Would anyone be interested in the idea of working to connect brain machine interfaces to OpenAI for building a platform/tools to help people think media into being? Please see this ArXiv paper I released recently and the abstract below.
https://arxiv.org/abs/2304.09858
I’m compiling a list of devices and looking for people who might know or enjoy learning coding stacks related to how the devices handle their data, via SDK/API. These are non-invasive EEG headsets and vary in capability and availability, but I’m interested in drinking Diet Mountain Dew and sprinting towards a demo; I think a less costly yet widely available and easily usable device may be best. Possibly Insight by Emotiv. I’m in the midst of applying to their dev license and exploring others. I could use help comparing and vetting api’s. Eventually all devices? But first a demo: quickest easiest. Hopefully <4 sensors but fallback on more sensors.
As you can see in abstract and paper the proposed method is to simplify text stream creation through Morse code because it is close to binary and amusingly someone has already vetted it with implants. And with a text stream magic is possible. There is some evidence of ML being able to infer letters so options are there.
Then the surprise and delight is to go from daydreaming media into being . . . into leaving a system on to cross over into recording dreams.
So hoping to gather and support interest and hopefully convince OpenAI and/or Microsoft to join in the fun. Vector databases and tensor search are unlocking memory for AI - tensor search is multimodal in theory, but Microsoft announced Kosmos-1, which might open up a great deal of fun and a kind of dream assistant.
I believe almost immediate fun could be had with existing tools and a quick sprint to typing or adapting some existing codebase.
There is also an intentional implicit joke on Elon Musk in the thought process, where if an implant researcher could be convinced to collaborate, the signal from it might optimize non-invasive EEG and render implants unnecessary. A light joke since Neuralink is seeking direct neuronal connections but still fun and helpful.
I can be reached here or on Linkedin and it would be fun to hear from you. My income dropped drastically after my contract was not renewed by SandboxAQ, and then I lost several clients who were spooked by potential recession. I may have found an income source for some base expenses and am talking to one potential angelic person. But I am interested no matter what.
Linkedin: http://linkedin.com/in/tekelsey
No longer a secret but keeping it somewhat on the lowdown. Pssst.
ABSTRACT
The present study proposes a novel approach to dream recording by combining non-invasive brain-machine interfaces (BMI), thought-typing software, and generative AI-assisted multimodal software. This method aims to sublimate conscious processes into semi-conscious status during REM sleep and produce signals for thought typing. We outline a two-stage process: first, developing multimodal software using generative AI to supplement text streams and generate multimedia content; second, adapting Morse code-based typing to simplify signal requirements and increase typing speed. We address the challenge of non-invasive EEG by suggesting a control system involving a user with an implanted BMI to optimize non-invasive signals. A literature review highlights recent advancements in BMI typing, sublimation of conscious processes, and generative AI’s potential in thought typing based on text prompts.