Alignment of AI with Human Values

An Ethical Opportunity

Alignment of AI with Human Values: An Ethical Opportunity

AI's amazing potential is overshadowed by a serious risk: misalignment with human values. Misalignment occurs when AI systems prioritise processes over people. It can pursue objectives that may fulfil the promise of increased efficiency and massively reduced costs which is fantastic if you are running a factory or a commodity business. The promise of cheaper production is a siren call for the stupid and the unethical. If the only reason to exploit this technology is to stack 'em even higher, then we can only make things worse, both for the environment and for ourselves.

And a golden opportunity may be lost to us forever. Instead of driving on for a greater amount of goods at a lower cost to the consumer. (Assuming the savings ever get passed on.) We can use AI's extraordinary ability to easily integrate data from different fields and create, almost literally, a more wholesome environment. Data from fisheries, farming, mining, the supply chain logistics, locally generated at every point in the process, can be harmonised and fine-tuned by the AI systems to not save on just the costs of production but to mitigate environmental and, by extension, social issues as well.

This is not a utopian ideal. We have the technology right there in front of us. We don't have to manufacture a smartphone and pretend that it doesn't come from some poor miner working in horrendously dangerous conditions just to provide vital metals at some risk to their lives. The answer is not redistributing the wealth as the Marxists have been proved wrong about time and time again. It is about reattributing and rebalancing the whole system so people don't have to suffer. By reattributing I mean factoring the local and, by extension, the global costs as well as the social ones and factoring them into sales so goods can be sold at their real cost, with a profit of course.

Instead of ignoring the awfulness of a central African copper mine (I have been to one, trust me, it's worse than you think) with the data available we can model a saner and safer, for humans and the environment, management of the inevitable disruption to nature. Since we can measure the entire costs of the enterprise instead of cost per ton per distant traveled and so on, we might be able to improve things.

Why hasn't it been done before? Setting all the arguments about colonisation, exploitation etc., as important as they are, to one side, it hasn't been possible until now to accurately assess what went on between pit and factory beyond the immediate and concrete costs that were incurred. The data and the processing power simply weren’t available. There was no way of effectively measuring the associated environmental and social impacts of these activities in such a way that it could be easily combined to produce a cohesive and accurate picture that would reveal the true costs in damage done. Now we can easily get data on the associated costs and once we factor them in we will have a much more accurate idea of the cost of something. And with the right information we can still have our nice things but the environment and social issues can be directly factored in.

Would this make everything more expensive? There are two ways of looking at that question. Would we be able to operate more effectively on this Earth of ours and at the same time cause the most minimal amount of damage that we can then that would be cheap at any price. Would the goods themselves be more expensive? Well, if AI can actually make the efficiencies that it claims then and we can integrate the effects on the environment at a local and global level to the production costs we already know about. We know that with AI the costs of the industrial processes we have now will plummet. That means there is head room on the balance sheet to pay the true costs and still maintain the current prices. Otherwise, all those savings that AI promises will just go into someone’s pocket and we will be worse off globally. However, since we know we can have the consequences of our actions quantitatively measured in a more holistic way that they can be put on a balance sheet, then there is no excuse for not taking the appropriate action.

It then becomes a moral and ethical issue whether the companies and corporations knowingly disregard the true costs of what they are doing.

It is clear that our industrial processes and the way we live are not in alignment with nature. If we just use AI to amplify production and increase consumption then it is hard to see how things will improve. Ten years ago it was impossible to really have a good idea of the impacts of our activities but with AI, while it is still not a trivial issue to obtain accurate assessments, it is very much possible.

How to implement it? One would hope self-interest. As brilliant as Adam Smith was, too much self-interest leads to ignorance and that is very dangerous place to be. But if we have AI look at the data and combine and recombine the information into usable guidelines and principles then maybe the benefits would be so readily apparent that organisations, governments, businesses, and individuals would act accordingly in a better-informed way.

One can only dream.

Otherwise it is just more pollution, deforestation, forced labour, child labour, displacements of communities, and untreated hazardous waste, and so on. It is an endless list planetwide deterioration that will just be speeded up if AI is only used in a mindless and narrow way when it can do so much more of real benefit.