How AI Works: From Sorcery to Science
How AI Works: From Sorcery to Science by Ronald T. Kneusel
My rating: 4 of 5 stars
I think this book is a good introduction into how AI works. It’s concise and it explains the history as well as the neural networks and modern Large Language Models (the history part is more fun to read than it sounds). The end about using LLMs is not that good, but overall this book is a fine 4/5.
The liked:
My favorite chapter has to be the chapter explaining the image classification and image generation models. This was a really interesting application of the neural networks and I’m glad I understand it a little more. The principle of how the diffusion model works is also so fucking cool!
I’m also glad I got to read about the Symbolic AI models. I wish they were more precise so I could use them in practice - they are really nice algorithms.
The disliked:
The last two chapters about LLMs aren’t great. They aren’t about how LLMs work and it’s specifics, but rather a showcase of something it can do. I skimmed over most of this part, because I am not interested in reading AI generated stories or code. This chapter is basically ancient at this point.
The next chapter does not save this book either as it’s full of AI hype and speculations about what it can do. First of all I agree that AI does have some emergent behavior that you won’t expect, because it might have it in it’s training data.
However, at one point the writer wrote that the AI could simulate another AI model (Random Forest 300) in it’s neurons and paste the output. He even asked the AI to confirm if this is the case (the AI of course confirmed it could do this, because that’s just what AI does). That’s just plain bogus :D - a small AI psychosis in this kind of book is dangerous.
The summary:
Yeah, all in all it’s a good book. I received a good overview of this technology and it was a useful supplement to the practical understanding I got from simply using this technology. It won’t teach you everything, but it will give you greater understanding of all the different AI technologies.
Use your head when reading the last two chapters and think critically.
View all my reviews