2020 — and AI has still not taken over.
Sure. The last decade has brought us a lot of dramatic progress in the field of machine-learning. Computer programs can now learn to classify cats and dogs (and thousands of other categories), can translate text better than most of us (while still making some funny mistakes every kid would do better), and can drive cars (outside inner-city areas and even than still occasionally crashing in a way that by humans is perceived as unpredictable).
And yet, compared to what most humans (and most science-fiction writers) intuitively understand under “intelligence”, all of this is still only mimicking intelligent behavior. Funny enough tech believers keep repeating the same naive formula of predicting the future of AI by simply interpolating based on only vaguely related (or entirely unrelated) observations. Such as: computers could do cognitive task X1 in year Y1 and and task X2 in year Y2, so it will reach human-level in year… 2020.
One of the most prominent examples from the last 1–2 decades was from Ray Kurzweil (e.g. in his book “The Singularity is Near”). He’s essentially basing his predictions on the number of calculations and for quite some time 2020 seemed to be the year where human levels of intelligence would be met. Of course, as with any good divination, most of those predictions are kept vague enough to leave room for speculation. In addition the predicted years get updated as time progresses. Quite convenient.
In times of deep learning, people sometimes take the number of artificial neurons as a measure how close we are to human-level intelligence. Which seems equally naive. As if an extremely complex, living neuronal cell can simply be modeled by an artificial neuron which is nothing more than multiplying and summing numbers…
So, I would easily dare to make the prediction that no human-like intelligence will come from current machine-learning approaches. Not in 2020. And not this decade. But no doubt that the field will continue to grow and bring a lot of exciting (and often not predicted) new abilities, use-cases and debates. And about the human-level intelligence? Well, luckily there’s plenty of human intelligence around, we just have to make sure we use it right this decade.
Speaking of which: Happy New Year to all of you!
