AI – A very tricky subject for creatives.
Lets just say – before I start this, I am VERY against AI as a creating tool, making images and calling them “original” as they are not. AI generated images do not replace “art” or any work created by a human for a simple reason, they are not created from emotion, from a soul.
AI created images are plagiarism on a grand scale (although lets be honest – most creative work now is influenced by previously created work and therefore has some small level of plagiarism) but if created by a human, it will always be created with an emotional heart.
Obviously press photography operates to different criteria – it exists to accurately record events unfolding in front of the photographer where the human element is to remain impartial, to be faithful to what actually happened, using fractions of a second to tell a moving story., something that an AI image creator cannot really be trusted with at the moment.
So all that being said – is there a place for AI in our industry at all?
With other industries moving forward with AI technology, streamlining their workflows to save time and money, is there any room for the AI tools in our industry, for agencies or freelancers? If so, is the use of them just hypocritical given all I have voiced above?
As I started with – a very tricky subject.
Long time readers will not be surprised to learn that I have been investigating various tools, over the years this blog has shown, I think, that I am quite pro-active when it comes to using technology to streamline my workflow so that I can concentrate on what matters to me – making images. So what have I been playing with, how what have I learned so far?
Probably my first sustained use of the technology has been in my bookkeeping (and that is probably true of most of us not using spreadsheets). The way my system scans, analysis and tracks my expenses has been a huge help over the past 12 months, so much so that I have increased the level of my subscription on Zoho Books to lower my burden further. This of course is a fairly rudimentary AI and rule driven task set.
Then of course there are the AI aids I am using to help correct my (awful) spelling and grammar as I type this (the first post I have written outside the wordpress platform, before copy and pasting it in).
My experimentation has moved on. For the past couple of months, as a background task, I have been captioning and keywording iPhone photos I have taken over the last 5 years, uploading them to a well known stock site. That site has a contributors forum, where I became aware of multiple people using AI to help in this specific task, so I started to give it a go.
My bank includes the pro version of perplexity as part of its pro account subscription fees, so this has been my platform of choice.
My starting point was captioning each image myself, writing the caption as I have always done. I then fed that caption into the AI to ask it to generate the keywords for me, based on the caption (which outlined what I thought was important). This was my workflow for quite a few weeks and I was generally happy with the keywords generated. They were very specific and to the point, listing only what I thought was important.
It was time-saving and a success, but definitely not as time saving as it could be.
So following a good bit of research (would anything I upload be used to train the model).
Are any images I upload here used to train ai models before they are deleted
No, images you upload to Perplexity are not used to train any AI models before they are deleted. They are processed only for your query responses and not retained for training purposes.
Image Processing and Privacy
Your images are analyzed using multimodal models to answer your specific query, such as identification or captioning. They are not stored on servers beyond what’s needed for the thread, ensuring privacy during use.No Training on Uploads
Perplexity does not use uploaded files or images for AI model training; they stay private and are only for customizing your responses. Agreements with third-party providers like OpenAI and Anthropic explicitly prohibit them from training on Perplexity data, including images.Retention and Deletion
Images are retained for 30 days (7 days for Enterprise Pro) for follow-ups, then contents become inaccessible while thread context remains. Delete them anytime via the web (click X next to the file in the search bar), or delete the thread entirely; use the support form for immediate removal requests.Opt-Out for Other Data
While uploads aren’t used for training, you can opt out of Perplexity using your search data for its own AI improvements via Account Settings > Preferences > AI data retention toggle (Enterprise excluded by default). This doesn’t affect image handling or core features.
I thought I would try uploading watermarked images directly to see the result. It has taken a few days playing with the command text to get a result I am happy with (and no, I am not giving away my crown jewels here) but I do have it generating captions and keywords.
It is not 100% reliable (as a few examples below will show) but it is pretty good and cuts the time this (horrible) task consumes.
No matter how I tried, the tool always captioned the image as the Hercules statue at Portmeirion – which of course, it is not. However, the rest of the caption is pretty good.
I am far too lazy with stock imagery to get into that level of detail with a caption, you can decide if it is suitable or not.
Of course, here the spelling of Colourful is not right… Also with this image – I had to correct the caption to read frozen, the AI insisted it was a puddle!
This is unaltered from the AI result.
I spent yesterday morning captioning and keywording. I completed and uploaded over 100 images. Using my old long-hand workflow this would have taken me longer than a day to research and complete.
As anyone who has used AI, the key is the prompt is everything.
As perplexity thinks it shows where and how it is researching – to create what I used , it scoured many resources that of course included stock image sites. So is this use also unethical?
Then there is the news/live event use – how about using AI to recognise the people in a Red Carpet image? I know there are agencies using this technology – is this unethical?
I finish as I started – I am very against the use of AI to create an image, it is plagiarism on level we have never experienced before. However – what about the uses I have outlined above – where do you stand on these? I really expect a discussion!
I look forward to hearing from you.
J
Julie Edwards-X 