Hacker News Reader

Comment by hyperpape | Hacker News Reader

I assume he means something like "computers will reach a point where they can take non-code descriptions written by a human and implement code based off of them." However, I agree that this is not a plausible thought

1. I doubt that this is coming that soon

2. If it does arrive, it may well be with AI so sophisticated that the human has no role even for the "creative" parts of the process. A system that can handle non-formal requirements and implement code sounds close to general purpose AI--the kind that just replaces humanity.

This article seems pretty pointless.

View on HN

> AI so sophisticated that the human has no role even for the "creative" parts

That is a crucial point that is often missed by people outside of the industry. There is nothing magical about activities that don't require the assistance of a computer today. The very principle of computability applies equally to any activity, not just mathematics.

In a way, art is the product of a very complex Turing machine that relies on physics and biology to compute it.

We are already seeing impressive pieces of art coming out of silicon computers, DeepArt being an example: https://www.facebook.com/deepart.io/.

It is, to me, extremely likely that within 20 years, we will have generated novels compete on the market with best-sellers.

The creative process will increasingly rely on mastering and combining more and more exotic techniques, and that may require a deep understanding of the new tools at hand.

That isn't 'art' as an expression of emotion (if that's how you want to define it). Those are interesting images generated from combining other interesting images.

> It is, to me, extremely likely that within 20 years, we will have generated novels compete on the market with best-sellers.

Where are we with that now?

> It is, to me, extremely likely that within 20 years, we will have generated novels compete on the market with best-sellers.

I'd say it's going to be a lot sooner, given that they're already doing well in writing contests[0].

[0] http://www.digitaltrends.com/cool-tech/japanese-ai-writes-no...

There's another way of reading it, which is that code generation will become sufficiently advanced that a huge number of the high-paying software engineering jobs that exist today become things that people can do with very little training.

It's not so much that you can take informal descriptions and turn them into software, but more that the software can be built by anyone with a few months of training, thus driving down salaries to a point where mere coding is no longer a white collar job.

That doesn't seem impossible, but the reason for its unliklihood is far more subtle.

This is always an argument in tech. But as has been stated by folks before the easier coding gets, the more ambitious and complex the tasks we attempt to accomplish with it. So the difficulty at the top end never gets any easier just because the bottom end does.

Exactly right. And that's more obvious on the web than anywhere else.

My point was just that seeing this is a lot more subtle -- it's not a purely technical observation, but also contains a social/economic component.

>So the difficulty at the top end never gets any easier just because the bottom end does.

Personally, I think the end of construction came after the lever and fulcrum were invented.

After humans didn't have to exert effort to build things, construction became a thing of the past...

Oh wait.

Reply on news.ycombinator.com