If artificial general intelligence existed, it would be able to perform many tasks better than humans. For now, the machine learning systems and generative AI solutions available on the market are band-aids to reduce the cognitive load on engineers until there are machines that think like humans.
Generative AI currently dominates the headlines, but its backbone, neural networks, has been around for decades. These machine learning (ML) systems have historically served as cruise control for large systems that are difficult to manually maintain on an ongoing basis. Modern algorithms proactively respond to errors and threats, alerting your team and logging anomalous activity. These systems can also be further developed to predict specific outcomes based on previously observed patterns.
This ability to learn and respond has been adapted to all types of technology. One thing that sticks is the use of AI tools in envirotech. At this stage of development, AI offers great freedom, whether it's enabling new technologies with massive data processing power or improving the efficiency of existing systems by intelligently adjusting inputs to maximize efficiency. It is highly advanced and can theoretically be applied to any task.
Co-founder of VictoriaMetrics.
The undeniable strength of AI
GenAI is not inherently energy-intensive. While models and neural networks are no more energy-efficient in operation than other software, a large portion of the energy costs are incurred in the development of these AI tools. The justification for this energy consumption is that the future benefits of the technology will be worth the energy and resource costs.
Some reports suggest that many AI applications are “solution-seeking solutions” and that many developers spend vast amounts of energy developing tools with questionable energy savings at best. One of the biggest benefits of machine learning is its ability to read large amounts of data and summarize insights in a way that humans can act on. Reporting is a tedious and manual process, but you can transfer the time you save from reporting to leveraging machine learning insights to proactively address your business-related emissions.
Companies are under increasing pressure to start reporting Scope 3 emissions. Scope 3 emissions are the most difficult to measure and represent the largest source of emissions for most modern companies. Capturing and analyzing these disparate data sources makes smart use of AI, but still ultimately requires regular human guidance. Surveillance solutions that reduce the demand on engineers already exist on the market, so taking this one step further with AI is an unnecessary and potentially harmful innovation.
Replacing engineers with AI agents reduces human effort, but only removes complex interfaces and adds equally complex programming in front of them. This is not to say that innovation should be stifled. It's a noble goal, but don't be sold a fairy tale that this will happen without a problem. This technology will eventually replace some engineers, but the industry will need to tread carefully.
Consider self-driving cars. They're here and they're doing better than the average human driver. However, it can be dangerous in some cases. The difference is that this danger is much easier to recognize compared to the potential risks of AI.
Today's “smart” machines resemble naive humans
At this stage of development, AI agents are equivalent to human employees. It requires training and supervision, and will gradually become obsolete if not retrained from time to time. Similarly, models can degrade over time, as observed with ChatGPT. The mechanisms that cause this degradation are not clear, but these systems are carefully regulated, and this regulation is not a permanent state. The more flexible a model is, the more likely it is to misfire and not perform optimally. This can manifest itself as data or concept drift, a problem where the model becomes invalid over time. This is one of the problems inherent in applying probabilistic models to deterministic tools.
A concerning area of development is the use of AI in natural language input, which aims to make it easier for less skilled employees and decision makers to save engineering jobs. Natural language output is ideal for translating expert-specific output from a surveillance system so that even non-data savvy people can access the data. Despite this strength, even summaries can be subject to illusions if the data is fabricated, a persistent problem in LLM that makes it difficult for AI to summarize mission-critical reports. If used, costly errors can occur.
The risk is creating an AI overlay for a system that requires deterministic input. It's great to try to lower the barrier to entry for complex systems, but these systems require precision. AI agents cannot explain their inferences the way humans can or truly understand natural language input to resolve real-world requests. Additionally, it adds layers of energy-consuming software to the technology stack to minimize profits.
We can't leave everything to AI
The rush to “AI everything” is creating a huge waste of energy, and while there are 14,000 AI startups today, how many are creating tools that actually benefit humanity? Is it? AI can improve data center efficiency by managing resources, but in most cases, free capacity can be used for another application using the saved resource headroom and still required cost. This ultimately results in no meaningful energy savings. More AI-powered tools now available.
Can AI help meet sustainability goals? Probably so, but most proponents stumble on the “how” part of the question, suggesting that in some cases AI itself will come up with new technologies. I am. Climate change is now an existential threat with so many variables that it is beyond the understanding of the human mind. Rather than tackling this problem directly, tech enthusiasts are handing over the responsibility to AI in the hope that it will provide a solution at some point in the future. The future is unknown and climate change is happening now. Relying on AI to save us is simply folding our hands and hoping for the best in futuristic guise.
We've listed the best collaboration platforms for your team.
This article is produced as part of TechRadarPro's Expert Insights channel, featuring some of the brightest minds in technology today. The views expressed here are those of the author and not necessarily those of his TechRadarPro or Future plc. If you're interested in contributing, find out more here. https://www.techradar.com/news/submit-your-story-to-techradar-pro
