Artificial Intelligence (AI) has been one of the most heavily misunderstood areas of our time.
In the past several years there have been 2 streams of throught from people that I have met:
- AI is MAGIC
- AI is EVIL
Actually, AI is actually neither. AI cannot think on its own - so it is definitely not magic. AI cannot therefore do good or bad without any ability to think - so it isnt evil either.
A vast majority of AI is mathematics, logic and reasoning - which is also a heavily misunderstood and enigmatic field of our time.
AI is complicated mathematics and logic, and like math and logic, it is enigmatic.
Artificial Intelligence is a machine that is programmed to emulate a human using mathematics and logic.
Anything powered by AI presents a mathematical approximation of a problem. AI can never do better than humans at subjective problems (e.g. identifying humorous comments), but it does really well at objective problems (e.g. answering factual questions, playing chess).
To use AI in your business or product, you need to convert the problem into a mathematical one that an AI system can learn from.
Today, AI is mainly mathematics, instead of expert systems and logic. This means that in order to use AI in our pipelines, we need to first convert the input data into mathematics. This means - words, images, sounds are all converted into a mathematical representation called vectors (each vector being a sequence of numbers).
Deep Learning and AI
The lastest and greatest in the field of Artificial Intelligence has been the advent of Deep Learning ([DL] (https://en.wikipedia.org/wiki/Deep_learning)). Deep Learning is the field of study of AI where data (in the form of vectors) is processed through large (deep) neural networks to solve problems.
In order to analyze these large vectors using deep learning, one needs tremendous amount of processing power - more than multi-core CPU’s - usually GPU’s and TPU’s. This is one of the reasons why deep learning based AI has been limited to large organizations that can spend the $$ to buy cutting edge processors, or hire the elitest developers and scientists.
Machine Learning and AI
Machine Learning (ML) is an approach to statistically learn from large amounts of data by recognizing patterns. ML is different from DL based on the architecture and algorithms use in teh backend. DL is built over neural networks, where are ML is based on statistics.
Natural Language Processing and AI
Natural Langauge Processing (NLP) is the subfield of AI where the input data provided is in written textual form. In order to do NLP one often uses techniques from DL, ML and others.
AI is not magic, or evil. Artificial General Intelligence doesn’t exist yet - because machines cannot think and reason or have commonsense knowledge like humans.
For further information about AI check out John McCarthy’s answers to questions about AI.
John McCarthy is also known as the father of AI.