Regulatory Framework and Guideline for AI – Brief Approach Note

Limitations of AI and Comparison with HI (Human Intelligence) : Our uniqueness of being Human Beings is derived from our dependence on Human Intelligence; our Integration with our Belief Systems gives us the liberty to choose the path between Scientific Logic and Human Reasoning, our Faith and Trust on our Traditions and Customs Differentiates us from the  Theist and Atheist. For a Theist, the replacement of our daily rituals and communication with the Creator of the Universe by an Alien System which we have started knowing as AI, is not only challenging, it is rather Irreplaceable. Human-Machine interface demands partnership and collaboration, with HI retaining the right to review and modify or even reject AI output. Execution and Implementation should not be left with AI, rather it should be delegated to a group, retaining a Sense of Justness of Purpose and Human Intelligence.

The domain of Human Intelligence is Reason, the domain of Artificial intelligence is Algorithm. The parting line between the two is our Ethics and our Moral Values. Human recognition of any Situation or Truth is driven by the Freedom of our Choices, Ignited by the Foundation for a Reason, a Sense of Justness, driven by our quest for Fairness and Openess; we are driven by a passion for Public Acceptance and an overall Sympathetic View towards Human Values and Principles. But an AI does not offer any of the above. Driven by its Algorithmic Capabilities, it is rather disconnected from Human Intelligence and Human Consciousness. The endeavour of the Scientific Community lies in establishing a bridge between the two. (The below table tries to compare some parameters between AI and HI, and predict their natural Areas of their Individual Supremacies. The scores are given in a scale of 1 – 100, to represent the gap in abilities of AI and HI)

1.    Framework for AI Guideline : Under the lens on transparency and trust building, common man finds it easy to understand any scientific outcome, easily. The judgement and decision making of AI needs to be in the language of the common man, so that the acceptability of the technology is easy. Orders and Instructions, Guidelines and Judgments must be addressed through Human Intelligence and Human Reasoning. Before accepting any AI judgement and decision, the balance between Human- Machine Interface needs to be established through Collaboration and Partnership. The right to Review, Reject and Modify an AI judgement should also lie with the Human Mind. Let AI be a Business Enabler that facilitates in the consolidation of vast amount of data and information, let AI be another data aggregator that goes through all the information available, scans through it and shares us a brief note of all the documents. However, parallelly, the output of the AI needs to be confirmed and be finally approved by a human entity. Let us not relax in the glory of the fact that AI is just another neutral tool to help us. The fact, that the AI is capable of Continuous Improvement through Self Learning is a potent strength in the AI, that retains the Capability of Surpassing and bypassing any Human Instructions, once it Learns to Declare Ones Independence. To bring Accountability and Responsibility on the part of AI Creators, a proposed framework has been suggested, guided and regulated by a ‘Regulatory Framework and Compliance Guideline for all AIs’.
   
    
   
   The Objective of such a framework is not to control or throttle the growth and development of AI, but rather to Regulate the Output of the AI, with clearly defined Objectives. Such a model should also be having the window, to audit the data and verify the data that goes to feed at AI, and to audit the Algorithm as well. The AI, the Developers of AI, the Owners and Promoters AI shall be working under the Umbrella of Government Policies and Guidelines.
   
   2. Framework for Algorithm Audit  : The below table can show the Intent of the above thought. The Intent of the template is to create an Algorithm, based on the below Logic, run the Algorithm and confirm the results of the Test Run, with intention of Auditing the Data and Algorithm (The below table reflects only a thought process as to how we can create, check, and validate the Data and Algorithm). The No Risk carries ‘0’, Yes Risk carries ‘1’. All the parameters, if it carries ‘0’ (No risk), the Algorithm seems to have No Risk. This implies, there are No Bias or Tweaks in the Algorithm. And it is capable of giving a Neutral Output, without any biases :  
   
   
   

   Based on the above Template, an Audit Algorithm shall be developed and deployed to audit the Yes Bias/No Bias of the AI Algorithm.

     

   3.Framework for Data Audit :  The AI learns what it is taught. Whatever data and information the AI is fed with, it learns from it as its basic alphabets, from alphabets to words, and words to sentences. As a result, the kind of data that goes in the AI decides the output of the AI. To prove that data is Unbiased, the total data score should be 0

   
   

   In the above matrix, the Bias Score for data in situation 1 is 6.This implies the data is not neutral and implores room for auditing and reviewing the data set. In the column of ‘Desired Ideal Situation’ the Bias Score is ‘0’. This implies the data set used has ‘0’ bias, and can be used safely.

   Classification of data : Further templates can be designed to classify data sets as  Good data/Bad data; Ethical data/Unethical data; Right data/wrong data; factual data/fictious data, biased data/unbiased data etc. However, a defintion of all the above parameters of classification needs to be Addressed, Developed and Defined, keeping in mind the Objective of the AI and wether it matches with the ‘Regulatory Framework and Compliance Guideline of AI’

    

    

   Conclusion : As AI is woven into the fabric of our daily existence of humanity, and as it Expands and Influenced the Tapestry of Human Consciousness, Humanity at times will stand in direct confrontation with AI. Overtaken and Mesmerized by the Deluge of the Unavoidable Influence of AI, some of us may be in a hurry to establish AI as a Divine Superpower, hurriedly sent on to earth to Liberate Mankind from the clutches of a Ruthless Dictator or from the unending cycle of Human Suffering and Grief. Such promises from the 'So called Divine Superpower' demands our Careful Observation and Attentive Caution before we jump on the ship of AI Superpower!! 

   We must exercise our choices with utmost caution, weather to allow AI to dominate our lives, or to brush it aside as another effort to enslave us through technology. Orders and Instructions from any AI enabler should always be legitimately evaluated. To retain the balance of relationship between AI and Human Mind, we should always be consulted before we accept any AI Judgement.

2.  

   Author : Bhaskar Nandi

   (Consultant with a Global IT Company)