This guy wants to make AI easier for people who know nothing about tech
Laodis is exploring how much you can get out of ChatGPT and others
Hey Laodis! Can you tell us who you are and how you ended up doing GenAI?
Hey Ben, was great meeting you and thanks for the opportunity! My name is Laodis, I’ve been in the startup world for about 10 years now and I’ve always been passionate about disruptive tech. In 2019 I wrote a science fiction novel where AI held a prominent role, and last year we built 2 projects on text and image with my co-founder, for which AI was transformative.
At that point we understood that one of the biggest challenges of GenAI would be to make it accessible to everyone, so we decided to build Argil, a no-code platform to train AI on your own data & scale any use case with automations.
Argil is a bit like the Zapier of GenAI. We can easily train models and use API keys to connect them to whatever we want. We can also do a lot of refining to get the best out of the latest GenAI models such as Stable Diffusion XL. Can you tell us a bit more about the tech and how you manage to get really good results? It can even beat Midjourney from time to time which is a great achievement!
One of our main goals is to make AI insanely easy to use for non-technical people, so we have a lot of built-in features for that. Everything on the platform has an easy mode and an advanced mode, whether it’s training models, using the image studio or creating workflows.
We also have a lot of hidden features built in to enhance the prompts. For example, in the Studio we have an automatic GPT rephraser for prompts: if you type in “a cat” it will generate a detailed prompt that drastically improves the results. We want our platform to feel like “magic” to the non-AI veterans.
Can you share with our audience what are the best use cases you’ve witnessed using Argil so far?
We’ve had a lot of fun with different use cases – when it comes to model training, we can decorate an interior in 2 clicks (img2img), get a storyboard for a photo shoot of a physical product with 5 input photos (model training), get a personal photographer with 15 photos of you as input, illustrate a script for your YouTube videos in any style (basic workflow + styles)…
We’re also planning to add text, which will open up a whole new field: Publishing? Have any book you receive read by AI, sorted in an Excel file and illustrated. Investor? Create a workflow that generates an entire investment memo using any set of notes as input. A million tasks that can transform the day-to-day of any job.
What are the next features on the roadmap that will be a game changer?
We’re about to release the “workflows” feature, which will allow you to create multimodal automations in a few clicks, in no-code or via API. You can then invoke them wherever you want, even in ChatGPT. The synergy between GPT and our platform is amazing – from the GPT plugin you can create prompt templates, workflows, train models, and you can do the same the other way round, creating complex workflows on our platform and generating the results in GPT.
What do you think is the next big thing that gonna happen in the next couple of weeks/months, ai-tech related?
I don’t think there will be a “next big thing”, but rather a series of small breakthroughs in each area that will cumulatively push AI to a new ceiling – Claude is working on context, Meta is working on multimodal, RunwayML and WonderDynamics are doing wonders in video, GAN models are also improving, GPT’s code interpreter is opening up a whole new world, Stability is doing great things for open source, large companies with big data moats are training their own models (e.g. Bloomberg).
Every field will have its own revolution in 2023. I truly believe that everything we do creatively and professionally will become massively easier within 24 months as each of these breakthroughs are introduced into our daily lives.
With your experience using GenAI Tech, you probably discovered a couple of tips and tricks. Which ones would you be ok to share with our audience?
The best trick for images is to use ChatGPT for a series of prompts – with the right guidance, it can generate entire sets of images into a JSON. Most of the image templates on the platform were created this way in a few seconds. It also works for benchmarking specific criteria in images.
Is there anything else you want to share with our audience?
We’ll be doing a ProductHunt launch early June, and would love any support there. Otherwise, come visit us on our Discord!
Post a comment