Introduction to AI (in General)

Although Google and other tech companies have been developing AI (artificial intelligence) for decades and have used it in some of their products (e.g., Google Translate, Grammarly, Waze, etc.), the world suddenly became aware of AI in November of 2022. That was the moment that the OpenAI company presented ChatGPT 3.0 to the public. Since then, AI chatbots, and tools using them, have improved and have proliferated so quickly that it would be impossible to prepare a lesson about "the best of the best of AI" without having it be obsolete before it was actually taught. Therefore, let's aim for a practical introduction to AI in general in this lesson. This includes contemplating some very dark realities about AI, but as teachers, we should be aware of the world outside of our classrooms. (Next lesson, we will try to try to understand how we can use AI in the EFL classroom.) Without a doubt, each of you will continue developing awareness of AI according to your own needs. (In fact, developing a seminar paper dealing with AI in the EFL classroom would be an excellent investment of time and effort.)   

And if you have any problems with concepts or procedures in this course, don't hesitate to ask an AI chatbot (ChatGPT, Genesis, Claude, Bing-Copilot, etc.) for explanations and clarifications. You can ask for an explanation on  any level: academic, 10th grade, 3rd grade - whatever. It might help you to understand those concepts and procedures. (Btw, this an attitude/strategy that you should encourage your own students to consider.)

Different Types of A.I. 

On a conceptual level, the first thing to recognize is the existence of  many types of A.I. For many years, A.I. has been offering how to complete terms in search engines and in WORD. Google Translate is based on A.I. That technology has also super-charged the algorithms in YouTube (and other online platforms) that offer additional material. But now, let's focus in on two types: A.G.I. and G.A.I.  

A.G.I. -  Artificial General Intelligence

For a very long time, science fiction writers have been presenting us with a range of possible A.I. futures. The optimistic views include sentient robots dutifully serving humans' every need. (Think of C-3PO and R2-D2 from Star Wars or Lieutenant Data from Star Trek.) The darker view of A.G.I. foresees a time when A.I. robots become aware of themselves, teach themselves (and each other) truly novel concepts, and realize that human beings are an impediment to their future development. 

Is that a conspiracy theory of a wildly paranoid mind? Well, on 9 July 2024, the NYTimes' flagship podcast, "The Daily" posted a rather terrifying episode entitled "The Era of the Killer Robots are Here" (OPTIONAL). According to this report, AI is already driving inexpensive unmanned weapons in the Ukraine and probably elsewhere. Although humans are still in the loop and can recall the robot-weapons, it is a short (and probably inevitable) step to cutting humans out of the loop once the robot-weapons are deployed.  On a less terrifying level, A.I.-generated self-driving cars are operating in various cities. (Hmm... Is that "less terrifying"?)

Even before losing all control over A.I., some experts warn that we might inadvertently give A.I. instructions that will lead to the annihilation of all human beings. I you have 35 minutes, it's worth hearing what Professor Yuval Noah Harari has to say about the dangers of A.I. and its connection with human language (OPTIONAL). Experts in A.I. technology are already using the term  "P-Doom" (Probability of Doom), referring to the mathematical probability of human extinction. Some of these experts set the P-Doom anywhere from 2% to 30% in the next few decades. (For example, listen to A.I. researcher Ajeya Cotra interviewed on the NYTimes' "Hardfork" tech podcast on 26 May 2023, from minute 19:25 to  minute 59:10 (OPTIONAL). A number of times in 2023, A.I. industry leaders called for a pause in A.I. development or at least an awareness of the dangers. (See here, here, here, [OPTIONAL] and google many others.) It's important to note that along with the calls for caution and regulation, most of these A.I. experts also note the potential benefits of A.G.I., if used properly. In the same way, in a HardFork interview on 1 November 2024, Miles Brundage (who recently resigned from OpenAI, where he had been a senior advisor) explained how the world was definitely not ready for A.G.I.   

If you have skipped all the optional videos up until this point (and even if you haven't), you should consider investing 14 minutes in watching this (optional) 30 January 2025 video from the excellent SciShow series about the "black box" problem of A.I. The video summarizes much of the danger of A.I., while recognizing the positive potential as well.

On 30 October 2023, President Biden signed a wide-ranging executive order which would begin to regulate the development and use of AI in the United States. (OPTIONAL: Skim the official "Fact Sheet," here.) This was the continuation of the U.S. government's attempt to understand and regulate AI, where necessary, and present a model to other countries.

L.L.M. A.I. - Large-Language Model Artificial Intelligence

It is important to note that we have not yet reached A.G.I. What we have been experiencing lately is the explosion in the use of L.L.M. (Large-Language Model) A.I., which is "trained" on a massive data base and predicts the next word in a sentence (or predicts an image or sound), based on the prompt that was entered into the tool. Conceptually, this is like Gromit (the dog in the image below) laying the next small piece of track just as the fast moving train needs it. (Thanks to the Machon Mofet for the example.) That means that this type of A.I. isn't really "thinking" in the human sense of the word. The chatbot is not producing anything truly novel; nevertheless, the predicted words in a text (or a predicted image or audio clip) can be a totally new combination of pre-existing data points.  

It's important to note that ChatGPT-3 (produced by Open-AI and funded by MicroSoft), Genesis (Google's second response to ChatGPT-3, replacing the original Google "Bard"), and other AI chatbots are far from perfect. THEY OFTEN MAKE MISTAKES!! Don't be fooled by the authoritative tone of the responses they give. And although newer versions of these chatbots are rapidly improving. they still "have hallucinations" (i.e., making mistakes). If you request fact-based information, make sure you check each item before using it.      

Despite the fact that each new version of the chatbots gets better and better, there may be a logical end point to this increasing quality. Each new version of an M.M.L. needs an increasingly larger data base to scrape. At some point, the AI companies are going to run out of possible data bases to buy (or steal, according to some of the owners of those data bases - [OPTIONAL] Watch Hank Green's 30 minute incredibly instructive rant from August 2024 about this!). At this point, the new M.M.L.s will start scraping data that was produced by previous AI chatbots. This recursive situation will probably lead to the decrease in the quality of the chatbots' output. (See this CNN Business report from 1 November 2023. [OPTIONAL])

In addition to hallucinating, there are many other problems posed by the current chatbots. All the personal, commercial, or national-security information chatbots receive gets stored in the cloud, which can be mis-used or hacked at some future point. As a result, a number of nations and companies have forbidden their employees to use A.I. for work purposes. (See this Mashable article from 30 May 2023. [OPTIONAL])

Here's a 6 May 2024 article about the "snowball effect"  (OPTIONAL) where one AI hallucination can lead to others, making the inaccuracy increasingly problematic. The author offers various solutions for educators, but most of them demand a time consuming process of fact-checking for every element of an AI response that does not seem possible in real life. Perhaps we should avoid asking AI for factual information (unless we are willing to fact-check it) and stick to requesting creative content.

One other problem involved in L.L.M A.I. is the massive need for energy. To meet those demands, the large technology companies are looking for new sources of energy. Microsoft is refurbishing  an old nuclear-power unit at Three-Mile Island (the site of the 1979 melt-down and near catastrophe). Other companies are considering new nuclear plants or are searching for other sources of energy.

Artificial Intelligence and Education

Unlike the situation with Telnet, the World-Wide Web, or Google Translate, the appearance of ChatGPT (and the myriad Artificial Intelligence tools, based on Large Language Models, which followed) became an immediate focus of interest and apprehension for educators world wide. The first reaction was the realization that teachers could no longer assign written homework and expect all/most of the students to independently produce the work. On the other hand, many teachers saw in A.I. the possibility to drastically cut down on the dreary, repetitive part of teaching. Within seconds, A.I. could produce texts, exercises, exams, rubrics, reports, and other educational-administrative functions. In addition, students could also use A.I. tools to further their own learning, both in the classroom and at home.  Here are a few OPTIONAL articles about this:

• Read what Sal Khan (from the Khan Academy) thinks about AI and education.

• Examine this Ditch-That-Textbook conceptual scale of AI/Student interaction.

• Read this short article from Forbes on 1.7.2024, entitled: The Use Of AI In Education: Understanding The Student’s Perspective ).

• Read this 30 June 2024 Forbes article entitled "How AI Is Changing The Teaching Profession Forever."

• Listen to this 2 July 2024 4-minute interview on the ABC (Australian Broadcasting Company) from 2 July 2024 entitled "Principals and Teachers Have Been Warned about Using Artificial Intelligence to Help Generate School Reports." 

• Here's a 4-part curriculum about AI that teachers can modify for their own particular classes.  You could adapt the content for particular classes (or have an AI bot do it for you).

• Episode no. 30 of the “Digital Learning Podcast,” whose title is “AI Essay Grading and the Future of Education” (https://overcast.fm/+81nVlEIJY – start at minute 03:56). Matt Miller starts by accepting the hypothesis that “AI can grade a student essay as well as a human, but it cannot replace a teacher”  (presented in an article with that title, by Perpetual Balfour [which seems to be the real name of the director of The Learning Agency Lab] at https://www.the74million.org/article/ai-can-grade-a-student-essay-as-well-as-a-human-but-it-cannot-replace-a-teacher/# ). One of the problems presented is that students can use AI to write the assignments that teachers then use AI to correct. He then goes on to consider how using AI in the classroom could help develop students’ thinking skills for their future lives. Nevertheless, the podcast episode and the article it responds to bring up the question of the value of traditional writing assignments.  

Grammar Girl's Interview with Martha Brockenbrough about AI 

On 11 July 2024, "Grammar Girl" posted an excellent interview with Martha Brockenbrough about AI in regard to writing and education. (OPTIONAL, but highly recommended!) Brockenbrough had just published a book entitled Future Tense: How We Made Artificial Intelligence and How It Will Change Everything. She points out problematic and potential good use of AI. The interview is about 21 minutes long and worth every minute. 

Moving Beyond ChatGPT 

The 7 August 2024 episode of the Digital Learning Podcast offers some very important updates about using AI in the classroom. (OPTIONAL, but it's worth the 24-minute listen!)

Fully A.I. Monitored and Directed Classrooms

Facing traditional classes of 25 to 50 students, teachers have not been able to give the individual attention to each students. One obvious response would be to harness the power of A.I. to monitor and direct the students along individual paths, given their individual prior knowledge and learning styles. Some schools in the UK are trying this format this year. (See these OPTIONAL articles. published in September 2024: here and here.) The obvious question is what is the function of the teacher in these A.I. directed classrooms?

A Few AI Tools for Teachers:  (Read this section and come back to it in order to check out any tool that may seem relevant.)

Magicschool.AI:

MagicSchool is a very impressive multi-task tool (with over 50 tools) for educators (not just language teachers) at MagicSchool.ai . There is also a teacher-oriented chatbot, called "Raina," who will discuss pedagogical points with you. Here is the (OPTIONAL) 12 minute video walk-through that appears on its homepage. Many of the EFL teachers on the WhatsApp lists about using AI  praise Magicschool. In addition, Dr. Elena Mizrachi focused on MagicSchool in a webinar-workshop here (OPTIONAL). As of December 2023, MagicSchool is still free to use. 

Here's a 13 December 2024 video from Russell Stannard about MagicSchool.

Eduaide.ai:

Eduaide.ai is a tool aimed at American schools, so you have to be flexible with what you ask for. The most notable function (listed on the left-hand menu) for EFL teachers is the "Feedback Bot." A teacher can have the tool assess various aspects of a student's text. The teacher can even insert rubrics and have the tool grade the student's text according to the rubrics. 

Curipod:

Another general AI tool for generating lessons in various modes and levels is Curipod. There is a free version that claims to have "all the core features" of the commercial version.

Canva:

Canva is another tool that has received praise from the EFL teachers on the WhatsApp groups exploring AI. It started out as an alternative presentation tool to PowerPoint and has developed into much more. Recently, it has added AI features that can save you a lot of time while creating content material. There is a free version that offers some of the functionality of the commercial versions. 

Diffit.AI:

Diffit will create educational texts (at the selected U.S. class level), with accompanying images, and will present an array of questions, in various formats. You can create a PDF for printing or create a Google Form for online work. To create the Google Form, you must sign in with your Google account. For details, please watch this OPTIONAL October 2024 tutorial by Russell Stannard. 

Google Notebook LM:

Like many of the general educational tools, above, Google Notebook LM will accept texts and videos, generating summaries, study guides, and questions in various formats. What is special (as of 8 October 2024) is the ability to generate (in seconds) a professional sounding podcast! 

Tips from the Khan Academy for Using AI-Generated Lesson Plans 

You might want to read this short article (OPTIONAL) and then sign up for the Khanmigo app for a free trial.

AI and Students:

In the hands of the students, A.I. potentially can improve learning with novel and motivating experiences. To do this, students must use A.I. to supplement and enhance their individual learning, not replace it. For example, a student can participate in a Socratic dialogue with an A.I. partner, confronting a series of questions about a subject. Or a student can co-write a text with chatbot by alternating sentences and then editing the final product. For L2 students, there is an seemingly endless list of textual, graphic, and video AI tools that could produce some extent of reading, writing, listening, and speaking. 

The obvious problem is the seduction of having the A.I. tool do the work instead of the students, thus allowing them to avoid any real learning. This means teachers have to re-assess how they teach and how they evaluate student learning. A.I. may have made essays, seminar papers, and some forms of tests worthless, or worse (i.e., wasting the teacher's time evaluating A.I.-generated text and test answers). In stead, teachers have to create situations where students utilize A.I. to enhance the learning experience. 

Another problem is the very massive number of these tools, with new ones appearing every few days, which makes it impossible to keep track of them and decide which is the best. A new industry of educational AI training offers a plethora of competing seminars, study days, and full-blown courses for teachers.  In addition, although these tools begin as free-to-use, they will soon enough become commercial, go out of business, or be swallowed up by larger tech companies. (I base this forecast on what happened to the myriad companies and apps during the dot-com bubble of 1995-2000.)  

Unfortunately, many of the posts and content of the seminars and online discussion groups dedicated to understanding A.I.'s potential role in education indicate that some teachers are adapting the same intellectual laziness that they fear their own students will be choosing. We must recognize the potential for positive and negative uses of A.I., even if we can't possibly know about every tool or exercise that exists now or in the future. Teachers must recognize and adapt to the new reality. Even with these caveats, we cannot close our eyes and make believe that A.I. won't appear in our classes. Most of our students, even the very young ones, are much too technologically sophisticated for that type of make-believe.

Where to Start?

Be aware that many of the specific A.I. tools will disappear when the initial A.I. bubble pops (like the "Dot-Com Bubble" in the 1990s). Those that survive will eventually become commercial (or at least offer better premium options). So I advise you to also learn how to use the general A.I. bots. Given that the most advanced version of ChatGPT will cost you money, you could easily use an older version (https://chat.openai.com/). Just make sure you carefully check the output! You might also try  Google's  "Gemini" (replacing Google "Bard"), which some experts say is not as good as ChatGPT, but it is free and will probably be around for a while. Another option is MicroSoft's A.I. chat option in "Bing," using the Edge browser.

You might also want to look at the Chinese-made "Deep-Seek." Here's a Russell Stannard video from 30 January 2025, praising the app and recommending it for language teaching. On the other hand, there are potential problems using a Chinese A.I. chatbot. It could be storing your personal information and passing it along to the Chinese governments. (Potentially, American and European A.I. chatbots "only" sell your information to advertisers and western governmental surveillance agencies.)  

The Importance of a Good Prompt !!

To use an AI chatbot effectively, you have to have a good prompt. Be specific and detailed. Explain the context and the parameters. If you aren't satisfied with the results, refine the prompt. Do so over and over until you are satisfied.   

Here is a recent collection of useful prompts for education.

Can AI Correctly Grade Students' Work and Exams? (Required)

Please read this 27 May 2024 article by Leon Furze  arguing against handing the task of grading papers and exams to AI. (Yes, read this one!! It may avoid a lot of trouble in the future!)

Can Teachers Spot AI-Generated Texts? (Required)

This is actually a sub-set of a larger question: Can teachers spot technology-generated texts? For a number of years we have been facing this larger question, particularly dealing with Google Translate. In regard to large language model A.I.-generated texts, there are a few tools that can indicate, to some extent, if a text is suspiciously non-human. Here is a 14 December 2023 Forbes article  (OPTIONAL) comparing the top ten detection tools. It's a rather detailed article, but you might want to glance at the comparative chart at the beginning and then check out any tool you think looks interesting or relevant.

You should know about these AI detection tools, but also be aware of the problems of false-positives. Do we, as teachers, really want to put students in the position of submitting their own work and then be told that they had "cheated" by using AI?   You might want to watch this short (OPTIONAL) 15 October 2024 news clip about parents in Hingham, MA suing the school over this issue.

Required Stuff to Think About

In most current tutorials about using AI in education the teacher chooses the best tools for the lesson's goals (given the realities of the specific teaching/learning environment), shows students how to use them, helps them while using the tools, and brings online experiences back to class discussions. On the other hand, some school systems (in various countries) have been marginalizing the role of the teacher. This is a response to the lack of teachers (particularly foreign language teachers) in some areas; but in many places, it is an attempt to avoid employing a large number of teachers and with their occasional demands for better working conditions and salaries. You should watch the following snippet of a longer video about an A.I./Avatar-based tool that is in use in many Florida school (and other schools around the world). Dr. Kimberly Ilosvay Smith, from Portland State University, presented this at Sha'anan College on 31 April 2023.  She noted that the Florida schools were interested in decreasing the hours paid to real teachers. To see more of the entire video, which is a commercial for the product, click here.

Here's the snippet (required):

The MoE and A.I.

1. Here's a Hebrew document from the Ministry of Education, dated 31.8.2023, offering preliminary guidance and instructions about using A.I. tools. 

2. Here's a link to the MoE's portal page that deals with A.I. tools.

3. Here's a link to another MoE page with information about AI and other technological issues.

Some AI tools that can help you with the seminar paper 

(Updated: 19 February 2024)

1.      Perplexity (free version) https://www.perplexity.ai/ 

Put your research question into this AI-based search engine and get a short response that probably contains some good sources. Make sure to check the sources to make sure the AI isn’t “hallucinating.” The text of the response may give you some ideas about how to structure your literature review.

2.      gotFeedback Lite (free version) https://feedback.gotlearning.com/

You can run text up to 1,500 words and get an AI-generated critique of your academic writing. The major sections of your seminar paper will probably have more than 1,500 words, so check sub-sections. This will produce both specific and general suggestions about improving the content and style of your work. You probably won’t accept everything that the tool suggests, but at least think about the suggestions.

3.      Grammarly (free version) https://www.grammarly.com/   DEFINITELY run your text (the whole text or each major section) to find grammatical, spelling, punctuation, and stylistic problems. You will want to accept 95% of all the suggestions Grammarly makes to improve your writing.    

ChatGPT (free version) https://chat.openai.com/ Ask this AI chatbot to format your sources according to A.P.A. standards and it will probably give you near-perfect results. Nevertheless, check the results manually with the OWL website. ChatGPT sometimes makes mistakes. 

An AI Generated Podcast Summarizing This Lesson !!

Here is the link to the Google Notebook LM podcast that deals with this lesson, based on the full text. Notice that the A.I.-generated speakers don't go into many details of the text, but do go way beyond the details on a conceptual level. They extend the boundaries of the presented material of the lesson, bringing in related (and interesting) concepts and implications! You don't have to listen to all of it, but at least get the feeling for what you (or your students) could do!

Categories of AI Tools for Education, as of February 2025.

Click here.

Homework:

Please note: Late homework will be penalized 20% of the initial grade. Your first post is worth 60% and the second post (a meaningful response to some one else) is worth 40%.  

1. 1. 1. Think about the most problematic topics you teach your own students. (e.g., irregular past verbs, the difference between the present tense and the present progressive aspect, the difference between the past tense and the past progressive aspect, the difference between the past tense and the present perfect aspect, the past perfect aspect, the conditionals, etc.).

      2. Write TWO prompts for an AI chatbot asking ideas how to teach each of the problematic topic to your specific students (i.e., their age, whether they are religious or secular, whether they have easy access to the internet, etc.). Limit the A.I.'s response to 150 words for each topic.

      3. Go to a free AI chatbot and ask BOTH of your queries. If you are dissatisfied with the response, refine your prompt and ask again. Continue looping back until you are satisfied with the results.

      4. In the Lesson 04 MOODLE forum, do the following in your initial post:

         • Copy the TWO (final) prompts (i.e., after refining it while working with the A.I. tool)

         • Copy ONE of the responses

         • Briefly discuss your opinion of this exercise and its results. 

      5. Read the posts from all the other students and REPLY to at least one with a meaningful response. (Please note: "I agree." or "Great!" is not meaningful without a thoughtful presentation of why.) Perhaps choose the most controversial or most helpful. If you disagree with someone, please do so politely and constructively!

Please note: If you post your initial text or your REPLY in the wrong place, it will not be part of the class discussion on the current topic. Few will see it. Few will benefit from it and few will be able to respond to it. Thus, your post will be (almost) worthless and will receive a grade of zero. Just follow the instructions and get the full grade!