Werner Herzog in conversation with Paul Cronin (2014) • Hacker News Board

Re: Werner Herzog in conversation with Paul Cronin (2014)

#31

Post by atoav » Sun, May 31, 2026, 5:04 PM UTC

enos_feedler wrote: post #16

Buried in this interview is a great insight about the limits of AI: "Just viewing all of Treadwell’s material would have taken me at least ten days, but I had four assistants who went through everything, melting it down to about twelve hours of footage. I gave them precise instructions about what I was looking for, but sometimes scrutinised what they had put aside and found extraordinary moments they had dismissed. T…

The problem I see is that LLMs have the tendency to pull everything towards the mean of whatever you ask it to. But the really, really good stuff is often totally out there on the fringes.

You can get some good results with really precise instructions, but this requires someone who has already thought a lot about what they are after.

Re: Werner Herzog in conversation with Paul Cronin (2014)

#32

Post by arwhatever » Sun, May 31, 2026, 8:02 PM UTC

gobdovan wrote: post #8

arwhatever wrote: post #5

Earlier quoted context omitted.

I always enjoy hearing from him because he’s so unorthodox and I never have any idea the approach he’s going to take when giving an interview or answering a question. And I always feel the need to point out that Grizzly Man was a truly good movie. I’d heard about it for years and based on the premise expected to have a low brow appeal, something for dumb people to feel superior to someone. But no, it was a respectful…

If you want to see the extent of his respectfulness and depth, with his courage for self-analysis added on top, My Best Fiend would be my choice. It's a documentary about his relationship with Klaus Kinski, who was a pretty unhinged actor. It's fascinating to watch Herzog hold his own with Kinski, and to observe the strange mental space Herzog is in, somewhere between crystal-clear vision and complete madness.

Watched it. It reads interesting indeed. Makes me suspect that, at least during the Fitzcarraldo period, Herzog might’ve thought misery was valuable or even necessary to good filmmaking - misery in the form of stunts and geographic setting, and misery in the form of lead actor’s personality.

Best part of My Best Fiend was Herzog’s own thoughts (soliloquy?) on nature and the jungle itself. Those bits were some pure solid gold Herzog.

Re: Werner Herzog in conversation with Paul Cronin (2014)

#33

Post by AngryData » Sun, May 31, 2026, 11:42 PM UTC

enos_feedler wrote: post #16

Buried in this interview is a great insight about the limits of AI: "Just viewing all of Treadwell’s material would have taken me at least ten days, but I had four assistants who went through everything, melting it down to about twelve hours of footage. I gave them precise instructions about what I was looking for, but sometimes scrutinised what they had put aside and found extraordinary moments they had dismissed. T…

Yeah I think LLM story lines will probably have the exact same problem as story lines made in commitee. They end up as average and predictable results that might not be bad but almost never become something great.

Subverting expectations by itself can make or break a movie, but it takes real insight to not just be "lol random spork, im edgy" or just baffling, as we have seen in many soon forgotten works.

Re: Werner Herzog in conversation with Paul Cronin (2014)

#34

Post by enos_feedler » Mon, Jun 01, 2026, 12:01 AM UTC

atoav wrote: post #31

enos_feedler wrote: post #16

Buried in this interview is a great insight about the limits of AI: "Just viewing all of Treadwell’s material would have taken me at least ten days, but I had four assistants who went through everything, melting it down to about twelve hours of footage. I gave them precise instructions about what I was looking for, but sometimes scrutinised what they had put aside and found extraordinary moments they had dismissed. T…

The problem I see is that LLMs have the tendency to pull everything towards the mean of whatever you ask it to. But the really, really good stuff is often totally out there on the fringes. You can get some good results with really precise instructions, but this requires someone who has already thought a lot about what they are after.

Coming up with precise instruction about what we are after isn't always possible. In many cases we only know what we are after once we see it. And if you can't express what you are after into tokens you can't outsource that choice to anyone else except your own direct experience of the process.