It seems like if ChatGPT had literally any other way to turn a profit, they would have done it before resorting to this. Kind of fees like this is their last ditch effort to show investors that they can somehow become profitable.
Same. If chatgpt actually could do what they claim, would they be wanking (heh) around with erotica and job boards, or would they be selling $5k/mo subscriptions that replace junior employees? The answer is obvious.
Is the idea that hackers can review the source for vulnerabilities?
I would imagine any internet-facing application would not have any issues with a source code "leak" like this, as all inter-app communication is based on short-lived credentials.
> Nobody wants a computer. They want what it does.
This is exactly why we have AI assistants in the first place.
The typical person wants the damn phone / computer / tablet out of their life, they just want to say "Siri, make my dentist appointment for 9:30 Monday morning" or "Copilot, find all the pictures of my dad and put them together in a collage for his birthday".
The biggest problem with AI assistants and this thought process in general is that they do not work.
As the article states, if the AI assistants actually did things we wanted, these companies wouldn't need to shove them down our throats so much. We would just... use them.
Agree on it being AI, but what really screamed "AI slop" to me was the emojis. I don't know any tech bloggers who use emojis, but everything that ChatGPT or Gemini generates always has too many emojis.
I've used GIMP and Photoshop for a very long time (close to 30 years).
In ~1998, GIMP was not quite as good as Photoshop 5 and was more awkward to use, but you could see how it could close the gap. It had impressive underlying tech that could handle large images on computers at the time. There was an expansive library of weird and neat plug-ins and scripts. It felt like we were at the start of a great shift in which OSS software would "catch up" and eventually replace desktop power tools, just as Linux had done with web servers. It was... the year of Linux on the Desktop!
By ~2005, GIMP was starting to really catch up to where Photoshop was in 1998, but Photoshop had added lots of quality of line features like adjustment layers and layer effects, way better text rendering, and amazing new features like spot healing brushes, vanishing point warping, etc. The gap was widening. But GIMP still did all the core stuff, and Photoshop was annoying users by shoving Adobe Bridge down their throat, etc. So people were still hopeful for a replacement.
By ~2012, GIMP was adding.. an awkward single window mode? It lacked tons of by-now-basic features that made it totally impractical for professional use. Photoshop, meanwhile, was adding amazing time-saving features like Content-Aware Patch and Move that seemed "magic" at the time. The tech gap was widening, but Adobe was also pushing subscriptions down users' throats, which was very unpopular, so GIMP still had a chance to make a come back.
By ~2018, GIMP was finally adding.. basic CMYK support for printing, something which literally no one uses GIMP for professionally and was a dying need? Meanwhile, Photoshop was demoing an AI object selection tool that could magically select objects without needing to trace them, which came out in 2019. Using GIMP felt like using software from a decade previous.
The last 5 years have been the worst for GIMP. Photoshop has been improving at an astonishing rate. Now it's literally what photo editing looked like in 90's movies - you just open an image, click "select object" and it perfectly selects it, and lets you move/drag/add elements with AI, etc. You can do edits in seconds now that used to take hours, and the results are really good.
None of this is a complaint about GIMP or all the people who contributed to it. It's impossible for a few volunteers to complete with infinite money and hundreds of full-time employees. But Photoshop and GIMP are no longer in the same league. And Adobe knows this, which is why it can get away with punitive subscription-only pricing.
It used to lack non-destructive editing ("adjustment layers" in Photoshop parlance) until recently, it's a core foundation of editing workflows for designers and photographers, it lets you layer transformations of over immutable rasters. This was in Photoshop since 2005.
It has been a very long time since I tried GIMP (>15 years) to remember everything I found wanting, but as I recall, GIMP lacks both macros and batch editing, the former letting you record a set of actions to a hotkey so you don't have to repeat them yourself all the time, and the latter letting you apply a set of actions to hundreds or thousands of images at once. I would literally have to spend hundreds of hours to do things in GIMP that can be done with no effort in Photoshop, to the point where it would actually be easier to just program something myself from scratch than it would be to use GIMP, if Photoshop didn't exist.
I see that GIMP has since gotten a UI revamp, but the multiple window UI from the time I used it was also unbearably bad and one of the main things that sticks out in my memory.
Have you looked into script-fu? It would probably be a very steep learning curve.. BUT there is an opportunity to do something impossible 10 years ago, and that is to use AI and an external application. BATCH-FU is one such attempt but it seems to be a 'select action from a menu' thing.
But Gimp developers: implementing batch in one go is a big ask I know. But a great first step might be to create a channel in Gimp where correct script-fu is emitted for operations in progress. Being able to connect to that from outside would allow 3rd party projects to assemble "record by doing" macros that could be turned into Photoshop-like batch capability.
Macros are on the roadmap (https://developer.gimp.org/core/roadmap/#macros-script-recor...), and in fact we did a lot of prepwork for them during 3.0's development (internally, several features like filters and plug-ins now have configs that store settings, which will be used by macros in the future to repeat operations).
reply