MasterPolygraph

MasterPolygraph

Felix Farley

Felix Farley

London, England

I am seeking to create a polygraph that tailors its diagnostic model to each of its subjects using machine learning.

Artificial Intelligence

  • 0 Collaborators

  • 0 Followers

    Follow

Description

This is my BHAG (big hair audacious goal) for the Intel Ambassador program, so bear with me. Should all go according to plan, this project will culminate in a polygraph that tailors its diagnostic model to each of its subjects using machine learning, based on biometric statistics recorded through a series of statements. Ultimately, the plan is to revolutionise the justice system by applying the 'MasterPolygraph' to identify lies under oath in a courtroom with a higher probability of accuracy than courts have historically achieved.

My first step in the long journey toward creating this model is to amass biometric data recorded as people make a range of scripted statements under examination, with each statement varying in truth and falsehood. I will feed this data into my hypothetical AI and try to teach it to accurately diagnose the truth or falsehood of each statement. The AI will begin its diagnoses with a statistically random output, but I will implement a reward incentive in the algorithm to create a genetic machine learning function, so that the AI optimises its matching of given biometric patterns with the statistical likelihood that a given statement is true.

Over time, I will seek the highest statistically-significant correlation possible between the AI's diagnosis and the veracity of the statement. Once this is done, the model can be extrapolated for use in the analysis of unscripted statements. This is the second phase of research. In turn, the results yielded in the first phase of research will be recorded and the accuracy of the model will be gauged. I will seek to identify consistent variables in the biometric patterns with which a correct diagnosis is associated and feed the statistics back into the model over an extended period of model refinement, until such a point that the precise biometric patterns have been identified for a given sample group.

The third phase of research is to then apply the refined model in experiments with random participants, extrapolating the algorithmic methods developed in the second phase of research. Of course, the MasterPolygraph could never be 100% accurate and it is concerning to think that justice might be entirely outsourced to artificial intelligence, so its purpose would only ever be to produce highly accurate diagnoses for consideration by the courtroom, in addition to other evidence.

Medium screen shot 2017 10 23 at 6.55.41 pm

Felix F. created project MasterPolygraph

Medium e6704ae2 e196 44ec 9a5e 43f005e47ccd

MasterPolygraph

This is my BHAG (big hair audacious goal) for the Intel Ambassador program, so bear with me. Should all go according to plan, this project will culminate in a polygraph that tailors its diagnostic model to each of its subjects using machine learning, based on biometric statistics recorded through a series of statements. Ultimately, the plan is to revolutionise the justice system by applying the 'MasterPolygraph' to identify lies under oath in a courtroom with a higher probability of accuracy than courts have historically achieved.

My first step in the long journey toward creating this model is to amass biometric data recorded as people make a range of scripted statements under examination, with each statement varying in truth and falsehood. I will feed this data into my hypothetical AI and try to teach it to accurately diagnose the truth or falsehood of each statement. The AI will begin its diagnoses with a statistically random output, but I will implement a reward incentive in the algorithm to create a genetic machine learning function, so that the AI optimises its matching of given biometric patterns with the statistical likelihood that a given statement is true.

Over time, I will seek the highest statistically-significant correlation possible between the AI's diagnosis and the veracity of the statement. Once this is done, the model can be extrapolated for use in the analysis of unscripted statements. This is the second phase of research. In turn, the results yielded in the first phase of research will be recorded and the accuracy of the model will be gauged. I will seek to identify consistent variables in the biometric patterns with which a correct diagnosis is associated and feed the statistics back into the model over an extended period of model refinement, until such a point that the precise biometric patterns have been identified for a given sample group.

The third phase of research is to then apply the refined model in experiments with random participants, extrapolating the algorithmic methods developed in the second phase of research. Of course, the MasterPolygraph could never be 100% accurate and it is concerning to think that justice might be entirely outsourced to artificial intelligence, so its purpose would only ever be to produce highly accurate diagnoses for consideration by the courtroom, in addition to other evidence.

No users to show at the moment.

No users to show at the moment.