AI – the good side and the bad side
Nobody can say for sure where the AI world is heading. We can hope that the power of modern technology will let us achieve a great deal of good. In fact, there is no reason to put it off until tomorrow – we can act today by spotting and removing potential threats and backing positive, worthwhile initiatives.
A bold attempt at guessing what lies ahead for artificial intelligence and humanity.
To readers in the distant future
We are in the spring of 2023…
Some people claim our civilization is in good shape and that a remarkable era of prosperity awaits us. Others foresee an unavoidable collapse in a dystopian future not far away. And some hold that nothing really changes: wars and plagues take turns with better, more peaceful periods.
I struggle to side with any of these views. As an Earth and environmental scientist, I am aware that we have made a huge mess of things. Undoing human-made climate change, reversing species extinction and easing the pressure we place on nature is hard, if not impossible. As a machine learning engineer, however, I follow developments in artificial intelligence with bated breath. I am convinced that some future AI-enhanced technology will let us undo the harm we have caused.
AI can be used in many ways to build a better future or, conversely, to do something awful. I have no doubt that people are working on both kinds of solutions right now. In the end, there is no shortage of recipes for using AI for good and for evil. Pick up a few sci-fi books and let your imagination roam.
The challenge of forecasting what the future holds has always tempted me, and that is where this article comes from. In it, I try to work out how artificial intelligence will affect people and nature in the years to come. My main reason for writing it is to see how my outlook shifts over the next few years. A disclaimer: this is a set of loose reflections rather than a serious read. Still, I encourage you to share your own thoughts and ideas – above all, to show me where I’m wrong.
You, reader from a not-so-distant future, will either praise my spot-on predictions or enjoy seeing how mistaken I was!
AI for bad
Let’s begin with the hard subject of AI causing harm to future humanity. For those of us who are AI experts, actively countering this will be crucial. The earlier we begin, the better.
Fake science
What will help us tackle problems on a global scale is not faith or good intentions, but high-quality scientific research – and that quality is about to decline.
Don’t misunderstand me. I have hopped on the generative AI and large language models (LLM) hype train myself, but the effect this branch of AI has on our everyday lives frightens me. As far as science is concerned, I’m not troubled by scientific articles being churned out en masse, or by authors starting to sound alike because they all use the same writing-assist software.
What really worries me is data. Data set augmentation has been a machine learning technique for a long time, and I’m a fan of it myself. But what happens once we are able to generate entire datasets that look realistic yet do not reflect any real natural phenomenon?
In every scientific discipline, collecting data is an extremely complex undertaking. It is hugely expensive and frequently drags on for a long time. It would be easy to skip the hassle of obtaining data by simply asking an LLM the right questions. The answers won’t be real, but they will seem realistic.
Will we cut corners to keep pace with the publishing cycle? Before long we will witness a small scientific revolution in which, besides innovation, the key value will be rigorously repeating experiments, not just traditional peer review.
To what extent will generative AI change how we do research?
Praise the machine god
We put a great deal of trust in AI. Machine learning and deep learning solutions do indeed help us enormously in automating all sorts of tedious processes, and ML/DL solutions frequently take part in decision-making as well.
AI is becoming an everyday thing, and we grow ever more familiar with it. We begin to believe in it – much as we come to trust a person simply because we have known them for longer and longer. Yet it remains just a tool, and like anything else it may let us down when we least expect it.
Meanwhile, AI is getting harder and harder to comprehend. At some point we won’t be able to keep pace with it, and all we’ll be able to do is drift along with the technology. I wonder how fast we will come to terms with living in a world governed by algorithms we don’t understand. How quickly will we accept being harmed by rules we can no longer control? Perhaps this has already happened.
At what point will AI knowledge become too much for us?
An unhappy accident
My worry is that some AI-enhanced technology will appear whose effects we are unable to control or even notice. I don’t mean a robot rebellion – let’s be serious. What concerns me more are mechanisms that might substantially and negatively steer the way society develops.
I realize this is hard to picture. Imagine it starts as a useful solution which, after running for a long time, creates certain social imbalances. It could be a citizen rating system, a clever algorithm trading on the financial market, or an app delivering entertaining content to us online. If such systems are left untended and unmonitored for years, their effects may be hard for us to notice. The rating system could bring a new form of discrimination, the financial algorithm could put particular individuals at a disadvantage, and the content recommendation app could make us less sensitive or dumber.
Could AI development hold an unpleasant surprise for us somewhere down the road?
Power consumption
AI could give us an astonishing degree of control over even the most intricate processes. In many demanding tasks it already beats humans, and with each year there will be ever more AI-based solutions.
This comes at a price, though. Today, ML model training and inference demand substantial computing resources – and not only to serve user traffic. The real issue is that models have to be retrained over and over. That is essential for the solution to function properly and withstand data drift. On top of that, models keep growing in complexity, and working with them costs more than time. Above all, it takes electricity.
This is already a serious problem today, and it can only get worse. We will find ourselves spending energy on ever more comforts that benefit only us, humans. We’ll end up with excellent air conditioning, cryptocurrencies and a multitude of AI models in a world whose natural environment has been ruined.
Will we have enough electricity to power all our ideas?
Obsolescence
It’s well known that every now and then some professions vanish and new ones emerge. My children are probably being trained for jobs I haven’t even thought of. What I’m unsure about is whether this is happening too quickly. AI will push automation even further. In the worst case, we will hand everything over to well-designed algorithms and remove the human from the equation altogether.
I’m not overly concerned about people who use new media, work in STEM fields and so on. They won’t run out of work – just the opposite. I would focus instead on those who, for one reason or another, never had the opportunity to get enough education to take part in the change we are now seeing. Along the way, many tough decisions will have to be taken so as not to widen the social inequality that already exists.
Are we prepared to make the education system more flexible and resilient?
Loneliness
This is what troubles me most. With AI backing us up on every front, we won’t need anyone else’s help with our everyday tasks. In the end, we’ll be left alone with our computers. They will do our shopping, assist us at work, do our homework, play games with us and suggest films and music. We’ll also be able to chat with them about all kinds of things and spend time together. As it gets better, it will take the place of our human companions. Automation everywhere will make sure every need of ours is satisfied – and we will be alone. Creepy and unsettling.
Will our society end up disconnected?
AI for good
Up to now, the picture has been gloomy and grim. Time for some good news.
Research boost
Luckily, we don’t need to wait for this one. AI is already helping uncover new things across many science and engineering disciplines. I’m particularly excited about all the discoveries awaiting us in the life sciences: new kinds of treatments that will let us live happier, healthier lives, or substances that will prevent pollution or even reverse it. Thanks to a breakthrough in AI, we can come up with more new ideas and test them more quickly. Here, AI will serve future researchers as a tool for carrying out more complex experiments.
Which industries and scientific fields stand to gain most from adopting AI?
Quick response
One big strength of AI is that it makes decisions efficiently and can work without a break – something we humans cannot manage. AI-based systems can also handle far more complex data than human experts. This opens up amazing possibilities for observing all kinds of phenomena and responding to them; weather monitoring, earthquake prediction and disease tracking are just a few examples.
Going forward, we will enhance the mechanisms we use to monitor our planet. I’m looking forward to the creation of a complete digital twin of the Earth that will let us follow and forecast what happens on our planet.
How quickly can we deliver dependable model predictions to societies?
Accessibility
We are entering an era of quicker access to knowledge. Machine learning models keep improving at locating information and presenting it in a way that suits us. I can’t wait to see how AI will support learning and shape the education system. I’m also certain that we’ll soon be flooded with all sorts of AI assistants. Some of them will likely be worthwhile ideas for supporting people with disabilities or assisting in the delivery of medical treatment.
In what ways will we use AI to help others?
Sustainability and growth
My biggest expectations and hopes concern the use of AI in sustainable development. Applying AI across the various systems that support modern societies could help us arrange particular areas of human activity so that they run optimally. AI will be capable of handling the great complexity of the many sub-processes involved in using natural resources and distributing them fairly. I have in mind AI systems that monitor and manage the condition of water and its industrial use, AI techniques in precision agriculture that reduce the amount of fertilizers applied, better transport of goods and services, and, lastly, support for environmental and wildlife conservation efforts. I also hope these systems will be integrated with one another, so that rather than operating in isolation they reach a certain degree of synergy.
How can we use AI to turn Earth into a wonderful place to live?
Conclusion
Nobody can say for sure where the AI world is heading. We can hope that the power of modern technology will let us achieve a great deal of good. In fact, there is no reason to put it off until tomorrow – we can act today by spotting and removing potential threats and backing positive, worthwhile initiatives.
Any thoughts about how AI will shape our future are welcome. Links to any exciting projects supporting the positive development of AI are welcome too.