The article Shocking Numbers: Report revealed environmental impacts from AI first appeared at the online magazine Basic Thinking. You can start the day well every morning via our newsletter update.
How heavily contaminates artificial intelligence? The French company Mistral has presented a detailed environmental balance of its AI models for the first time. The report thus throws a new light on CO2 emissions and water consumption of chatbots.
Artificial intelligence promises productivity and innovation – but has high electricity consumption. What many have already guessed is now occupied Environmental report The French Ki start-up Mistral Ai: The development and use of large language models are a real climate killer.
The background: So far it was unclear how big the environmental pollution from AI really is. According to Mistral’s first environmental audit, a single chat with the in-house AI “Le Chat” (with around 400 generated words) causes 1.14 grams of CO2 and uses 45 milliliters of water. For comparison: this corresponds to about ten seconds Netflix in the USA or a four to 27 second zoom call.
How much does AI harm the environment?
But what may sound harmless is a real problem on the mass of the systems. In the first 18 months after the launch of the “Large 2” model, Mistral emitted a total of 20,400 tons of CO2 due to the AI training and the millions of use of the chatbot. This roughly corresponds to the output of 4,500 combustion cars a year. In addition, around 281,000 cubic meters of water evaporated, which could fill 112 Olympic swimming pools.
According to the study, most of the emissions (over 85 percent) do not fall when chatting itself, but in AI training. Huge amounts of data are led by processors for months – mostly in data centers that are not operated with green electricity. The construction of the hardware also costs ecologically negative.
Mistral has therefore proposed a new metric: a ratio of overall inference to the entire life cycle balance of a model. Or simply said: “Is the training worth it ecologically compared to later use? In the future, corresponding figures will help to better compare the efficiency of AI models.
A small step towards transparency
Mistral environmental report is on the one hand a milestone, but on the other hand, it also reveals weaknesses. Experts criticize that important details, such as the exact methodology or energy consumption, are missing across the entire life cycle.
In the future, disclosure and comparability could lead to a real “eco-score for AI”-similar to the energy consumption of electrical appliances.
The claim seems to be clear: If consumers use AI, you should be able to understand the environmental balance in the future. Mistral takes a step in the right direction and invites the competition to do the same to the company.
Also interesting:
- Lithium of sea water – how a new membrane should help
- This AI helps to cool down – through material analysis
- Openaai shows: This is how GPT-4 feels “intuitive”
- How a start-up with green cloud wants to set new standards
The article Shocking Numbers: Report revealed the environmental impact of AI first appeared on Basic Thinking. Follow us too Google News and Flipboard Or subscribe to our update newsletter.
As a Tech Industry expert, I am not surprised by the shocking numbers revealed in the report about the environmental impact of AI. The rapid growth and adoption of artificial intelligence technology have undoubtedly contributed to an increase in energy consumption, carbon emissions, and electronic waste.
While AI has the potential to revolutionize industries and improve efficiency, it is crucial for companies and organizations to be mindful of the environmental consequences of their AI systems. This includes implementing energy-efficient algorithms, optimizing data centers, and responsibly disposing of outdated hardware.
As the demand for AI continues to grow, it is essential for the tech industry to prioritize sustainability and invest in green technologies to mitigate the environmental impact of AI. By taking proactive measures now, we can ensure that AI remains a force for good while minimizing its carbon footprint.
Credits