Understanding the Brain:
Functional Connectivity Analysis

How do we meaure hypnotic trance?

There are three primary activities in the brain to consider when in trance:

1. A decrease in dorsal anterior cingulate activity
- This is responsible for decision making, evaluation processes, and emotional regulation as well as physiological functions such as blood pressure and heart rate.

2. An increase in the connection between the dorsolateral prefrontal cortex and the insula
- This leads to a stronger brain-body connection; more specifically, the pain perception, social engagements, emotions, and autonomic control functions of the brain more readily impacts the executive functions such as working memory and self-control in the dorsolateral prefrontal cortex.

3. Reduced connections between the dorsolateral prefrontal cortex and the medial prefrontal cortex
- This heightens the disconnect between actions and awareness of actions

How do we record and analyze these signals?

An intracranial EEG-based, noninvasive BCI can collect general brain activity data, localized in specific regions of the brain and over time. This data is analyzed using network analysis to determine the functional connectivity between different regions of a brain during a specific cognitive task or process.

Then, using neural input from EEG BCIs, where each electrode represents a node, we can conduct functional connectivity analysis using a method called “phase locking.”

Phase locking is the detection of where the neural signals (which come in the form of oscillations) are synchronous based on frequency. Between a pair of nodes, the weight (a representation of connectivity) is defined by the high gamma phase-locking value (70–100 Hz) for that pair.

Thus, we can use this proposed BCI analysis method to provide a portable means of measuring hypnotic trance.

Content Generation Process

Script Generation

First, we will build a dataset of a large number of script lines consisting of calm, hypnotic suggestions, which we will then put through a GPT-3 text generation model.

The GPT-3 will write out custom scripts based on the type of addiction the user wants to overcome. For example, if one wanted to stop drinking whenever I’m out with friends, the text generation script could create a custom script based on the circumstantial trigger.

Script Audio Delivery

Scriptwriting is only half the battle, however. Once we have the script, we need to figure out the best content delivery. Now, we could just use a simple text to speech, but doing so could mean that the user's subconscious would just reject the suggestion.

Our ML model will be run on a series of different factors, including audio volume, tone inflection, speaking speed, and background sound effects. All of these different attributes will, in the right proportions, have a positive effect on the user’s experience.

Improvement and Training

We will use biofeedback (EEG-based BCI) as well as conscious feedback in order to determine the effectiveness of our script / delivery. Conscious feedback will include data such as the number of drinks a user consumed in a week, or the number of temptations a user felt throughout a day.

Using a text-to-speech voice synthesizer that will relay our script with the aforementioned qualities mentioned above, we can test out different combinations of attributes to custom tailor and make our suggestion delivery methods as effective as possible for each user.

GPT-3’s language generation is based on 175 billion parameters and is by far more accurate than its predecessors. For example, GPT-2 had only 1.5 billion of parameters, and Microsoft Turing NLG only had 17 billion of them. Our hope is that PGT-4 will be based off of 200 billion parameters, which should allow it Parameters are network calculations that apply particular weights to different aspects of data. Thus, every data aspect receives its value and data perspective. Thanks to this massive amount of data, the language is capable of meta-learning. It means that GPT-3 can do tasks without any training, from a single example. OpenAI GPT-3 also has some advanced creative capabilities due to its context-based nature.

Once a user puts in a request, the language analyses it and provides the most probable answer. The text predictor processes all of the text existing on the Internet, calculating the most statistically expected output.

Therefore, GPT-3 is extremely powerful without understanding a single word it produces. Without specific tuning & adjustments, the language can write stories, blog articles, PR materials, and even hypnotic sessions. Most of the time, the outputs feel very similar to those written by a human, something that will only increase as their ai’s get better.

Technology Deep Dive