Sign in or Join the community to continue
What is AI today?
Posted Sep 10, 2026 | Views 0
# Large language models
# ChatGPT
# Claude
# Gemini
# Copilot
Share
Speaker
Makenna Bartow
LearnAIR Educational Content Designer @ LearnAIR
SUMMARY
Artificial intelligence turns vast amounts of data into patterns, predictions, and decisions, from recommendations and fraud detection to computer vision and autonomous driving. With tools like ChatGPT, Claude, Gemini, and Copilot, that same power becomes accessible to everyday people, helping us think, create, analyze, and work in ways that once required highly specialized technology.
+ Read More
CONTENT & TRANSCRIPT
0:00 Hello everyone, and welcome to our AI Basics video on “What is Today’s AI?” I think this is a term that encapsulates a lot more than people tend to think of.
0:11 Often, when we think of AI, we associate it with platforms like ChatGPT, but AI actually encompasses much more.
0:20 So, in today’s video, we’ll learn exactly what artificial intelligence is. Then we’ll apply that understanding to large language models, which will enable us to better understand how AI systems like ChatGPT work.
0:33 Artificial intelligence is the capacity for a machine to emulate human thinking, problem-solving abilities, and understanding.
0:43 It does this by using lots and lots of data to identify and predict patterns, which can result in behavior that resembles human thinking.
0:51 I have a few examples here of artificial intelligence that we see throughout our daily lives. The first one depicts recommendation systems.
0:59 Netflix, Spotify, Amazon, and many other systems use artificial intelligence to identify patterns. They might look at similar content that you have interacted with on the site, or look at users who have similar interests to you and identify patterns in the types of content they engage with in order to predict what you might like next.
1:20 In the next image, with the car and the license plate, if you’ve ever driven through a parking garage and it was able to recognize your car, it may be because it was using computer vision.
1:38 Computer vision is another type of AI. It can look at an image, break that image down into tiny squares called pixels, analyze the color values within those pixels, and use those patterns to create meaning from what it is seeing.
1:47 So that’s another type of AI: computer vision.
1:51 In the bottom-left image, we see a self-driving car. Think of something similar to Waymo cars in San Francisco. If you’ve ever been there, you probably know what I’m talking about.
2:01 These cars use machine learning and reinforcement learning to observe what is happening around them and use those observations to predict the next best move.
2:11 In this case, some of those observations might include a pedestrian walking, a cyclist, an upcoming traffic light, surrounding traffic, and the car ahead.
2:25 The system then uses models that optimize toward a goal. In this case, that goal would be to get the passenger to their destination safely without crashing.
2:36 The final image depicts fraud detection. This model is a type of machine learning that can use a classification algorithm to determine whether a transaction may be fraudulent based on patterns in the activity.
2:54 Hopefully, you can see through all of these examples that they rely on patterns and data.
3:04 And that brings us to our next category of AI: large language models. Large language models are trained on enormous amounts of information, including sources such as books, articles, websites, and other forms of text, in order to learn patterns in human language.
3:19 This is very empowering because you have access to this technology through tools like ChatGPT, Claude, Gemini, Grok, and Copilot. You have access to systems that have learned from enormous amounts of information, and that is incredibly powerful.
3:32 Large language models are built using neural networks, which were originally inspired by the structure of biological neural networks in the brain.
3:42 During the training process, the model takes all of that data we just mentioned, such as text from books and websites, and breaks the text down into smaller pieces called tokens.
3:56 Those tokens are then represented numerically so the model can process them mathematically.
4:10 Those numerical representations move through many layers of the neural network. In diagrams like this, you’ll often see little circles representing neurons or computational units.
4:26 Connections within the network have numerical values called weights. These weights influence the calculations the model uses to predict what should come next.
4:39 During training, enormous amounts of text are passed through the model. The model makes predictions, compares those predictions with the expected result, and then adjusts its weights to improve.
4:52 If you repeat that process enough times, across enormous amounts of data and billions of parameters, the model becomes increasingly capable of predicting patterns in language.
5:03 This training process is one reason these models are so expensive and time-consuming to build.
5:17 Once a company like OpenAI or Anthropic has trained a large language model, developers can build a product layer around it.
5:31 That product layer contains the features we interact with in these systems, such as memory, file uploads, web tools, projects, and many of the other capabilities we talk about in our training.
5:43 So when you combine the model, which you can think of as the brain behind the system, with the product and user interface, you get tools like ChatGPT, Gemini, Claude, and Copilot.
5:55 I hope that after this video, you appreciate these tools a little more deeply.
6:01 The fact that we can go from numerical representations and mathematical calculations to technology that can actually support our work and save us human time is truly incredible.
+ Read More
1
Comments (0)
Popular
Watch More
2026 model comparison video
Posted Feb 04, 2026 | Views 62
# Model comparison
# Digital employee
# Image creation
# Managed memory
# Personalization