A.I. Crash Course: Systems
In a previous article, we talked about the six pillars of artificial intelligence:
- Processing
- Information storage
- Algorithms
- Data separation
- Neurons
- Basic Logic
Those come together to form a system and we call this system a Neural Network. This system is the first full assembled building block of other systems. Think of it like chemistry: the broad use case and study in chemistry is the interactions between elements and molecules, so while there are subatomic particles and components, that area of focus is specialized and not as relevant to the widespread usage of elements and molecules chemical sciences. Neural Networks are like the atoms in chemistry. You learned the subatomic particles because those have bearing on the bigger picture, but ultimately, I brought you through this part of the journey for comprehension, not practical use. In the next section of this discovery, I will only refer back to these sub-components in Neural Networks (which you’ll often see abbreviated as ‘NN’) when important or relevant such as when I discuss weight and bias.
If you understand a NN, then let’s cover some more ground, because these terms are often used interchangeably which they shouldn’t be, and this is where a lot of confusion is derived and why many people (even the ones who know better) simply say “A.I.”
Models and Training
Training a model (which is broadly: a collection of Neural Networks) refers to the process of adjusting the parameters of a machine learning algorithm in order to improve its performance on a specific task. This process typically involves presenting the model with a large dataset and allowing it to learn from the data by adjusting its internal parameters. The goal of training a model is to achieve the best possible performance on the task for which the model was designed.
There are several different approaches to training a model, and the specific approach used will depend on the type of model and the task it is designed to perform. For example, some models may be trained using a supervised learning approach, in which the model is presented with a dataset that includes both input data and corresponding labels, and the model is trained to predict the labels for new data based on the input data. Other models may be trained using an unsupervised learning approach, in which the model is presented with a dataset that includes only input data, and the model is trained to identify patterns or relationships within the data.
Regardless of the approach used, the process of training a model involves adjusting the model’s internal parameters in order to minimize the error or loss function, which is a measure of how well the model is performing on the task. This process is often done using an optimization algorithm, which adjusts the parameters of the model in a way that reduces the error or loss function. Once the model has been trained, it can be tested on new data to evaluate its performance and make any necessary adjustments to improve its accuracy.
Machine Learning
Is a field of computer science that focuses on the development of algorithms that can learn and adapt over time, without being explicitly programmed. Machine learning is based on custom-built neural networks that have been programed to classify data, make predictions, and perform a myriad of other tasks. Often times when used in the context of industry or application, this is the form of A.I. that one is actually referring to. ML is a simpler system that is often used to perform tasks that would be difficult or impossible to do manually. It is also worth noting that neural networks are only one type of a machine learning algorithm, but there are many others, including decision trees, support vector machines, random forests, and more.
Deep Learning
Deep learning is a subfield of machine learning that focuses on the use of complex neural networks, which are ones with many layers or complex synapses. Deep learning algorithms are particularly useful for tasks that require the processing of large amounts of data which don’t have clear linear separation, such as image and speech recognition.
So WTF is A.I. then?!
Artificial intelligence is a broader term that refers to the development of systems that can perform tasks that would normally require human intelligence, such as problem-solving, decision-making, and learning. AI can be divided into two main categories: narrow or general. Narrow AI is designed to perform a specific task, while general AI is designed to perform a wide range of tasks. Neural networks, machine learning, and deep learning are all subfields of AI.
It is important to note that these systems are not separate things, but rather each is a small system that is embedded within a larger system. When two subject matter experts want to communicate a wealth of information succinctly, this is where the term AI comes into play.
I’ll give a real-world example of why this terminology matters and how I can differentiate between someone who uses “A.I.” as a buzz word or not. Unicorn has been an investor and built the MVP for a startup called ReVision, which is basically like Grammarly for bias detection. If I was talking to another engineer or a tech-savvy investor I would probably describe it like this:
We built our bias-detection AI with two primary layers: a supervised, general generative pre-trained transformer (GPT) model which lets the user thumbs up/down results. It also has an unsupervised narrow model which is powered by semantic data sets.
There’s a lot of subtext in that statement that we’ll unpack. But note that I said “model” twice and I specified “general” vs “narrow.” The term model is an indication of a machine learning algorithms powered by neural networks. You can reasonably assume that if you hear someone say “AI” and “model” in the same breath, they are assuming and drawing from all of these principals you just learned.
In summary, neural networks, machine learning, deep learning, and artificial intelligence are all related concepts that are used to develop systems that can perform tasks that would normally require human intelligence. While each of these systems has its own specific components and characteristics, they are all interconnected and interdependent, with each serving as a building block for the larger system.
Subscribe to get on the waiting list and early access to our upcoming book. If you want leading experts to guide your business in AI, technical patent work, or consulting, get in touch with the Unicorn team.
