Serendipity35

When I was teaching in a high school, I used to tell students (and faculty) that we were not preparing them for jobs. I was sure many of our students would end up in jobs with titles that did not exist then. There is a song by The Byrds from the 1960s titled "So You Wanna Be a Rock 'n' Roll Star." In 2024, it could be "So You Want To Be An AI Prompt Engineer."

The role of AI prompt engineer attracted attention for its high-six-figure salaries when it emerged in early 2023. What does this job entail? The principal aim is to help a company integrate AI into its operations. Some people describe the job as more prompter than engineer.

There are already tools that work with apps like OpenAI’s ChatGPT platform that can automate the writing process using sets of built-in prompts. Does that mean that AI will replace AI prompt engineers already? For now, the prompter works to ensure that users get the desired results. They might also be the instructors for other employees on how to use generative AI tools. They become the AI support team. AI can automate "trivial" tasks and make more time for work that requires creative thinking.

What kind of training leads to getting this job? You might think a background in computer science, but probably a strong language and writing ability is more important. People who write in the corporate world might justifiably fear AI will take their jobs away. Being a prompter might be an alternative.

Still, I suspect that there is a good possibility that a prompter/engineer's job might be vulnerable as software becomes better at understanding users’ prompts.

If you are interested in being an AI prompt engineer, I posted last week about some free online courses offered by universities and tech companies that included three courses that relate to creating prompts for AI.

AI Applications and Prompt Engineering is an edX introductory course on prompt engineering that starts with the basics and ends with creating your applications.

Prompt Engineering for ChatGPT is a specific 6-module course from Vanderbilt University (through Coursera) that offers beginners a starting point for writing better prompts.

Another course on ChatGPT Prompt Engineering for Developers is offered by OpenAI in collab with DeepLearning and it is taught by Isa Fulford and Andrew Ng.  It covers best practices and includes hands-on practice. 

If you are interested in taking some free AI courses offered by Google, Harvard, and others, here are 8 you might consider on a variety of approaches. For Coursera courses without the trial, go to the course you want to take and click 'Enroll for free', then 'Audit the course'. You'll need to create an account to take courses, but won't need to pay anything.

Google offers 5 different courses to learn generative AI from the ground up. Start with an Introduction to AI and finish having an understanding of AI as a whole.  https://lnkd.in/eW5k4DVz

Microsoft offers an AI course that covers the basics and more. Start with an introduction and continue learning about neural networks and deep learning.  https://lnkd.in/eKJ9qmEQ

Introduction to AI with Python from Harvard University (edX) is a full 7-week course to explore the concepts and algorithms of AI. It starts with the technologies behind AI and ends with knowledge of AI principles and machine learning libraries.  https://lnkd.in/g4Sbb3nQ

LLMOps are Large Language Model Ops offered by Google Cloud in collaboration with DeepLearning. Taught by Erwin Huizenga, it goes through the LLMOps pipeline of pre-processing training data and adapt a supervised tuning pipeline to train and deploy a custom LLM.

Big Data, Artificial Intelligence, and Ethics is a 4-module course offered by Coursera from the University of California - Davis that covers big data and introduces IBM's Watson as well as learning about big data opportunities and knowing the limitations of AI. I think the inclusion of ethics is an important element.

AI Applications and Prompt Engineering is an edX introductory course on prompt engineering that starts with the basics and ends with creating your applications.

Prompt Engineering for ChatGPT is a specific 6-module course from Vanderbilt University (through Coursera) that offers beginners a starting point for writing better prompts.

Another course on ChatGPT Prompt Engineering for Developers is offered by OpenAI in collab with DeepLearning and it is taught by Isa Fulford and Andrew Ng.  It covers best practices and includes hands-on practice. 

Recently, a former colleague asked me what I thought about push versus pull learning. I knew the terms more from social media marketing but hadn't really used them in learning situations. In marketing, examples include whether to decide to subscribe to a newsletter by email or snail mail (you pull that information by choice) or a newsletter that comes to you automatically (it is pushed at you).

In general, I think people prefer to pull (choice) over having it pushed at them. Companies might prefer to push, but that probably comes with the option to stop that push (unsubscribe.)

Moving these approaches - or just the terms - to education makes some sense.

In a push approach, teachers decide on the information, approach, delivery method, and speed of delivery. It is how education has been done for centuries. It tends to start with what Bloom and his taxonomy would categorize as knowledge-level remember and understand questions. These would build toward more critical and creative thinking. With pull, students enter into creating, evaluating and analyzing that requires them to seek knowledge and understanding.

This conventional classroom-styled learning is not the only approach in the 21st century. Pull learning allows learners to access information at the point of need, the way they prefer (in some settings) at the speed they find comfortable. I think that the initial surge of MOOCs back in 2012 is a good example of learning that learners pulled as needed.

Pull puts learners more in control It flips the teacher-centered learning setting. However, we must acknowledge that learning in school at all levels is still very much push learning. Fortunately, the idea that students should be able to pull some learning as they feel they need it is gaining more acceptance and is being incorporated in instructional design planning.

Currently, pull learning experiences are probably best suited to workers who have learning needs based on job roles, personal knowledge, and advancing their career interests.

Ideally, learning is "push-pull" with appropriate information provided by a push and additional information required to complete tasks and goals pulled as needed. This is not really a new approach. When you were a student, you were certainly pushed information, but you might well have gone beyond what was provided and pulled additional information that you felt you needed.

MORE
https://www.responsiveinboundmarketing.com/blog/the-difference-between-push-and-pull-learning

https://www.teachthought.com/education/push-teaching-vs-pull-teaching-thinking/

https://barkleypd.com/blog/pushing-or-pulling/

Andrew Ng's Machine Learning course on Coursera has been revamped and updated and it is getting good student ratings.

There are fewer online courses that I consider to be true MOOCs now. Massive is small. Open is more closed. But the "OC" portion remains for many. The three courses that make up the Machine Learning Specialization offered by DeepLearning.AI and Stanford on the Coursera platform still fit the MOOC definition more closely.

You can earn a certificate at the end, and enjoy the full experiences including quizzes and assignments if you enroll and pay a monthly subscription but the courses are free (Open) to audit and view the course materials. The Massive in this course is massive with over 20,000 students enrolled.

Andrew Ng is the co-founder of Coursera and was the founding lead of Google’s Brain Project, and served as Chief Scientist at Baidu. He then did two artificial intelligence startups - Deeplearning.ai (a training company founded in 2017) and Landing.ai (for transforming enterprises with AI). He remains an adjunct professor at Stanford University. His course on Machine Learning was one of the very first courses from Coursera when it first launched in 2012. I audited the course that year though I knew the content was way above my abilitiees but I was curious as to the structure of the course from a design perspective.

At that time, Machine Learning was a new concept and was close to applied statistics. Ng goes way back because his Stanford lectures were on YouTube in 2008 and got 200,000 views. Then, he converted them to an online format in Fall 2011 and they were offered for free. He had 104,000 students and 13,000 of them gained certificates.

On the tech side, this updated version:
uses Python rather than Octave
expanded list of topics including modern deep learning algorithms, decision trees, and tools such as TensorFlow
new ungraded code notebooks with sample code and interactive graphs to help you visualize what an algorithm is doing
programming exercises
practical advice section on applying machine learning based on best practices from the last decade