> "Almost overnight, the quality of [Google's] translations skyrocketed." In certain languages, and still dependent on the human brain to reparse the not-quite-gibberish into something that makes sense.

But it happened replacing a system carefully crafted for 10 years, for another one that barely needed a couple of months to get better than the old one. And it can be made better than google's, today: https://www.deepl.com/translator

> "Driverless cars soon" - and flying ones, too, "in production in a few years" for a century now.

We had driverless car prototypes for a century? Wow. /s

- Fair point. Conflating machine learning with strong AI (as evidenced on the machine translation example) still dubious.

- We had flying cars promised for a century, "any year now." Very similar status to "level 5 autonomous driving any moment now, don't look at the edge cases."

"Promised" but without any workable and remotely affordable prototype. It may happen soon, though: https://lilium.com/

I see both hand in hand. I don't think it will take off (metaphorically) without mandatory self driving AI near populations.

"It may happen soon" is the mantra of vaporware, QED ;)

Yes, it is indeed. However I base my conclusions on existing prototypes and their cost.

Truck drivers are held up as the classic example of a job that will be decimated very soon. A million jobs, wiped out, read the fear mongering headlines.

The hysteria is remarkable. Ten Million plus jobs are created and destroyed every single year in the US. We're facing at least a 8-12 million person jobs gap, in terms of not having enough labor, over the next ten years alone (with 6.5 million unfilled full-time jobs right now and the U6 rate pushing down toward two decade lows). And that's assuming modest average economic growth of perhaps 1.5% annually (2.5-3% and we'll have an extreme labor shortage).

To amplify the actual context problem, if we had a perfect solution for automating away truck driving jobs today, it would take decades to fully deploy it. In reality, we're still a decade away from meaningful scale commercial testing.

Yeah I'm a little bit dubious about this article too, the author appears to have absolutely no training or experience related to AI and it doesn't look like this article is based on any actual research.