AI in Higher Ed: the Best Is Yet to Come
AI in Higher Ed: the Best Is Yet to Come
UPCEA Annual Conference 3/23/2023
Ray Schroeder
UPCEA Senior Fellow
UPCEA Annual Conference 3/23/2023
Ray Schroeder
UPCEA Senior Fellow
(Can you suggest something else of this sort that has or will touch nearly every industry, business, career?)
Udacity CEO and Co-founder, Stanford Professor Sebastian Thrun joins Becky Anderson on CNN International and shares his thoughts on the impacts of AI and what that means for the future of education.
Artificial Intelligence - the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages. (https://www.lexico.com/en/definition/artificial%20intelligence
Machine Learning - Machine learning is an application of artificial intelligence (AI) that provides systems the ability to automatically learn and improve from experience without being explicitly programmed. Machine learning focuses on the development of computer programs that can access data and use it... learn for themselves. [https://expertsystem.com/machine-learning-definition/
Deep Learning - In practical terms, deep learning is just a subset of machine learning. In fact, deep learning technically is machine learning and functions in a similar way (hence why the terms are sometimes loosely interchanged). However, its capabilities are different. While basic machine learning models do become progressively better at whatever their function is, but they still need some guidance. If an AI algorithm returns an inaccurate prediction, then an engineer has to step in and make adjustments. With a deep learning model, an algorithm can determine on its own if a prediction is accurate or not through its own neural network. https://www.zendesk.com/blog/machine-learning-and-deep-learning/
Algorithm - In computing, an algorithm is a precise list of operations that could be done by a Turing machine. For the purpose of computing, algorithms are written in pseudocode, flow charts, or programming languages. . https://simple.m.wikipedia.org/wiki/Algorithm [Ray - example Python]
Supervised and Unsupervised Learning - In a supervised learning model, the algorithm learns on a labeled dataset, providing an answer key that the algorithm can use to evaluate its accuracy on training data. An unsupervised model, in contrast, provides unlabeled data that the algorithm tries to make sense of by extracting features and patterns on its own. https://blogs.nvidia.com/blog/2018/08/02/supervised-unsupervised-learning/
Reinforcement Learning - In this kind of machine learning, AI agents are attempting to find the optimal way to accomplish a particular goal, or improve performance on a specific task. As the agent takes action that goes toward the goal, it receives a reward. The overall aim: predict the best next step to take to earn the biggest final reward. https://blogs.nvidia.com/blog/2018/08/02/supervised-unsupervised-learning/
Imitation Learning - Generally, imitation learning is useful when it is easier for an expert to demonstrate the desired behaviour rather than to specify a reward function which would generate the same behaviour or to directly learn the policy. The main component of IL is the environment, which is essentially a Markov Decision Process (MDP). https://smartlabai.medium.com/a-brief-overview-of-imitation-learning-8a8a75c44a9c
Neural Network in AI - A neural network is a method in artificial intelligence that teaches computers to process data in a way that is inspired by the human brain. It is a type of machine learning process, called deep learning, that uses interconnected nodes or neurons in a layered structure that resembles the human brain. It creates an adaptive system that computers use to learn from their mistakes and improve continuously. Thus, artificial neural networks attempt to solve complicated problems, like summarizing documents or recognizing faces, with greater accuracy. https://aws.amazon.com/what-is/neural-network/
DALL·E 2 is a new AI system that can create realistic images and art from a description in natural language. https://openai.com/dall-e-2/
GPT-3 - In May 2020, Open AI published a groundbreaking paper titled Language Models Are Few-Shot Learners. They presented GPT-3, a language model that holds the record for being the largest neural network ever created with 175 billion parameters. It’s an order of magnitude larger than the largest previous language models. GPT-3 was trained with almost all available data from the Internet, and showed amazing performance in various NLP (natural language processing) tasks, including translation, question-answering, and cloze tasks, even surpassing state-of-the-art models. https://towardsdatascience.com/gpt-3-a-complete-overview-190232eb25fd
GPT-4 - In March 2023, Open AI published the advanced fourth version. See https://openai.com/research/gpt-4 It marks a signficant improvement and enhancement over version 3.5. The fourth version is multi-modal - it can recognize and analyze images among the improvements. CNN reports on five key improvements: https://www.cnn.com/2023/03/16/tech/gpt-4-use-cases/index.html#:~:text=GPT%2D4%20is%20an%20updated,AI%2Dpowered%20Bing%20search%20engine.
Generative AI - refers to unsupervised and semi-supervised machine learning algorithms that enable computers to use existing text, audio and video files, images, and even code to create new possible content. Generative AI allows computers to abstract the underlying patterns related to the input data so that the model can generate or output new content. https://indiaai.gov.in/article/here-is-how-generative-ai-is-reinventing-the-creative-space
ChatGPT - is a large language model chatbot developed by OpenAI based on GPT-3.5. It has a remarkable ability to interact in conversational dialogue form and provide responses that can appear surprisingly human. Large language models perform the task of predicting the next word in a series of words. Reinforcement Learning with Human Feedback (RLHF) is an additional layer of training that uses human feedback to help ChatGPT learn the ability to follow directions and generate responses that are satisfactory to humans. https://www.searchenginejournal.com/what-is-chatgpt/473664/
Model: GPT-4
rayschroeder:
I am preparing to have a conversation with a group of higher education leaders on the topic of the impact of generative AI on the future of higher education. Please write a brief informal introduction for me.
=
Hello everyone,
I am delighted to be here today among esteemed higher education leaders to discuss a topic that is both fascinating and critically important to the future of our education system. As we gather to explore the impact of generative AI on higher education, I look forward to engaging in thoughtful and open-minded conversations that will shape the way we approach and embrace this technology in academia.
The rapid advancements in generative AI have undoubtedly brought new opportunities and challenges to education, and together, we have the responsibility to navigate this new terrain. From revolutionizing teaching methods to influencing curriculum development and research, the possibilities are immense. However, we must also be mindful of the ethical considerations, as well as the potential consequences of AI on the job market and society at large.
I'm eager to learn from your experiences, insights, and expertise, as we collectively envision the future of higher education in the age of generative AI. Let's embark on this exciting journey together and shape a future that empowers both educators and students to thrive in an AI-driven world.
Thank you.
(Again, I decided to go right to the source for this answer...)
ChatGPT
Chatbot GPT (ChatGPT) is a language model based on the GPT-4 architecture developed by OpenAI. It functions by utilizing deep learning techniques, specifically, a type of artificial neural network called a Transformer. The GPT in ChatGPT stands for "Generative Pre-trained Transformer."
Here's a high-level overview of how ChatGPT works:
Pre-training: The model is pre-trained on a large corpus of text gathered from the internet, which helps it learn the structure of human language, grammar, facts, and some reasoning abilities. During pre-training, the model learns to predict the next word in a sentence, given the context of the previous words. However, it is important to note that the knowledge of the model is limited to the data it was trained on, and its knowledge is frozen at the cutoff date.
Fine-tuning: After pre-training, the model is fine-tuned on a narrower and more specific dataset, usually generated with the help of human reviewers. This process helps to refine the model's behavior to respond more accurately and safely to users' inputs. Fine-tuning also incorporates an ongoing feedback loop with reviewers, which helps to improve the model's performance over time.
Tokenization: When you provide input text to ChatGPT, it first tokenizes the text into smaller chunks called tokens. Tokens can be as short as one character or as long as one word, and the model processes these tokens to understand the input.
Context vector: The input tokens are then embedded into a high-dimensional space, creating a context vector that represents the meaning of each token in relation to the others.
Attention mechanism: The Transformer architecture uses a mechanism called self-attention, which allows the model to weigh the importance of each token in the input relative to other tokens. This process helps the model capture the context and dependencies in the text.
Decoding: The model generates a response by predicting one token at a time based on the context it has processed. The output tokens are then converted back into human-readable text and presented as the model's response.
It's essential to understand that ChatGPT is a statistical model that generates responses based on patterns it has learned from the training data. It does not have personal experiences or emotions and might sometimes generate plausible-sounding but incorrect or nonsensical answers.
OK... Let's ask a human
Google, Wolfram Alpha, and ChatGPT all interact with users via a single line text entry field and provide text results. Google returns search results, a list of web pages and articles that will (hopefully) provide information related to the search queries. Wolfram Alpha generally provides mathematically and data analysis-related answers. ChatGPT, by contrast, provides a response based on the context and intent behind a user's question. You can't, for example, ask Google to write a story or ask Wolfram Alpha to write a code module, but ChatGPT can do these sorts of things. Fundamentally, Google's power is the ability to do enormous database lookups and provide a series of matches. Wolfram Alpha's power is the ability to parse data-related questions and perform calculations based on those questions. ChatGPT's power is the ability to parse queries and produce fully-fleshed out answers and results based on most of the world's digitally-accessible text-based information -- at least information that existed as of its time of training prior to 2021.
March 15, 2023
Much of the interesting work in the coming months will be to design interfaces to adapt the technology to actual work roles in both supportive and possibly in replacement modes. Generative AI already creates lesson plans, grades assignments, advises students, and answers learner questions. Can it competently take on class management and the other associated administrative tasks? And, certainly, generative AI has an important role to play in research. Will it become a formal or informal co-investigator and co-author of research? What status will we give generative AI in higher ed? And, what will happen to the human faculty aspirants who fail to measure up?
Some ExcellentTools You Can Access Free Right Now:
ChatGPT https://chat.openai.com/chat
Perplexity https://www.perplexity.ai/
You.com https://you.com/
Chat with Any PDF https://www.chatpdf.com/
(Here the two flavors of GPT from OpenAI are described - they have different features for different uses!)
In February, Microsoft announced its integration of ChatGPT to its Bing search engine called Bing Chat. This chatbot has some major differences from ChatGPT, with the biggest difference being access to the entirety of the internet. So which chatbot should you reach for your AI needs? We compared ChatGPT (the free version) and Bing Chat to help you make your decision easier.
https://www.zdnet.com/article/chatgpt-vs-bing-chat/
To get the result you want, you need text prompts that are as concrete as possible. Besides simple prompts for blog or social media posts, ChatGPT can cover many more, sometimes curious, use cases where you need to come up with more specialized prompts. ChatGPT processes up to 4,096 tokens in an input, any character beyond that is ignored without a message. A token is roughly equivalent to a word, depending on the use case. If too much knowledge has been exchanged during a chat, it may be helpful to start a new chat so that subsequent replies are not corrupted.
https://the-decoder.com/chatgpt-guide-five-basic-prompt-strategies-for-better-results/
ChatGPT and tools like it have made AI available to the masses. We can now get all sorts of responses back on almost any topic imaginable. These bots can come up with sonnets, code, philosophy, and more. However, while you can just type anything you like into ChatGPT and get it to understand you, there are ways of getting more interesting and useful results out of the bot. This “prompt engineering” is becoming a specialized skill of its own.
(How can you get ChatGPT to provide sources and citations? - It's simple)
One of the biggest complaints about ChatGPT is that it provides information, but the veracity and accuracy of that information is uncertain. That's because ChatGPT doesn't provide sources, footnotes, or links to where it derived information used in its answers. But that's not fully true. If you know how to properly prompt ChatGPT, it will give you sources. Here's how.
https://www.zdnet.com/article/how-to-make-chatgpt-provide-sources-and-citations/
My friends, I have never before reached out to you personally to read a report - but read this. [message I sent out on 3/16/23 - in decades of sharing such reports - I have not read anything that matches the impact and potential in this report and accompanying materials] Also linked is a recording of the live demo.
Ignore AI at your risk. In many ways, my quick acceptance and learning of pc-based spreadsheets accelerated my career. The tools allowed me to turnaround work faster and be assigned to engagements that allowed me to showcase my financial acumen. I also learned about computers; after all, the first Apples and PCs required you to take them apart and swap out items for increased memory, new processors, new hard drives, etc. I think we are poised for a similar awakening with large language model AI systems. I read three articles related to AI use in education this week. [by Paul LaBlanc, Ray Schroeder, Sal Khan] I’ll provide a few summaries of each and a conclusion at the end.
https://wallyboston.com/you-dont-have-to-know-how-to-code-ai-you-have-to-learn-how-to-use-it/
(This from the co-founder of Lumen Learning - what does this mean to the future of OER and copyrights?)
Excerpted from AI, Instructional Design, and OER - David Wiley, Open Content
In his application to register a work he created using AI software like Stable Diffusion, Steven Thaler wrote that the work “was autonomously created by a computer algorithm running on a machine” and that he was “seeking to register this computer-generated work as a work-for-hire to the owner.” In other words, he applied for copyright protection for a work he created by providing a prompt to a generative AI tool. The US Copyright Office rejected his attempt to register copyright in the work – twice. In their final response they wrote:
Copyright law only protects “the fruits of intellectual labor” that “are founded in the creative powers of the [human] mind.” COMPENDIUM (THIRD) § 306 (quoting Trade-Mark Cases, 100 U.S. 82, 94 (1879)); see also COMPENDIUM (THIRD) § 313.2 (the Office will not register works “produced by a machine or mere mechanical process” that operates “without any creative input or intervention from a human author” because, under the statute, “a work must be created by a human being”)….
While the [review] Board is not aware of a United States court that has considered whether artificial intelligence can be the author for copyright purposes, the courts have been consistent in finding that non-human expression is ineligible for copyright protection….
Courts interpreting the Copyright Act, including the Supreme Court, have uniformly limited copyright protection to creations of human authors…. For this reason, the Compendium of U.S. Copyright Office Practices — the practice manual for the Office — has long mandated human authorship for registration…. Because copyright law as codified in the 1976 Act requires human authorship, the Work cannot be registered.
In other words, as far as the US Copyright Office is concerned, output from programs like ChatGPT or Stable Diffusion are not eligible for copyright protection.
https://opencontent.org/blog/archives/7129
In mid-February, the U.S. Patent and Trademark Office (PTO) released a “Request for Comments Regarding Artificial Intelligence and Inventorship.” The request, which is part of the PTO’s broader effort to engage in issues at the nexus of AI and innovation, invites responses on questions including “How is AI, including machine learning, currently being used in the invention creation process?” and “If an AI system contributes to an invention at the same level as a human who would be considered a joint inventor, is the invention patentable under current patent laws?” Responses are due by May 15, 2023.
https://www.brookings.edu/blog/techtank/2023/03/06/ai-inventions-policy-options-and-a-path-forward/
AI regulation in the United States is still quite nascent. Congress has passed important bills focused on government AI systems. While the Trump administration issued two relevant executive orders, these oversight efforts have so far been largely ineffectual. In 2022, the Biden administration issued voluntary guidance through its Blueprint for an AI Bill of Rights, which encourages agencies to move AI principles into practice. This White House has also issued two executive orders asking agencies to focus on equity in their work, including by taking action against algorithmic discrimination. Many individual agencies have taken heed and are making progress within their respective jurisdictions. Still, no federal legislation focusing on protecting people from the potential harms of AI and other automated systems appears imminent.
(Looking at impact in one professional field - 3/21/2023 -we can extrapolate to others of similar contexts)
Law firms that effectively leverage emerging AI technologies will be able to offer services at lower cost, higher efficiency, and with higher odds of favorable outcomes in litigation. Law firms that fail to capitalize on the power of AI will be unable to remain cost-competitive, losing clients and undermining their ability to attract and retain talent. Consider one of the most time-consuming tasks in litigation: extracting structure, meaning, and salient information from an enormous set of documents produced during discovery. AI will vastly accelerate this process, doing work in seconds that without AI might take weeks.
https://www.brookings.edu/blog/techtank/2023/03/20/how-ai-will-revolutionize-the-practice-of-law/
After years of being relegated to nerdy, wallflower AI status while self-driving cars, robot dogs and the future of the AI-powered metaverse got the spotlight, generative AI’s email-writing, blog-producing, copy-powering abilities are suddenly popular. And top companies from startups to Big Tech are developing tools to gain admittance to the generative AI bash.
https://venturebeat.com/ai/the-hottest-party-in-generative-ai-is-productivity-apps/
The case is made that generative AI has the potential to impact teaching in many fields. The researchers applied an AI Occupational Exposure measure, developed in 2021, to determine which human requirements for a position most overlapped with generative AI capabilities. In one of the listings reviewed by the researchers of most likely to be affected career fields, 14 of the top 20 occupations were higher ed faculty positions in a wide range of fields.
A CBS News report on the paper discusses the relative extent to which this predicts job augmentation or rather job substitution.
https://www.cbsnews.com/news/jobs-chatgpt-occupations-industries-exposed-teacher-telemarketer/
(Follow the development of a generative AI-integrated course designed for Fall 2023)
This letter embarks on a first step along the path to rethink the academy from the ground up: re-building a single course.1 Courses are the basic unit from which the university is constructed.2 With the bigger picture in mind, I sketch out a roadmap to prepare teaching in the fall term – and we will implement this over the coming months. I will re-imagine a course that I plan to teach in the fall term, from the ground up, based on everything we know so far about generative AI, while using our best estimates for its developing impacts, and our best predictions for future changes. The course that we craft will serve as a model for what can be done, a model that can help others in their own work, either by adopting some of the patterns, or by figuring out what they don’t like and how to go their own way.
We have a new partner in higher education to assist with fulfilling our primary mission. What is our primary mission? I turned to our new partner to briefly articulate that “the primary mission of higher education is to provide advanced learning opportunities beyond the high school level, with the aim of developing a highly skilled and educated workforce, advancing research and knowledge, and promoting critical thinking and personal development.” That was the leading sentence of ChatGPT Plus on Feb. 18, 2023, using default parameters responding to the prompt “What is the primary mission of higher education?” The generative artificial intelligence app went on to note that “Higher education institutions, such as universities, colleges, and vocational schools, offer various degree programs, certifications, and training opportunities that provide students with the knowledge, skills, and experiences needed to succeed in their chosen fields.” In order to accomplish that lofty mission, we face the intimidating new challenge of preparing our learners to integrate and optimize AI into their careers and vision of the future. We would be negligent if we were to fail to prepare the learners for what Stanford professor, innovator and CEO of Udacity Sebastian Thrun says is “the single biggest thing we are ever going to see in our lifetime.”
(Some interesting survey questions/answers that go beyond education to society-wide perspectives)
Chances are, you’ve heard the name ChatGPT a million times in recent weeks. With so much information scattered around the web, bold claims, and endless social media posts, it might be hard to grasp the full essence of what ChatGPT can do for us. To some it seems like another AI tool (just a bit more hyped up), to others it signifies the end of the world as we know it. We decided to dig deeper into what society thinks about this whole big thing that generative AI is. After sending out a survey to internet users, collecting 945 responses, exploring notable examples of ChatGPT’s capacity, and experimenting with ChatGPT—it’s safe to say that some things managed to surprise us.
An open letter calls for a six-month break on powerful AI training efforts. The idea is to develop safety and oversight systems and otherwise allow time for consideration of the tech’s rapid development. But the letter also urges more coordination between AI developers and policymakers. It describes a “robust AI governance” regime that would include AI-dedicated regulators; oversight and tracking of AI systems; watermarking to help people discern real and fake images; liability for any harm caused by AI; and public funding of AI safety research. An open letter calls for a six-month break on powerful AI training efforts. The idea is to develop safety and oversight systems and otherwise allow time for consideration of the tech’s rapid development.
https://www.govtech.com/biz/ai-pause-urged-by-musk-wozniak-and-other-tech-leaders
(What can possibly be better than GPT-4? Johns Hopkins has a bionic answer!)
Right now, it feels as though AI is an unstoppable force—but scientists at Johns Hopkins University believe one thing might overpower it. It’s called organoid intelligence, or OI, and it uses actual human brain cells to make computing “more brain-like.” OI revolves around using organoids, or clusters of living tissue grown from stem cells that behave similarly to organs, as biological hardware that powers algorithmic systems. The hope—over at Johns Hopkins, at least—is that it’ll facilitate more advanced learning than a conventional computer can, resulting in richer feedback and better decision-making than AI can provide. Johns Hopkins’ OI “brains” are too small to produce researchers’ desired outcomes. The current organoids hold about 50,000 cells, and they’d need to hold closer to 10 million to facilitate OI. Once they’ve reached that point, they’ll contain approximately 125 trillion synapses, each of which can store 4.7 bits of information. The world’s fastest supercomputer contains 58 billion transistors, each storing 4 bits. Using human brain cells to power computers has obvious ethical implications, which the researchers openly acknowledge.
https://www.extremetech.com/extreme/343498-biocomputers-using-human-brain-cells-could-overpower-ai
https://sites.google.com/view/upcea-gpt/home
~~~~~~~
Ray's Daily Curated Reading Lists and Social Media. Blogs with daily updates on the field of online / continuing learning in higher education
Inside Higher Education - "Online: Trending Now" https://www.insidehighered.com/blogs/online-trending-now
UPCEA Professional, Continuing and Online Education Update http://continuingedupdate.blogspot.com/
Economics and Change in Higher Education by UPCEA http://recessionreality.blogspot.com/
OER Update by UPCEA https://oerupdate.blogspot.com/
Ray's Curation by UPCEA https://rayscuration.blogspot.com/
Alternative Credentials in Higher Ed by UPCEA https://altcred.blogspot.com
Twitter https://twitter.com/rayschroeder
rschr1@uis.edu ~ rayschroeder@gmail.com - ray@upcea.edu
Senior Fellow, University Professional and Continuing Education Assn.
Professor Emeritus, University of Illinois Springfield