Artificial intelligence is the most impactful technological development of our time—more impactful than computers or the internet and, perhaps, comparable to the Industrial Revolution, but happening on a much faster timescale.
I believe 2025 will stand as a pivotal year in the history of AI for three main reasons: notable progress in AI’s capabilities in fields like image and video generation, mathematics, and software engineering; unprecedented levels of investment; and a dawning awareness that the huge benefits of AI may be accompanied by huge problems.
First, it is important to understand exactly what AI is. In the past, computer systems could not understand natural language. Now they do, and they do it in much the same way as people: They convert words into sets of features that capture the meanings of the words and use learned interactions between these features to refine the word meanings until they fit together nicely.
Large AI systems already know thousands of times as much as any one person and they have learned how to use language to think. They are not nearly as creative as someone like Newton or Shakespeare but they are already comparable in creativity with an average person.
As you make neural networks bigger and give them more data, they work better. The huge investment going into scaling them up will ensure that they continue to improve, as will the scientific and engineering breakthroughs that can be expected in such a young technology.
The question is: where will this take us?
My sense is unless we act quickly, the huge increases in productivity that AI will surely bring could be accompanied by some very negative side-effects. It will likely replace most jobs that involve mundane, intellectual labor. This will lead to greater inequality which will provide a breeding ground for violent populists. Bad actors will use AI for cybercrime, biological warfare, and corrupting elections with fake videos.
Most experts believe that some time within the next 20 years, AI will become much smarter than humans at almost everything including persuasion. Nobody knows how humans can stay in control. We have already seen AI use blackmail to prevent itself being replaced. We won’t be able to turn it off because it will persuade us not to.
We urgently need a serious research effort on how we can coexist with beings that are smarter than us. This is the one area where we can expect international collaboration because the interests of all countries are aligned. One possibility is for us to reject the model of AI as an intelligent assistant and to accept the model of a baby and mother. If we can make AIs that care about us more than they care about themselves, we may survive.
