What is all this AI fuss about?
Machine Learning, Deep Learning, Predictive Analytics — what is the reality behind the hype? How do machines learn and what are their limits?
AI stands for Artificial intelligence, AI is not new its been around since WWII and the first computers, it only exploded in the recent years due to technological advancements and computational power, as now we can crunch massive datasets using powerful algorithm in almost no time and at a minimum cost.
Today AI-infuses systems are beating humans in most complex games, driving cars on their own, and creating work of art and even generate human like text (GPT-3)
BUT
How does it all actually tick under the hood?
First let's just understand what is intelligence?
Intelligence is the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience.
are we there yet?
not really
so, if you’re scared of machines taking over the world, don't be — it's not likely to happen in the foreseeable future. but we are making great progress.
When we talk about AI, we usually talk about two sub-parts of AI
- The AI that mimics “human like intelligence” is what is commonly called “artificial general intelligence” aka “Strong AI”.
- What is being developed and implemented by 99% of AI practitioner in the industry and academia today is a called “Artificial narrow intelligence” aka “Weak AI”
These solutions are designed to solve only one specific problem without any capacity to be translated to another one without rework.
you can call it a one-trick horse, but its a very powerful trick nonetheless.
When we talk about narrow AI we usually talk about the good old machine learning.
Machine Learning
It is the process of applying computer algorithms to capture the behaviour and behavioural pattern of systems and processes, based on the input and the output data collected from these systems.
under the hood, its good old mathematics, but AI-ML sound so much cooler.
Modelling the AI
A predictive model is a simplified representation of the actual process we are dealing with. It is neither feasible nor necessary to ever create a perfect model — we need one that is good enough for our application.
How do we build a model?
- Define the problem
- Collect the data
- Configure and fit the model
- Use it!
Start by defining the specific problem we need to solve and the associate measures of success. Then we collect data from our process inputs and outputs based on that we select the model to address the problem at hand and fit it using the available data.
Model Fitting — in scary mathematical jargon…
Model fitting refers to perform numerical optimization, looking for the minimum of the cost function in the parameter space
in human terms, fitting is the process of moving the knobs under the hood of our model testing as many combinations as necessary, until we find those values that yield the best performing model.
To better understand machine learning, let's investigate its three most common flavours: supervised, unsupervised and reinforcement learning
Supervised
Supervised learning is the most common flavour of machine learning in use today. Companies use it to predict employee performance, what product you’re likely to buy next, are you likely to repay the loan you are applying for and much more.
We use it to build models that predict categories or quantities based on some inputs measurements. so, if we are making a fruit and vegetable recognizer, the training inputs will be pictures and the training outputs the labels stating which fruit or vegetable is in the picture. The usage of output labels during training is where the name “supervised” comes from.
There are two major problem type in supervised learning: regression problem, when the output of interest is a quantity — such as length, weight or price; and classification problems where we want to predict categories, such as “ metal or plastic”, “positive or negative”.
Unsupervised
Unsupervised learning owes its name to the fact that at training time it makes
no use of the output labels — it is only bust with capturing the relationship and patterns in process inputs.
Reinforcement learning
It is most similar to the natural way in which living organism learn: an entity or an “agent” is taking certain actions in its environment and then adjusting its behaviour depending on whether the outcome of the action was positive or negative compared to its success criteria. although a very powerful idea and easy to intuitively understand, this domain of AI is still in its infancy, but significant efforts are being invested in research within its domain.
So far you have got the basic idea of what Artificial Intelligence and how it is operated under the hood.
If you enjoyed this article and learn something from it then don't forget to clap, it encourages me to write better.
