I have project files such as corridor lighting and the like and I wonder if ChatGPT can offer something in terms of optimizing the project, offering solutions. Can it? Maybe chatgpt vision can?
From the robot itself…
- Information Gathering: Start by explaining the specifics of your project – objectives, constraints, any existing issues, and goals for optimization.
- Image Analysis: If you can provide images, diagrams, or sketches that visually represent the project, I can analyze those. For example, for corridor lighting, you might provide images or drawings that show light placements, brightness levels, shadows, etc.
- Feedback and Solutions: Based on the information and visual representations, I can provide feedback on potential areas of improvement, suggest alternative design approaches, or provide insights into best practices related to corridor lighting or similar projects.
- Iterative Process: As with most design and project work, optimization can be iterative. You can implement suggestions, make changes, and then provide updated information for further feedback.
I would attempt number 2, it may be able to help if you have good images of what you need to show it from different perspectives. You could potentially even feed it design mock ups / blue prints but I wouldn’t know, sounds like a cool idea to me. Always review its answers of course.
You also could look into something built custom for you, but that would probably run you 5-10k for this use case.
I bet you it can. I’d upload some screenshots or images on GPT-4 of what you want to get done I’m sure it can offer solutions of what could be reworked in the lighting fixtures.
There is visual representations in the documents. I can share them as pdf if possible. Does file size matter? Each originally is around 40/50mb although I can compress.
From another thread here I found the answer
- You can upload 4 images at a time.
- It appears the max size of an image is around 10 - 15MB.
You could take a literal screenshot on your windows or mac and then upload that screenshot to the bot, and that would reduce the file size and make it so you could upload 4 images at at time.
To take a screen shot on Windows it is:
WindowsKey + Shift + S
to take a screenshot on Mac, press and hold these three keys together:
Shift, Command, and 3
Screenshot the visual representations in your PDF and upload those. Would this work?
What are you hoping to automate? Finding the right styles of fixtures? Finding right placement of lights? Staying within a specific budget?
You have “project files”? The AI can work with text, and specializes in answering factual questions.
Can it replace software like this? No.
Finding right placement of the lights would be the priority from what you’ve listed but in a perfect world I would rather have all of them combined.
Eventually there will be interoperability with Physics Engines, Gaming Environments, where AI can answer questions that involve advanced scenarios of physics setups, including about lighting.
Right now we have “multi-modal” which means images and sound, but in 10years we will have “Physics Modal” where the AI is running experiments and perhaps even looking for “the rest of” the laws of Physics, etc.
So sit back about 10 years, and then plz repost the question on this form. I’ll still be here. 
1 Like
lol try like 2 and a half if that. I don’t even believe the cliche “ten year” time lines anymore. Fuck it could be tomorrow
I meant 10 years away from AI doing theoretical particle physics research better than humans, but you’re right it could be only 2.
1 Like
Okay now, I checked out 100+ use cases of GPT vision and it apparently has document understanding skill but from my experience with chatgpt 2 and 3 it is probably aren’t capable of creating different scenario out of it like right placement of lighting as you say. From where I am I can’t get a refund if I make the purchase for ChatGPT and I’ll be only using it for my specific purpose so I need to know very well before. It will probably reply to me like; As a language model I am not blah blah…