It happened when the internet came up. It will happen with all the disruptions that have to do with information, a great element of historical power, if not the most: change is scary, employment is transformed.
With artificial intelligence (AI) the complexity is greater than with the internet. AI algorithms require training. There are two approaches, supervised and unsupervised learning. In the first of these, humans provide the computer with a set of images with labels indicating relevant information. For example, if they are images of dogs and cats, the tags can indicate which images correspond to dogs and which correspond to cats. Instead, unsupervised learning allows the algorithm to learn patterns and structures in the data on its own, without predefined guidance, in order to discover hidden patterns in the data.
ChatGPT, the best known generative AI today, trains the algorithm simply to predict the next word in the text that is used to train it. It has surprised everyone, including its designers. Understands and produces human-like text. It will soon have competition because other big tech companies are launching their own tools (like Google's Bard).
AI applications perform many tasks better than us. We share the infosphere with artificial agents that are increasingly intelligent, autonomous and even social. Applications, web bots, algorithms and software of all kinds, robots, driverless cars, smart watches and other devices. New jobs will also appear, because new tasks will be developed between automated services, websites or AI applications.
The prospect that AI-powered machines will overtake humanity, take over the world, and who knows what else they will do to us, a recurring theme in dystopian science fiction, is easy to dismiss: humans are still, today, having the control. But many AI experts take the doomsday prospect seriously. Consider the distinction between narrow AI and artificial general intelligence (AGI). Narrow AI can operate in only one or a few domains at a time, so while it can outperform humans on select tasks, it is still under human control. The AGI aspires to display human intelligence in general. That is why there is so much talk about regulation. The first initiative to govern artificial intelligence was announced by Singapore – a leading country in AI – at the last Davos Forum. Other regulatory interventions such as Italy have followed. It won't be the last.
What can humans do better than AI? Tasks that require a high degree of creativity, critical judgment, empathy and cultural understanding. Deeply human skills that are difficult, if not impossible, to automate. The future must be a world where human capacities and AI must be reconciled to improve the productivity and well-being of all.
