June 6, 2024
I have decided to update this in a blog format so that previous insights remain available, but new ideas can rise to the top. I typically get my information through the One Useful Thing blog and invite all of you to subscribe. Ethan writes in a way that even I can understand. Today’s article brought the state of AI up to date, but as Ethan points out, by publication it was probably already out of date. This is an incredibly fast-moving target. I do not hit on nearly everything covered in One Useful Thing. My goal is to provide you links to the various AIs and ideas for how to use them. There are quite a few out there now and each has strengths and weaknesses.
Today’s ideas are about things you can do with research. Input was fed to a series of AIs and asked for a variety of things. For example:
Music! Suno wrote a song based on prompts I gave it about oil boomtowns (yes, I am a one-trick pony). It was easy to figure out, the lyrics were visible, and simple to download. My oil boom song is here. Udio also wrote a song. I never found lyrics and I had to download it instead of getting a link. Both provided 30 second samples, but there was an option to extend the length. Both also had an option to reuse the prompt, so if you don’t like what you got, get something completely different. I encourage you play with both and keep tweaking until you get what you want. I only tried one on each site.
Illuminate is still in the beta stage. You can’t submit your own papers, but you can see how it works. It takes a full length paper and turns it into a conversation. It has multiple voices, rises and falls in tone and breath, and interactions between the voices. It makes sense as if you were following an actual conversation. I am really looking forward to this. Dusty, dense research can easily be turned into something more interesting.
Now, I will revisit the core uses of AI and the current best tools for that. Summarize, generate ideas, write outlines, produce reports, create strategic plans, and so much more. As always, I used the prompt “explain the 1917 oil boom in Ranger, Texas. Include citations.” The full text of each response is attached to the title.
Claude 3 Sonnet: Not as accurate as Gemini. There were only two citations, but both were real books well suited to accurate research.
Gemini 1.5: For what it’s worth, you cannot log onto this site through Ranger College. I have to open an incognito window and log in with my personal email address. I do use it frequently at work, however. The information today was generally accurate. The citations were real, but shallow. It even used Wikipedia as a source.
Chat GPT: This does not seem much improved over the model we started this process with two years ago. Neither citation, provided by hyperlink, worked. The TSHA article was easy enough to find by searching the TSHA site once you were there, but the Legends of America site had no article on Ranger at all.
LLMs change so fast that it is hard to keep up. I subscribe to a blog called One Useful Thing to help me. This morning (December 7, 2023) he published a blog called An Opinionated Guide to Which AI to Use: ChatGPT Anniversary Edition. Surprisingly, it is one year since ChatGPT took the world by storm. The blog discusses what has changed and what is coming. I have been attempting to keep up, but was shocked by what has changed. The bottom tab in this Guide is called Other AI Programs. Mollick only mentions two of the ones I had listed. I will have to investigate what is still alive and what has failed.
The most important takeaway from the article, which everyone interested in AIs should read, is the huge quality difference between ChatGPT 3.5 (free version) and ChatGPT 4 (paid version). Bonus – he explains a few ways to access ChatGPT 4 for free. Every one of his blogs provides insight into great ways to use AIs to increase productivity, but using ChatGPT 4 seems to be the most important aspect.
The most important takeaway from this post is that this page, and every other page you look at, is probably already out of date. That said, read on – there is a lot to learn.
What is a Large Language Model Artificial Intelligence?
A large language model (LLM) is a type of artificial intelligence (AI) algorithm that uses deep learning techniques and massively large data sets to understand, summarize, generate and predict new content.
LLMs have become increasingly popular because they have broad applicability for a range of NLP tasks, including the following:
*https://www.techtarget.com/whatis/definition/large-language-model-LLM?Offer=abt_pubpro_AI-Insider
This article explains many more things AI can do.
What is ChatGPT?
ChatGPT is an AI chatbot system that OpenAI released in November 2022 to show off and test what a very large, powerful AI system can accomplish. You can ask it countless questions and sometimes will get an answer that's useful.
For example, you can ask it encyclopedia questions like "Explain Newton's laws of motion." You can tell it, "Write me a poem," and when it does, say, "Now make it more exciting." You can ask it to write a computer program that will show you all the different ways you can arrange the letters of a word.
Here's the catch: ChatGPT doesn't atually know anything. It's an AI that's trained to recognize patterns in vast swaths of text harvested from the internet, then further trained with human assistance to deliver more useful, better dialog. The answers you get may sound plausible and even authoritative, but they might well be entirely wrong. More significantly, AI programs like this are not connected to the internet, doing searches in the traditional way. They only know what they are trained with, which may be two or more years out of date.
https://www.cnet.com/tech/computing/why-the-chatgpt-ai-chatbot-is-blowing-everyones-mind/
What is the problem with Artificial Intelligence in the classroom?
There are several significant concerns with AIs like these. The first, and probably most important, is that it gets its information from its programmers. It is fed all the available information from the general internet at a particular point in time. Occasional updates might occur, but in general, the information might be very old. Additionally, any information fed to it was derived from the general internet, which is notoriously unreliable. Imagine the damage this can do with bad medical advice. Here at Ranger College we expect you to do enough quality research to find accurate answers to your questions.
Along a similar vein, AI is subject to bias. Since it is drawing on the collective writing of millions of humans, past and present, it picks up biases as fact.
AIs can provide bibliographical information if requested, but it is often entirely made up. AIs do not incorporate journals, paywalled articles, or peer reviewed material.
Invention, or hallucinations, are a common issue with AI programs. Material is routinely invented based on use of keywords. This is a serious issue for the quality of the information.
AIs can struggle with the use proper language and grammar. That aspect is improving faster than other problems, but it often sounds as if it were written by a non-native English speaker.
AIs cannot tell between blatantly false information and real information. Information literacy skills have always been critical. When someone uses AI information without verifying the accuracy, false information gains ground even faster.