The creator explains the step-by-step process of how he built a sophisticated AI tool designed to predict human brain reactions to content and ads. He details sourcing open-source fMRI data, using various AI models for development, spending over $4,000 on GPUs, and explains how the final model works to map neural activity related to emotion, memory, and decision-making, concluding with a call-to-action for viewers to get access.
Post caption
I just built one of the most dangerous tools on earth. A tool that allows you to predict how a human brain will reach to any piece of content and ads.
Frustrated by Tribe V2 non commercial license I took matters into my own hands. I used Perplexity computer to run deep research for hours across the web to identify all available fMRI open source data that was available for training.
I then asked it to create a detailed plan on how to train a model using quad code that would beat or get close to Tribe V2’s outputs.
So it went out and generated it, and then I used Claude code to build the entire thing. I then proceeded to spend over $4,000 on Modal’s GPUs training our model fine-tuned specifically for content and ads.
And in just under a week, it’s live and ready to go. All thanks to the orchestration of multiple AI tools.
The barrier to do anything in this world has dropped to zero.
#ai #tech #startups #buildinpublic
Robert G
Transcript
Just built one of the most dangerous tools in the worse, and by the end of this video, you will have access to it. Over the last week, I've spent over $4,000 orchestrating countless AI tools to build the world's most accurate brain analysis tool for ads and content. I pretty much analyzed all available open source brain data from real fMRI scans of people hooked up to machines while they watched movies, shows, and content. We're talking over 200,000 data points mapped across thousands of regions of the brain. I then took this data yes for flexi computer plot code and $4,000 of modal GPUs to then build our own model on top of it. I'm talking about five AI models running in parallel, analyzing every frame of the video, processing the raw audio, and breaking down every word spoken. Then they all get fed through a custom transformer that learns one thing. When something looks good and sounds good, the brain does that. And through that, we're able to get back a full neural map of the brain, predicting activation across thousands of regions, like emotion, memory, reward, attention, and decision making. Giving you the power to optimize any content and ads to drive the outcome you want from any human brain here on Earth. Sounds fucking scary, but right now you have the privilege to access it. Just comment me down below, and I will send you the link. Have a good
