Artificial Intelligence for People in a Hurry . From a beginner to expert in 8 Minutes!!

8 min readDec 5, 2018

What is Artificial Intelligence?

Artificial intelligence (AI), sometimes called machine intelligence, is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and other animals. In computer science AI research is defined as the study of “intelligent agents”: any device that perceives its environment and takes actions that maximize its chance of successfully achieving its goals. Now Artificial intelligence is in its golden era.After the advent of General Purpose GPU (Graphics Processing Unit) the field essentially under went a massive transformation.

How Our Brain Works?

A biological brain consist of neurons.Which are small cells that are the basic building blocks of brain.A human brain consist of billions of neurons which collectively computes our reality and experiences.Basically our brain consist of billions of neurons which are interconnected to each other.The strength of the connection is modifiable.This is called plasticity. When a new information is to be stored in the brain these cells changes their connection strengths and thus facilitates the representation of the information in the brain. Brain was evolved to control and manipulate the tons of muscle tissues in the animal body to ensure its survival.Nature cannot figure out all the possibilities that a being might experience in its lifetime,that is why brains evolved.Since the programming by the nature is not enough to circumnavigate the myriad of problems and situations the animal is going to face.Humans are always fascinated by the inner working and secrets of the brain.The quest to understand how the brain works have always enticed the generations of mankind.Finally after so much of scientific progress and tireless research we are now creating artificial brains.An artificial brain or machine intelligence is analogous to the biological brain.Both functions in the same way.In an artificial neural network we can small mathematical function modeled after the neurons in the brain itself. A neuron is a simple cell that fires or get excited when a particular amount usually known as threshold amount of input is provided.In a similar manner an artificial or machine intelligence neural network get triggered only when the sum of the input reaches a preset threshold level.When this firing happens this neuron will trigger adjacent neurons and then the process continues.Every thought that we have in our mind are the result of some neuron firing.A biological brain does not have too many different components.It is having complex but similar arrangement of neurons.When we learn something the connections are modified.When computers became more and more powerful thanks to Moores law we were able to simulate small neural networks .But the only thing constraining us are the compute ability.Now the Moores law is being replaced by supercharged laws in which doubling of the transistors can happen in a very less time.

Leveraging the LOAR Low of accelerating returns and ever growing compute power we hope to see the development of super intelligent machines.A super intelligent machine will have orders of magnitude more compute capability than humans. Scientists look forward to a time when the compute capability of a machine far exceeds the capacity of a human and named it the SINGULARITY.It is excepted to happen in this century.The development of AGI (Artificial General Intelligence) is one of the hottest research and commercial opportunity right now.We are now seeing futuristic technologies like self driving car,machine translation,Deepmind AlphaGo,Open AI Dotabot etc.Every advancement in AI thanks to the neural network architecture.Usually these algorithms learns like us.They are fed with tons and tons of data and are equipped with the state of the art AI algorithms that are let loose on the data.These algorithms figure out a way to optimize the losses to get the target value as output.The algorithm powering this architectures is usually back propagation and in some cases evolutionary algorithms.

Consider the case when an AI achieves AGI,this would be the greatest invention of mankind.The fundamental advancement in AI right now are

Computer Vision
Reinforcement Learning
GAN (Generative Adversarial Networks)

The fundamental building block of a computer vision system is CNN (Convolutional Neural Network) and the fundamental building block of a Machine translation algorithm is an LSTM or GRU (Long Short Term Memory).All the hype happening in the Machine Intelligence space is the novel and interesting new combination of the above units.Some scientists even claims that it is only the matter of compute power to create an AGI.

The Effect of AI on Jobs

This topic has gotten so much attention nowadays because the emergence of new AI techniques that can replace most of the human labor in almost all the industrial verticals.What the stem engine did to the horses are gonna happen to humans .When the muscular power of the horses were replaced by machines an industrial revolution took place.And the same thing is going to happen to the human labor also.No job is safe from AI automation.Every industry will fall prey to this technology.But there is nothing to be scared because all the repetitive boring works will be done by ai. We can enjoy our self in more meaning full works. The present pace of development of technology is so rapid which is summarized in the below videos.

https://www.nvidia.com/en-us/deep-learning-ai/industries/ai-innovators/?ncid=so-twi-mds-28980#docuseries

Applications of Artificial Intelligence

According to a famous saying software is going to eat the world and AI is going to eat software and everything.Every industry is going to be transcended by AI.According to Kevin Kelly the formula to the next genre of startups are SOMETHING+AI.AI is often attributed to fire or electricity which signifies the level of change this technology can bring to the table. The fields that are ripen to change are soon to be impacted by AI .The recent advancement of robotics and also the application of computer vision is on itself going to change the fabrics of human living.The major sectors that are about to be transcended by AI are given below .

Each and every business can be powered by an AI.In the competitive world in order to stay ahead businesses need a super human help.And the fuel that is needed to supercharge the AI is your data.We can help you take your venture to the next level by supercharging it with some AI.We have automated the model creation ,tuning and most the heavy lifting behind the scenes to create an AI for you.Please provide your email and an AI expert will walk you through all the steps required to create an AI powered business.

Artificial Intelligence in the cloud and Compute

Moore’s law has been dethroned!Training speed has been increasing exponentially with a 3.5 month-doubling time (by comparison, Moore’s Law had an 18-month doubling period).A super charged law has come into existence.The ever accelerating computing power now doubles in every 3.5 month period. This is so fascinating because the compute that we can achieve within next 10–50 years is mind boggling.The overwhelming compute power acts as a catalyst in the growth in scientific breakthrough and AI applications. Since 2012, compute has grown by more than 300,000x (Moore’s law only 12x).

The major catalysts affecting the AI are:

Computer Power
Data
Algorithms

This three factors are the pillars of modern progress in AI.But the Data part can be replaced when powerful algorithms like reinforcement learning can learn without the aid of crafted data.The innovations like artificial curiosity and GAN can replace the dependency on the data part.According to Google founder Larry Page “AI is only a compute problem“.Once we have enough compute then making an AGI seems to be a achievable goal.Open ai researcher Ilya Sutskever is one of the flag bearers of the modern ai. He is super optimistic about the future of AI and AGI.He also believes that reinforcement learning is the key to AGI. A reinforcement Learning algorithm is a method pioneered by Richard Sutton.It is a simple algorithm in which a computer program tries to maximize the rewards that are given to the program.Reward usually means any value that we like to increase.For example in a game playing ai the reward is the score of the game itself. AlphaGo tries to increase the reward function that is obtained when the game is won.A reinforcement learning algorithm can be achieved through the simple following steps;

Do some actions randomly
Check for the feedback or reward
if the reward is positive increase the probability of those action sequences
else decrease the probability of those actions
repeat

Although this is a simple algorithm scaling it up with more complex methods and giving it more compute can do marvels.A reinforcement learning agent nowadays use a neural network to learn the task.All the breakthrough achievement recently achieved are done with the help of a neural network tied to the above algorithms.Deep Learning is a new kid in the block gained prominence after the 2012 Alexnet event.

Deep Learning

While traditional machine learning algorithms are linear, deep learning algorithms are hierarchical.Deep Learning networks have layers of neurons.Usually there are three types input neurons ,hidden neurons and output neurons. The outputs of the first layer is further processed by the second and output of second is processed by the third and so on and so forth. This type architecture of increasing complexity and abstraction lead to more learning capacity of more complex topics. In the brain similar process is undergoing which is called Hebbian learning.The cells that fire together wire together.For example when a dog is given food along with an alarm the dog brain cells receiving both the taste inputs and the auditory inputs from the alarm tied together and when the alarm is sounded without the food the dogs mouth starts watering.The fundamental ideas that powers a Deep Learning neural net is Back propagation algorithm .This algorithms nudges the neural networks in the direction of the target results.Which lead to the modification of the weights of the neuronal connections and give rise to new learning. A little bit of understanding of Calculus is the required to fully understand the back propagation.In simple terms it is the calculus chain rule on steroids.

The Equation of Machine Intelligence

The idea of Machine intelligence or Artificial Intelligence can be expressed in a simple mathematical equation.Here we represent Y as the output and X as the input and T as the target output.For example if X is an image of a cat and Y is the output of the network and then T is the name cat.Then we can simply formulate the equation as follows;

X=Input Image

Y=Output Value

W=Weight Matrix (AI Program)

T= Target Value

Y=X*W

so we need a W that can change the input to output label.So how can we figure out W?

W=Y*X^-1

Here W is the computer program that can transform an X into T.

Loss = T-Y

Once we are able to reduce the difference between the Target T and output of the network we can create a program W that is the magic formula to convert the inputs into valid output values.

Web

http://www.appdron.com

https://appdron.com/decentralized-peer-to-peer-mobile-intercom-system-2-4ghz-and-900mhz/

https://appdron.com/enterprise-management-android-application/

https://appdron.com/robotic-smartphone-assistant-deep-learning/

https://appdron.com/sample-page/

https://appdron.com/social-media-marketing-definition_appdron/

https://appdron.com/the-age-of-artificial-intelligence-applications-and-the-future/

https://appdron.com/wanna-laugh-artificial-intelligence-powered-application-that-can-detect-laugh/

https://appdron.com/machine_learning_ai_what-if-we-can-talk-to-the-web-ai-browser-voice/

https://appdron.com/worlds-fastest-and-best-qr-barcode-scanner/

Download Apps