Montreal City Weblog

Kate 10:09 on 2023-04-02 Permalink Reply  

• Ephraim 10:47 on 2023-04-02 Permalink

  Midjourney…. all generated photos. Weird. They don’t look genuine Montreal.

• Kate 11:21 on 2023-04-02 Permalink

  I noticed it right away in the first photo – looked more like Brooklyn. But I kept reading. Thing is, we’ve got to start listening to those doubts now when photos look off, and raise questions if it’s anything serious.

• DeWolf 11:33 on 2023-04-02 Permalink

  I’ve dabbled in trying to create accurate-looking Montreal scenes in Midjourney and it fails every time. All of the images in La Presse that are supposed to be from the streets of Montreal have a kind of generic Northeast US vibe. The architecture is wrong, the trees are wrong, the street furniture is wrong. Kind of surprising given that there’s no shortage of Montreal photos online.

• MarcG 11:34 on 2023-04-02 Permalink

  The face of the house on fire isn’t quite right, either.

• Kevin 11:38 on 2023-04-02 Permalink

  3 years ago we had hyper-realistic video filters that made an old balding man look like a supermodel.
  Ten years ago we had video of an eagle snatching a child on Mount Royal.
  50 years ago a shaky camera operator filmed ‘Bigfoot’.
  100 years ago someone created the Cottingley Fairies

  This kind of fakery is as old as photography — it’s just that now it’s easy enough, cheap enough, and realistic enough that anyone can do it.

  So seeing is no longer believing, and unless you’ve got a person – and some way to verify it’s a person – willing to testify that something is legitimate, we can’t trust it.

• Robert H 11:52 on 2023-04-02 Permalink

  That touching story about the dog, I must admit I bought it. I’ve been had, I’ve been played. I’m a sucker for cats and dogs and my impulse toward skepticism collapsed like a house made of plasterboard. I suppose the first clue should have been “Michel Tremblay.” That name alone sounds too much like an archetype, unless we’re talking specifically about the playwright. Even the old axiom about if it seems too good to be true (Soya! mon héros!) didn’t raise a red flag. I’m more impressed by what ChatGPT can do with text than what Midjourney can do with images, I agree with you all that those Montreal photos look too generic. But just think of the applications. I’ve seen the future, and it’s a Black Mirror episode populated by deepfakes.

• MarcG 11:57 on 2023-04-02 Permalink

  Is it a coincidence that this technology is developing at the same time as nefarious right wing groups are waging war on public trust in society and its institutions? Seems to align with, and propel forward, the world of alternative facts.

• Kevin 14:22 on 2023-04-02 Permalink

  MarcG
  No, it’s not that coincidental. A strong argument has been made that a billionaire tech bro bought a website specifically to destroy it because it was a place where journalists were able to communicate the truth — and he was just following in the footsteps of the *other* tech bro who teamed up with a wrestler to destroy several websites.

  As this article suggests, tech libertarians will never admit that anything developed by themselves or their brethren could ever be negative https://dgardner.substack.com/p/the-techno-libertarian-faith