Video: Inside the Mind of AI Tools: How AI Tools Think

By now, you must have used generative artificial intelligence (AI) tools like ChatGPT, Gemini, and Perplexity somewhere in your everyday life. Ever wondered how these tools operate behind the scenes? 

In this video James Sonne (Anatomy & Neurobiology, Ph.D. Integrated Biomedical Sciences) breaks down the mind-bending world of AI “thinking” and explores why machines don’t operate like humans do.  

Here’s a surprising truth: AI does not think, plan, or have goals the way people do! It all depends on how the AI model is trained. Humans aim for a result, but AI starts at one point and predicts the most likely next step based on probabilities.  

We then dive into one of the coolest parts of AI: its vector space. Imagine a massive outer-space universe, but not in 3D or 4D. A casual chat with ChatGPT itself tells us that its internal “space” exists across 16,384 dimensions! Each word (or token) the AI knows is like a star floating somewhere in this gigantic high-dimensional galaxy. When you type a prompt, AI turns it into a vector, plots it in this huge space, and then begins hopping from one token to the next, choosing whichever one sits closest in meaning.  

This is why words like “cell” can link to “membrane” in a biology context, but connect to “prison” in a criminal justice context. It all depends on where your prompt lands in this massive 16K-dimensional map.  

If complex AI concepts boggle your mind, this video simplifies the explanation with easy-to0grasp examples. Whether you’re a tech enthusiast, a student, or just curious about how gen AI tools form their answers, this video will completely change the way you see artificial intelligence! 

Break that writer’s block! Check out Paperpal and simplify your academic writing with AI assistance.