Hacker News Reader

Comment by kobeya | Hacker News Reader

"Bhautik Joshi posted a paper to arXiv on the topic of some consulting work he did for a short film, and listed the producer and actor as co-authors to garner publicity."

View on HN

Actually, she's not an actor in the (short) film in question, she is the director as stated in TFA. The ML technique is applied to achieve a specific effect the director wanted to achieve.

Did she submit PRs to TensorFlow? Probably not. Did she contribute to the paper? Most definitely, I think we can even go so far as to say it's plausible she fulfills the Vancouver criteria:

Substantial contributions to the conception or design of the work; or (..)

AND

Drafting the work or revising it critically for important intellectual content;

AND

Final approval of the version to be published;

AND

Agreement to be accountable for all aspects of the work

Assuming she at some point revised the paper critically for content, which is plausible, there is no denying she's on the author list.

It is plausible that she had some input to the paper beyond requesting the work, and artistic feedback. But it's definitely more likely that she didn't.

Would it be safe to say your opinion of Steve Jobs is similar?

It is plausible that Steve Jobs had some input to the iPod beyond requesting the work, and artistic feedback. But it's definitely more likely that he didn't.

Steve Jobs didn't co-author any papers on capacitive touchscreen technology or whatever.

But Steve Jobs got the patent: https://www.google.com/patents/US7479949

>But it's definitely more likely that she didn't.

Why is it "definitely more likely"?

The vast majority of programmers are men, and acting and AI research are pretty polar opposite activities. Most actors/actresses aren't tech geeks.

I guess I'll get downvoted for this too because it sounds anti-women or something, but it is the unfortunate truth.

Duh. She's a girl and everyone knows girls can't program. </sarcasm>

Perhaps more charitably, because she has no known history of having done so, and her talents in the acting profession are not generally considered directly transferable.

The point was that there is no requirement beyond conceiving of the work and intellectual (not just spelling and grammar) feedback for her to be considered a coauthor.

It's not very different from a professor having a idea, giving it to a grad student to implement, run and write the paper, and then the professor gives feedback on the paper before submitting.

Your professor can fully understand your work, could have done it herself if she had the time, and can fully check your work for any flaws or limitations. Your non-technical manager cannot fully understand your work, do it herself or check it.

If "Substantial contributions to the conception ... of the work" is enough to be a co-author, she should be a co-author. But that means that if Pointy-Haired-Boss tells Dilbert to "make the Skype call to Mars less laggy" and Dilbert invents FTL communication, PHB gets to be co-author of the paper (and resulting Nobel prize?). That seems too generous to PHB.

That was my initial reaction, but if you look at section 2, both paragraphs could plausibly have been written by Stewart and/or Shapiro; all the images are authored by Stewart as the director.

Trying to apply a technology is enough to publish ?

What constitutes a publishable paper depends on the subfield and journal/conference.

There are many computer science papers discussing the usability, implementation, application of tech ologies.

Definitely. And I wouldn't be surprised to see more methods papers from movies going forward.

If you can write something useful about the results or the things you discovered or developed in the process, sure. (Not in every venue of course, but in many)

I don't know enough about the specific subfields involved here to guess where and if this specific example would likely pass.

There are basically no barriers to "publication" on arXiv.

No, they actually did write parts of the paper.

Leaving them out as co-authors would be misconduct.

Source?

The paper is about how to iteratively tweak your rendering pipeline to achieve a certain look. It makes sense that the director of the movie would be involved in that process.

If I tell someone to blur a background during postprocessing, do I also get a coauthor credit on a paper about gaussian photo manipulation?

If the particular blur in question has a unique characteristic, you should push for it, sure.

So, you didn't read the paper, huh?

I did. They describe using a Keras implementation, written by someone else, of style transfer, which was invented by someone else, to achieve a particular artistic effect. They use a pre-trained model. They don't even have to buy any hardware, since they used AWS GPU instances. It's a case study, nothing new has been accomplished here.

Do I also get a Verge writeup if I publish a PDF to arxiv about how I compiled and installed tensorflow? The only reason we're talking about this paper is because the name of a Hollywood actress is on it, and because it has the AI buzzword.

Look at the title: "Kristen Stewart co-authors AI paper". 113 points.

The paper itself was submitted earlier, under its actual title, which does not mention Kristen Stewart: https://news.ycombinator.com/item?id=13443107 6 points! The paper is not very interesting. The only thing that makes it newsworthy is the coauthor.

Is any arXiv preprint more untrustworthy than any other or something [I mean, it's not viXra...]?

I can't help but think it would be degrading to have your name on a paper and it be deemed too advanced to be your responsibility..

You could ask the first author directly, I suppose; https://twitter.com/bhautikj

My question was directed at "salem" who stated "No, they actually did write parts of the paper.". That statement seemed like it should have reasoning behind it...

My source?

I would ask what is the source of that quote? The article no longer contains it.

Reply on news.ycombinator.com