Artificial intelligence was one of the most popular technology trends last year, and it shows no signs of slowing its colossal influence in 2024. This article is aimed at educating our readers about how to take advantage of the AI wave, authored by our Chief Technical Officer, Alex Stone.
via Giphy
Before we begin, don’t forget to connect with us on Facebook and X for more updates and discussions on AI and educational technology! You might also consider bookmarking this post to use as a reference.
Let’s dig in!
To launch our series, a bit of a primer is necessary. When I workshopped article ideas with other leaders at Filament, we quickly realized that even as long-time leaders in the edtech space, we actually didn’t have a shared understanding of the state of the art of AI, or even necessarily what AI is or is capable of now. I suspect this situation is playing out at companies large and small inside our industry, with a few notable exceptions where strong ML teams have already been built, for example, the fine folks at ETS AI Lab, and now Curriculum Associates through their acquisition of SoapBox Labs.
This article is just as much for my colleagues at Filament as it is for everyone else out there wading through the jargon soup. To begin, let’s take a survey of the types of AI models that exist today. Understanding the breadth of AI techniques will help you identify the opportunities that are relevant to your organization.
The most popular AI models follow into one of these categories:
- Generative Models
- Planning Models
- Classification Models
Generative Models conjure up text, images, and even music. You likely have interacted with them before without even realizing it! These are the most popular types of Generative Models:
Large Language Models, like ChatGPT, are the wordsmiths of AI. Their job is to understand and produce human-like text, making them useful for tasks like natural language processing and creative writing. State of the art LLMs also have demonstrated emergent capabilities of step-by-step planning, code generation and instruction following, making them useful to automate non-hazardous tasks.
This is actually the technology that underlies our AI-powered digital robotics prototype TAVIX, which acts as an engineering and robotics assistant in our robotics sandbox game RoboCo. You can check out the demo below:
Ever wonder how Siri or Google Assistant speak with such clarity? That’s thanks to Speech Synthesis, which converts text into spoken words. But it’s not just about Text to Speech – there’s Speech to Speech synthesis too, exemplified by tools like Respeecher. With advancements in technology, speech synthesis has evolved to sound more natural and lifelike than ever before.
Media Synthesis encompasses the creation of music, images, and video. Some examples of this model are DALL-E, Stable Diffusion, and Sora.
Unsung heroes, Planning Models are the doers of the AI world. They are not as headline grabbing as DALL-E or ChatGPT because they don’t usually interact with users directly. They are busy focusing on the bigger picture or controlling processes as they unfold over time. They’re behind autopilot systems, automated driving, and factory automation, orchestrating complex tasks with precision and efficiency. Paired with other AI types like Computer Vision, they bring intelligence to action.
Classification Models discern patterns and identify objects in the digital landscape. From Computer Vision’s role in autonomous vehicles to Speech Recognition’s assistance in virtual assistants, these models understand and interpret our world in real-time.
This is what keeps your car in its lane on the freeway, and how your phone tracks your location in augmented reality experiences. It’s also used in cancer screening technology. If you want to use Computer Vision in your edtech product today, check out these two platforms: V7 Labs and Viso Suite.
Speech Recognition is how Siri and Google Assistant understand what you are asking. It’s also used for dictation. Often this involves Speech to Text, but it’s possible to feed speech audio signals directly into semantic models.
Natural Language Processing analyzes text to extract useful information. Whether it’s classifying names and locations, summarizing articles, or gauging sentiment, this type of AI helps us make sense of the vast sea of textual data. NLP is often used to select between a set of possible commands or to assist in grading open-ended assessment questions.
Recommendation Engines suggest products, content, and experiences based on your preferences. Examples include Netflix’s movie recommendations and Amazon’s product suggestions – these engines help streamline decision-making in a world overflowing with options.
Regression Models are the fortune-tellers of AI, forecasting outcomes and trends. Used in inventory planning, investment strategies, and predictive analytics, these models provide insights into the future, helping businesses make informed decisions.
These were the key categories of AI models that I came up with based on my experience and research. Did I forget one? Reach out and I’ll revise my list! Next time, we’ll look at how different types of models can be put to work solving problems inside learning games and other digital experiences.
Ready to explore the endless possibilities of technology-driven learning? Reach out to us today about your game-based learning project!
More on AI: