’s Global Environmental Standards.”
In a statement provided to the vice president of public relations for Mistral AI, Audrey Helbrinstorp at USA Today, “We recently conducted the first comprehensive study to quantify the environmental impact of LLM.
In particular, it shows that the model’s footprint is strongly correlated with its size and that selecting a smaller or case-specific model can help reduce environmental impact.
Other efforts can also have positive results in the environment. In August, Google agreed to curb the use of power in AI data centers and ease the burden on the grid when demand surges. Google Gemini is becoming a popular and mainstream Genai tool.
Dr. Lucciioni says that by limiting what we rely on AI, we can also do our part. “Do I need to generate cookie recipes from scratch, or can I look them up in recipe books or online?” she asks rhetorically. “Today, people use AI as calculators, encyclopedias, and even therapists, so thinking about alternatives is key.”
AI to help combat environmental threats?
It is also worth noting that AI models can actually help combat climate change.
Although emerging technologies themselves have environmental impacts, AI can also improve resource use and energy distribution, increase efficiency, analyze and optimize waste flows, and predict and mitigate the impact of climate change.
“The most commonly used AI models to combat climate change are not the most harmful models,” explains Dr. Lucciioni. “These aren’t an issue in terms of environmental impact, as they are so efficient and can run locally on laptops, such as models that are typically used for climate forecasting and biodiversity monitoring, and new generation batteries and what you have.”
But for now, Dr. Lucciioni says that because of the huge demand for AI, the demand for AI is very large, as technological advances that increase the efficiency of resource use can paradoxically increase the overall consumption of that resource.
“Many tech companies are able to do more and more because they are increasing efficiency, including NVIDIA, including hardware for each generation, but people are using it even more. “For now, all this demand for all of this genai is really, really growing, so we need to find a meaningful solution.”
