Bruh, this isn’t a complaint about GPT-4, I’m just pointing out that it’s Not Magic

My phone takes pictures that are better than those of professional photographers from 20 years ago. And with those photos, it can zoom in on areas of the photos that are so sharp and clear that I can see every single one of my hairs or the downy hairs on a bird’s face. <br><br>This is the result of the camera having a ton of megapixels. <br><br>When I speak to my phone, it can select from hundreds of thousands of examples of words (even spoken by different people and with different inflections), giving high probability guesses at what words I’m saying. <br><br>Throw in a few decades of work on speech recognition, some machine learning algorithms, and the incredible computational power of this tiny device, and it can understand what I’m saying pretty well. <br><br>The GPT-4 model changes the paradigm slightly, in that it’s not just recognizing words from an existing set of examples, but predicting what word should come next, given the words that came before it. <br><br>This is still just those 2 components: a model of the words we use, and the power of a computer to perform predictions based on that model. <br><br>That’s also what’s behind my phone’s keyboard’s ability to suggest what word should come next as I type. <br><br>I know those examples are weak and simplistic, and of course OpenAI’s model is far more sophisticated and technologically impressive. <br><br>I’m just trying to say that while this technology is amazing, we should be able to see behind the curtain. It’s not magic. It’s not a person we’re talking to. <br><br>It’s such an overwhelming experience to first engage with GPT-4, and I think many of us take for granted what an amazing technological achievement this is, as we all fantasize about a world where we have a new friend or companion that we can talk to. <br><br>It’s just technology. Technically impressive, and a great foundation for a bunch of improvements over the coming years, but not magic.