To Understand How ChatGPT Works You Must Understand Neural Networks

  

Above is a video of Stephen Wolfram discussing ChatGPT. He talks in great length about how ChatGPT works. This is an informative video that gives the “soup to nuts” view of ChatGPT. 

To understand how chat GPT works, one must understand the concept of a neural network. I read a book that was published in 2009 on neural networks and the title was “Wired for Thought: How the Brain is Shaping the Future of the Internet” by Jeffrey Stibel. This was the first book to expose me to the idea that both computer networks and brains are similar in how they function. Stibel goes on to argue that both systems, the brain, and the Internet, operate based on the principles of networking, feedback, and scalability. He goes on to say that the way the brain rewires itself is very much like the way the Internet handles digital congestion by rerouting traffic to less congested areas of the Internet. 

What is a neural network? Imagine you have a big puzzle to solve, and you have a bunch of friends who can help. Each friend is good at solving one part of the puzzle, and together they can figure out the whole thing. 

In the brain, a neural network is like that - it is a group of little helpers (called neurons) that work together to solve a problem. Each neuron takes in some information, does some thinking, and then gives an answer. The answers from all the neurons are combined to get the final answer. 

For example, let us say you want to teach a computer to recognize pictures of cats. You would show the computer lots of pictures of cats and non-cats, and the neural network would try to figure out what features are common to all the cat pictures (like pointy ears and whiskers) and what features are different from the non-cat pictures. 

It is like a big game of "spot the difference" - the neural network looks for little details that are unique to the cat pictures and uses those details to make a guess about whether a new picture is a cat or not. 

That is the basic idea of a neural network - a bunch of little helpers working together to solve a problem by looking for patterns in the data they are given. 

In the book, “Wired for Thought,” Stibel’s goes on to mention how technology and neural networks are going to impact future society. Stibel implies that as technology advances, society may have greater cognitive abilities and the possibility of a “new type of hybrid intelligence.”

 

What does a neural network have to do with ChatGPT? When one first interacts with ChatGPT, the text is turned into numbers and fed into ChatGPT neural network. ChatGPT neural network processes this input and makes a “best guess” about what word comes next by examining large data sets about things that are written on the Internet by people and machines. As we know, everything on the Internet is true, so ChatGPT can only be as reliable as its data. 

With the advent of the Internet, teachers had to realize they were no longer the smartest thing in the room. This was a wide awakening for some teachers. With the possibility of hybrid intelligence not only is the teacher not the smartest thing in the room, very shortly, we may not be the smartest person in the room. But this is ok. Educators must give up the notion that we are teachers because we are "smart." Further gone are the ideas that people got into teaching to be in control. Have you been in a highly functioning classroom lately. It should be controlled chaos and most of them are.

Teachers should be in classroom now and forever because we should be providing
opportunities for cooperation, collaboration and conceptualization. Schools should expose students to topics and content that they have yet to encounter and may enjoy or excel at. Teachers need to teach students how to be safe, to share ideas and objects and participate in conversations and debates. Students should have the opportunity to feel the richness of learning even if they can not feel the richness of wealth or peace in their own homes.

Free Resources for Teachers: Project Look Sharp

Yesterday I went to a workshop at Erie 1 BOCES in West Seneca New York. I was really looking for resources on how to get my father-in-law to not just watch Fox News for his news of the world. But I got a lot of great resources and a reminder that we should be asking kids about the sources they're using. Project look sharp is associated with Ithaca College in Ithaca New York.

I think it's important to teach students to get different perspectives about an issue on multiple types of media.

I basically blame the insurrection of January 6th largely on the education system. I am struck by how difficult it is now-a-days to discern different sources of information and whether things are true or false. With the advent of social media there has been a shift in how people get their information. Social media companies, in the past, have done little to prevent the spread of misinformation. 

Heck. I am guilty of this too. In the last 12 years I have been a dog owner and have become an animal person. Some people would argue, “how can you be an animal person and still hunt?” That is a debate for another blog post. On Facebook, when someone’s dog gets loose or runs away from its owner, I would often, without looking, re-post the information about the dog. I spread a lie.

Fact check: Identical injured dog posts are a viral scam

Now, I just don’t repost as much stuff as I used to. A few bad apples have ruined it for the rest of us. 

Project Look Sharp is a free resource for everyone. One of the lessons I liked, since I was a past High School Social Studies Teacher, was the ending of the war in Afghanistan. The lesson looks at different newspapers from all over the world (written in English) and the students compare the source and the headlines that were written. This lesson teaches students to inquire about the source of the newspaper and the newspaper’s perspective. 

All of these lessons are based in the inquiry method of teaching, where the students discover, on their own, the answers. Teachers ask questions to get students to “go along the right path” to discovery. 

Below is a youtube video that displays a High School Inquiry around different media sources

I feel like if we as educators can do a better job of teaching students to be discerning adult customers of news, we can avoid future “January 6th” moments.

Why I left Twitter and went to Mastodon.

There are a few reasons. The first reason is Twitter has such an angry algorithm. Whenever you log in a Twitter you see a tweet by some angry person. Someone I have never heard of before. Twitter’s algorithm promotes speech that elicits responses both negative and positive. Anything that’s controversial moves up on your Twitter feed because of the algorithm. 

Another reason why I like Mastodon is that you are able to label certain posts as sensitive information or disturbing content. On Twitter it’s all just in your face. It’s not that way on Mastodon. Due to the majority of my social media time is now dedicated to Mastodon, I’ve not seen the video of the Memphis Murder. And I don't want to. 

Everyone on Mastodon has been completely nice and sane. When people would respond to me directly on Twitter, I would get a little “gun shy“ before I saw their tweet. I automatically assumed it was going to be something negative or an upsetting response to what I wrote. Often times it was. So far, direct mentions in Mastodon have been to keep the conversation going or to seek information. Not to annoy. 

Mastodon is also open source. No one can own it. It’s created by the people and for the people and is free. This has some good implications for its users. Third party applications that aggregate data and improve user experiences are more easily created since the code is open to all, specifically developers. Twitter just started “pulling the rug out from under some developers.” Things like Twitterific, Fenix and Twitterbot have all had their applications cut off at Twitter without much of an explanation. 

I also can’t support Twitter anymore. There are just too many people spreading lies. Lies that can actually hurt people and Twitter is doing nothing to police these lies now that has new ownership.

I’m not saying that Mastodon is not like Twitter; they are very similar. I’m also not saying that the things I dislike about Twitter won’t happen to Mastodon. For now, I need a timely, diverse and responsive personal learning network for my own growth. So, for now, Mastodon is that tool.