Hacker News Reader

Comment by JabavuAdams | Hacker News Reader

Marketing works. This is an example of marketing based on authenticity.

Regarding perfection: machines are much more capable of perfection re: manufacturing than humans. Where I think we get confused is that in the past only corporations had the capital to mass-produce, and they have an incentive to "race to the bottom" in terms of production cost. Although it's not the goal, usually quality suffers as a result. So, we associate mass production with poor quality.

I've been spending a lot of time studying historical armour manufacturing and also rocketry. It's clear to me that the goods manufactured by passionate, talented, (and largely under-capitalized) craftspeople are inferior to what could be made by automation. It's just that they're often superior to what is actually made by cost-cutting automation.

Note: I'm all for people "geeking out" about their interests. As we automate more work, people will need more hobbies.

View on HN

This raises the very interesting question, in my mind:

What is to be the proper center of the economy, the individual human or social equivalent (say, household unit)? Or the machine or social equivalent (corporation, institution, government)?

Personally, I'd rather live in a place where humans and social equivalents were the center rather than where we were at best pieces of a social machine. So maybe there is something to the social superiority of supporting small businesses as much as we humanly can.

> It's clear to me that the goods manufactured by passionate, talented, (and largely under-capitalized) craftspeople are inferior to what could be made by automation.

That is not true for all goods....I cannot forsee a computer making a better painting than a good human artist in the near future....In fact I cant even forsee a computer understanding what a good painting actually is in the near future...This holds true for music too....In general works of art are (and will be in the near future) better crafted by humans.

But can you please point to me some good books on this subject.I find this topic very interesting.

I can foresee software being able to discern good paintings from bad based purely on supervised learning from examples, with today's technology. I.e. humans label paintings as good or not good and the machine learns to extrapolate from this. Is there a consensus among humans about what makes a painting good, though? Perhaps we could settle on well-made. I.e. I may not like an artifact but I can agree that it's well made.

Creating good paintings seems like it should be doable, but take a couple/few decades.

Music should be a lot easier, as music is a much simpler signal than images. I'll probably start with music.

Making maille: http://www.mailleartisans.org/articles/articledisplay.cgi?ke...

Rolling plate armour edges: http://www.ageofarmour.com/education/armour_rolled_edges1.ht...

Armour reading list: http://www.ageofarmour.com/instock/books.html

Machine Learning: http://www.ml-class.org/

AI Composer: http://singularityhub.com/2009/10/09/music-created-by-learni...

I can foresee software being able to discern good paintings from bad based purely on supervised learning from examples, with today's technology.

Unless we manage strong AI, I expect this effort to get stuck in arguments over feature selection.

Is this so different from the human experience? :P

That's not art! Yes it is! ...

Noisy data labeling presents an additional problem, yes, but I was under the impression that there existed techniques for mitigating that problem.

No, computers aren't very good at creating unique things without explicit instruction on how to do so. But they are very good at replicating things to an exacting standard. In the case of a painting, a computer could probably make an exact replica of an already existing painting far better than any human artist could.

In the case of knives, a computer could make a copy so exacting that it would be indiscernible from any of the knives this guy produced if given the instruction and materials to do so. The only thing that would give the computer made knives away would be that they would be too perfect (if perfect is measured by how close it falls to the original), while the knives produced by hand will all have inconsistencies because they were made by something that isn't precise to a micrometer.

Speaking of which, that's part of what I think drives the hand-made movement: you're buying something unique. Even if that particular artist made a hundred of them, there are no two exactly alike.

>No, computers aren't very good at creating unique things without explicit instruction on how to do so. But they are very good at replicating things to an exacting standard.

Well there are heuristic approaches...so if a computer could actually determine what a good painting is.....it can keep rearranging the pixels (in a divide and conquer sort of way) to get to a good painting in a finite amount of time.

So I think the problem of making a good painting can be reduced to the problem of determining what a good painting actually is.Now I have read somewhere (dont exactly remeber how) that this is related to the halting problem.

Reply on news.ycombinator.com