Over the past year, AI has taken root in almost every aspect of the online business realm. Most companies are experimenting with this using tools like ChatGPT and Midjourney.
However, few people are familiar with the basics of this technology or definitions of common terms. This can be especially problematic for entrepreneurs who plan to leverage AI in their business.
We will be happy to assist you.
Here's a comprehensive glossary of 15 AI terms you need to know to make informed decisions about which tools to use.
A.I.
Starting with the very basics, there is AI itself (short for artificial intelligence). It is commonly used as an umbrella term to refer to a vast range of technologies. AI is a computer program that processes incredible amounts of data to mimic the cognitive functions of the human brain, such as creating language and art.
computer algorithm
Then there is the term “algorithm” or “computer algorithm”. It writes a computer program that follows a set of instructions, such as performing a specific set of tasks to process and analyze data.
Some of the most influential algorithms are those used by social media platforms such as Facebook and TikTok. The platform uses these to analyze user behavior and decide what content to recommend.
machine learning
Another term often used in the field of AI is machine learning (ML). This describes a specific type of algorithm that processes large datasets to identify invisible patterns and connections. Facial recognition is a prime example of ML and, given enough data, can identify individuals in a crowd.
model
We're not talking about Gigi Hadid here. In AI, a model is a computer program trained by an ML algorithm to perform one specific task. It is specialized and cannot be deployed in isolation. For example, Midjourney is a model trained to generate images. That's great, but you can't suddenly start generating text.
Not all of them are huge. Today's smartphone camera apps use AI editing tools that run on the device itself. More and more smartphones are now equipped with dedicated AI chips just for this type of use.
Generation AI
When it comes to generation, “Generative AI” is another term that gets thrown around a lot but needs some explanation. While many AI algorithms are used to process and analyze data, the mission of generative AI is to create “new” content. From DALL-E to ChatGPT, these models have received a lot of attention in the media recently.
However, it is important to note that generative AI lacks true creativity. The “original” content created is based on patterns identified from large training datasets. These patterns are regurgitated and reshuffled to create “new” content, but contextual nuances are often missed.
training data
We've already used the term training data several times throughout this article. This is a word that summarizes all the data a developer feeds his AI to learn patterns. This data has a significant impact on the performance of the resulting model. The model only works on patterns that are present in the training data. Similarly, if there is a bias in the training data (e.g. regarding pronouns), the AI will adopt that bias.
supervised learning
I mentioned training the algorithm earlier. It is important to know that there are two ways to do this: supervised learning and unsupervised learning.
In supervised learning, you give your algorithm a clearly labeled training dataset.
For example, let's say you want to be able to identify crops in satellite images. To train it, feed it a series of GPS points labeled with the type of crop you want it to recognize, such as rice or mango trees. We then take this data, examine satellite images, and identify the characteristics that correspond to each type. This will allow you to re-identify the same crop in other images.
unsupervised learning
In contrast, unsupervised learning trains an AI to identify patterns in input data without using a human-created training set.
Staying with the satellite imagery example, you can also train your algorithm by clustering pixels with similar characteristics. For example, you can distinguish between flooded areas and areas where no flooding events have occurred.
Neural network / deep learning
Then there are artificial neural networks (ANNs) used in an ML technique called deep learning. Neural networks mimic the structure of neurons in the brain. An ANN consists of multiple layers of interconnected nodes, each representing a neuron. Every layer processes data independently and outputs a result that the next neuron receives. The largest models, like the one underlying ChatGPT, have billions of these nodes.
parameter
When training an AI algorithm, parameters are elements that can be changed. For example, in the unsupervised learning example, you can tell the algorithm you are training how many types of clusters to classify your data into. Similarly, characteristic weights and sorting thresholds can be determined.
Natural language processing (NLP)
One of the more difficult tasks that AI has mastered over the past few years is natural language processing (NLP). This term is used to refer to a specific type of AI that can understand and imitate written or spoken language. NLP is responsible for voice-activated devices, for example.
transformer
Transformers, introduced by Google in 2017, are a type of AI architecture that relies on “tokenization.” This process converts strings of symbols into data and analyzes patterns. This approach is the basis for most of today's prominent AI models. For example, the “GPT” in ChatGPT stands for Generative Pre-trained Transformer.
token
Tokens are elements that are transformed to serve as input data for AI. For example, ChatGPT prompts are converted to tokens before being processed.
hallucination
Hallucinations are the output of an AI that seems plausible at first glance, but is actually complete nonsense. For example, a student submits a paper generated by her ChatGPT, only for the professor to discover that some of the papers cited in the work are completely fabricated.
Application Program Interface (API)
Finally, one common term often used in relation to AI is API (Application Program Interface). For example, the ChatGPT website communicates with the real model's API behind the scenes. This provides users with an easy-to-use graphical interface that they can utilize to send prompts without understanding the underlying code.
