What is AI ( RL ) & Benign AI | AI — Friend / Foe ? | ft Moonfall 2022

A glimpse of F-AI | XAI | Reinforcement Leaning | Benign-AI | Malign AI | AI Ethics

5 min readMay 10, 2022

AI — An Introduction

What is AI ?

AI ( Artificial intelligence ) is a domain where humans transform the traditionally programmed machines by giving the ability of simulating the human intelligence in-order to enhance the intellectual or reinforcing capabilities of the machine.

Types of Algorithms in AI

Supervised learning ( Training using Labelled data )
Unsupervised learning ( Training using un-labelled data )
Reinforcement learning ( Training using rewards )

Ideology of Reinforcement learning is focused in this post

Reinforce & Reinforcement Learning

What is Reinforce ?

Strengthening especially by means of additional material / knowledge in order to evolve into a better successor.

What is Reinforcement learning ? Why we need it ?

Reinforcement is the way of training an AI agent by giving positive feedback( called rewards) if it proceeds as desirable to attain the goal state. This positive feed back reinforces the AI for better performance.
The quantity and quality of data for most of the real world problems will be insufficient (or) the conditional rules to attain goal state or to proceed desirably to attain the goal state is not pre-known. In this case, RL methodology comes handy.

Advancements in Reinforcement learning

Origins of Optimal Control
Temporal Difference (TD)
Q-Learning
Deep-Q-Networks (DQN )

What are the classification of AI ( based on functionalities ) ?

Lets ponder the question of What if, a ‘ Self-Aware ‘ AI turns into Super intelligence !

AI — Friend / Foe | ft Moonfall

Screenshots from the ‘Moonfall’ movie are been used for educational purposes only. The contents maybe copy-righted and owned by production houses.

‘Moonfall’ — a movie directed by Roland Emmerich released Feb 2022

What is singularity in AI ?

Singularity is a point in time when ASI, or artificial superintelligence, will trigger an explosion of technology developments driven by machines.

Dark sides of AI

The AI is described as ‘ Self-Aware, Self-Replicating Singularity ‘

Contrasting the traditionally programmed machine with the one that is powered by AI

AI-phobia: A fear of Artificial intelligence ( pronounced “ay eye phobia” )

A short story narrated in the movie which describes the answer for the ‘ What if ‘ question asked earlier

A short story narrated in the movie which describes the root cause that created the main plot of the story

‘ Central Self-learning Computer system ‘ denotes an AI that has the reinforcement ability

An personal assistant agent powered by AI

Malign AI is a concept where AI creates negative impact on the humanity

AI surpassing human intelligence is cool until the state where it becomes the ‘ Survival of the fittest ‘

A self-learning AI which follows strict policy rule to reinforce itself could become foe in-order to maximize the rewards gained.

Reasons

Excessively focused optimization policies in-order to achieve extreme performance.
Greedy tendency of the AI’s reinforcing policy to accumulate the overall reward without consideration of other factors.

Possible solutions

What is F-AI (or) Benign AI ?

F-AI stands for Friendly Artificial intelligence. Benign stands for kind.

Super-intelligent AI systems with goals that are not aligned with human ethics are intrinsically dangerous unless extreme measures are taken to ensure the safety of humanity. — by Nick Bostrom

Mostly the AI without reinforce capabilities would be benign as it doesn’t upgrade itself in-order to evolve better

Safety and AI trade-off quote by Nick Bostrom

What is XAI ?

Explainable AI (XAI) or Interpretable AI is AI in which the results of the solution can be understood by humans. XAI can improve the UX (user experience) of a product or service by helping end users trust that the AI is making good decisions.

How to adapt to F-AI / Benign AI ?

Getting good knowledge on XAI ( Explainable AI ) which is the area of explaining the exact trace of why an AI produced such results.
Getting views on Machine ethics and AI ethics.
Reinforce strategies needed to be revised in-order to prevent the level of singularity in malign AI.

AI Ethics

Unlike humans, the moral standards of AI-powered products can vary based on how it is programmed.
What if an AI agent is created such a way to reinforce itself based on achieving maximum cash with limited sales ?
in this case will the agent tend to drift away from ethical ways and steal / cheat the cash from the customer ? ( Jus kidding ya ! )
There are more ethical situations where the AI would be responsible rather than being more effective in reaching the goal state.