People think it's different this time because the new machines are more generally capable than the specific purpose machines of the past.

It's not "the new machine can gin cotton" or "the new machine can weave cloth", it's "the new machine can handle an extremely wide range of problems, and we're moving from having to teach it each problem individually to having it teach itself."

This time really is different and it's crucial everyone realize why as soon as possible.

> It's not "the new machine can gin cotton" or "the new machine can weave cloth", it's "the new machine can handle an extremely wide range of problems, and we're moving from having to teach it each problem individually to having it teach itself."

It's funny you mention this because in the past few days on HN there have been a number of articles whose general point was that the machine learning stuff presently in use is well tailored to very narrow, specific problems, and that the AI tech that would tackle a broad range of problems is still very elusive.

This new machine that tackles a very wide range of problems without massive programming investment is something I've been wanting for 20 years, and I still don't see anyone with plans to bring it to market.

> the machine learning stuff presently in use is well tailored to very narrow, specific problems

It's more accurate to say the machine learning in use now works on narrow classes of problems. But even that is a giant leap forward from teaching the machine to solve each problem individually.

Teaching a car to drive very specific routes is one thing (and has been effectively done for decades now). Teaching a car to drive entire classes of routes is something else, and where we are now. In a decade or so cars will drive all classes of routes better than humans, and that will be something else again.

A thing that gets lost often in these discussions is that while human populations are fairly good at adjusting, individual humans often are not. Sometimes the end result to a tech shift eating your job is that you will never have a comparably good job, regardless of what you try. And maybe your kids will not either. This is a hard truth.

If it happens to you and enough of your peers in one area, it can devastate the local economy. None of the solutions for this are easy, and history is full of examples of it.

> while human populations are fairly good at adjusting, individual humans often are not

(...)

> None of the solutions for this are easy, and history is full of examples of it.

This is demonstrably true in the past, but it doesn't have to be that way in the future. But if you always do what you've always done, you'll always get what you've always gotten.

Solutions are all about the next generation. If a human did adapt to a change, it was because of training, mindset, and education they received before the change. Humanity has to get ahead of these things, and be prepared for changes, rather than trying to figure out what to do when they happen, when options are already drastically limited.

Probably the cheapest, best, most flexible way to do this is to educate them better in the first place. It's probably also the most efficient, since even if the world doesn't change, the education demonstrably improves their outcomes and has positive return.

Many of the "new things to be done" are done by machines from the very beginning now. No human or combination of humans can do what Google does.

And yet Google employs a lot of people to do what Google does.

No human or combination of humans can turn iron ore into stainless steel tubing. Yet a whole lot of people, acting in concert, can build and operate a bunch of terrifically complex machines which will do just that. The end result is an entity that you can draw a box around, which takes in that human labor plus iron ore and coal and minerals and produces steel tubing (and slag and acid rain). And it'll continue to do that, optimizing itself, as long as the market price of the outputs exceeds the market price of the inputs.

Google is similar (without the air pollution, though it has its own externalities). It's a bunch of people building and operating a bunch of terrifically complex machines which organize information, in such a way that the outputs have more value than the inputs inclusive of labor. It's a pretty nice business model, particularly when the inputs are basically free (because of the Internet).

But I don't see any reason to be especially threatened by Google, or at least there's no reason to be more threatened by Google, than the steel plant. Google can and probably will optimize itself to reduce labor costs, just as steel mills have optimized themselves; a modern steel mill can produce a hell of a lot of steel without very many employees, which isn't good if you're a steel mill worker. Google in 20 years may not employ very many people to maintain its search products, either. It's something to be aware of if you work in that industry, but it's not especially unique.

Most of Google's 57,000 headcount is ad sales reps. Before search personalization, the core search team was under 100 people. It's probably larger now, but nowhere near the size of the sales organization.

What's the relevance of the ratio between engineers and sales people?

Seems the point you're making is that it takes 57,000+ people to make google run. You'd think they'd want to automate those sales reps, except that's human work and they can't. Which is the point the previous poster was trying to make.

What aspect of the sales rep job is unable to be automated in the next couple of decades?

Sales reps mostly talk to people and convince them to buy things. Relationships and human interaction are paramount.

I'm unwilling to speculate a couple of decades out when I don't think anybody has any clue what's going to happen in the next 5 years, forget decades.

But what I will say is this - there is no plausible way to automate human sales relationships in the foreseeable future. I don't even know of companies who are even seriously trying right now.

If you take the unbridled enthusiastic position that everything can be automated in 20 years, sure, all those sales guys are toast. I just don't see a pragmatic basis for taking that perspective. Tech gets better and that's great, but what you're talking about would be some Manhattan Project stuff. People have been talking about the potential of those advances since the 1960s. Wake me up when someone announces a date when they'll ship a product.

You don't have to replace sales relationships, you just eliminate them.

My life has become so much better with the advent of Amazon Prime and related services. I go to a website, find what I want, click a button, and whatever I want shows up on my doorstep two days later. It's fantastic, and nobody is trying to mislead me, or upsell me, or convince me to add on some shit that I really don't want. More things should be this easy.