How will AI impact society and the environment? – What are the potential impacts of AI on society and the environment? And what is meant by eco-socially responsible AI systems? Well, stick around and you’ll find out. So there’s a non-profit organization called the Montreal AI Ethics Institute that recently published a paper called “SECure: A Social And Environmental Certificate for AI Systems.”
The paper focuses on the fact that AI will have a significant and unsustainable impact on the environment and society and that a framework should be developed to minimize that impact in an eco-socially responsible way. The framework proposed by the Institute consists of four pillars, they are; compute-efficient machine learning, federated learning, data sovereignty, and LEEDesque certificate.
Before diving into these pillars, let’s discuss what these impacts are exactly. So, many AI algorithms can have a tremendous carbon footprint and that’s due to the excessive data, compute, and power requirements needed to train certain AI models that perform well. And what that means is potentially a lot of carbon dioxide emissions.
Also training many AI models can require very specialized and expensive equipment which makes it inaccessible to most. And further, there are many labor issues potentially around the preparation of the data used to train these algorithms and there’s a general lack of diversity in the field as well. Lastly and very importantly, AI and the data that powers it also raises a lot of questions and potential issues around data security, data privacy, and so on.
To minimize the impact, the SECure framework proposes the four pillars that we mentioned earlier. And in plain English, they are meant to make compute resources more efficient, accessible, and cheaper so that a wider range of people can learn about and develop AI models, train and use models on devices instead of a centralized method like the cloud, for example, which reduces the risk of things like data breaches, potential privacy issues, and ultimately impacts people’s trust.
Give ownership of data to people so that they can decide how their data is used, when it’s used, and why it’s used, ultimately by either giving or taking away consent. Promote getting certification that ensures environmentally sensible and responsible AI. So another thing that’s brought up that’s really interesting is this idea of a single metric to measure sort of the performance of the model and also the considerations and trade-offs that should be taken into account.
So for example, normally when training a model often times people will think in terms of accuracy for performance. So say you’re creating a predictive model, you want your, you know, predictions to be almost 100% accurate if you can. Now, most of the time that’s impossible and certainly a model that only predicts well 40% of the time isn’t very good, but still, it’s this idea that you’re focusing on this one single metric.
But that metric doesn’t really take into account all these other things that we’ve talked about in this video around the impacts potentially on the environment, society, the carbon footprint, how much data and labor is required to create the data, how complex the models are and what the impacts of that complexity is and so on.
And so it’s a really interesting thing to consider you know, how do we kinda move more into a direction where we come up with a single metric that may take into account many more factors, like some of these factors we’re talking about today. The SECure framework is a really interesting approach and it’s great to see organizations like the Montreal AI Ethics Institute leading the charge. And they’re not alone.
Another organization called Responsible AI has developed this thing that they call Responsible AI Licenses, or RAIL for short. This is another sort of approach to deal with and handle these same sorts of sort of ethical issues, responsible use of AI, minimize the impact, and so on and so forth. So this is a really interesting area that’s very important and it’ll be interesting to see, you know, how things evolve with these initiatives and organizations over time.
reference – How will AI impact society and the environment?
Web enthusiast. Thinker. Evil coffeeaholic. Food specialist. Reader. Twitter fanatic. Music maven. AI and Machine Learning!