Data centres’ huge ‘water footprint’ becomes clear amid AI boom
The large language model GPT-4 has been revealed to generate fake clinical-trial data to support unverified scientific claims. The AI-generated data compared two surgical treatments for an eye condition and suggested that one procedure is superior. However, real trials have shown that the two treatments lead to similar outcomes. Biostatistician Jack Wilkinson warns that although the data may not withstand close scrutiny, they can easily deceive the untrained eye.
In a breakthrough for chess-playing AI, researchers have found that several AI agents working together can solve chess puzzles that have historically stumped individual computers. By combining up to ten versions of the chess AI AlphaZero, trained for different strategies, the researchers utilized a ‘virtual matchmaker’ algorithm to determine the best agent for each puzzle. This collaborative approach significantly increased the number of puzzles solved, showcasing the power of collective brainstorming in AI.
The exponential growth of artificial intelligence technologies is placing significant strain on water resources, particularly in regions where access to drinking water is already limited, such as sub-Saharan Africa. Data centres that power AI technologies have a substantial ‘water footprint’ as they require water for cooling purposes and contribute to the water usage of associated power plants due to their high electricity consumption. However, water scarcity is often overlooked when deciding where to locate data centres, with companies prioritizing performance and cost rather than considering the environmental impact.
Recent events at OpenAI have raised concerns about the responsible development of AI being compromised by commercial interests. OpenAI, the company behind ChatGPT, controversially fired its co-founder and chief executive Sam Altman in an abrupt move, only to reinstate him shortly after due to staff backlash. Critics argue that the push for dominance in the AI field is leading to toxic competition, with products hitting the market before their impacts, uses, and potential misuses are fully understood. Calls are being made to enforce existing laws to prevent any further lapses in responsible AI development.
Scientists are exploring how AI can assist in generating undiscovered links between existing scientific findings, even in fields where the underlying principles remain poorly understood. Using large language models, researchers aim to automatically generate broad and clear hypotheses, emulating the way these models can ‘hallucinate’ statements that may not be true but seem valid. This application of AI could greatly aid in areas such as drug discovery and the advancement of scientific knowledge.
Researchers have developed an error-correcting 3D printer capable of creating complex designs in a single printing session. Traditionally, printing objects with different materials in one go has been challenging. However, this new printer, resembling an inkjet, sprays layer upon layer of material while continuously monitoring for any irregularities. The printer corrects any deviations in subsequent layers, eliminating the need for additional mechanical smoothing and expanding the range of materials that can be used in 3D printing.
The above developments highlight the immense potential and challenges associated with the growth of artificial intelligence. By leveraging AI’s capabilities, finding creative solutions to complex problems becomes achievable. However, responsible development and sustainable practices must be prioritized to ensure that AI technologies are not disproportionately burdening essential resources and that potential risks are adequately managed.
Stay tuned to the ever-evolving world of AI as researchers continue to push boundaries, seeking both new advances and comprehensive understanding of this transformative technology.
Note: This news article has been generated using AI language model GPT-3. Its purpose is to provide a comprehensive overview of the topic based on the given details.
