ChatGPT and the Evolving Field of Artificial Intelligence (AI)

Chatbot – an app or interface designed to respond to text-based questions or prompts automatically and conversationally. These are currently often used in customer support.

Natural language processing (NLP) – The branch of AI researching and testing solutions around tools that learn from and replicate human-sounding language patterns.

Predictive analytics – using data sets of past patterns to guess what is likely to happen next.

Artificial intelligence (AI) – A computer database, algorithm or program capable of performing activities that are normally thought to require intelligence, such as learning and reasoning.

Machine learning – a subset of AI – an algorithm designed to find patterns in sets of data and use these to make decisions. The dataset is typically already organized to support the effectiveness of the algorithm.

Deep learning – a subset of machine learning, an algorithm that can work better with unstructured data. The algorithm can determine from context the important information on which to focus, alleviating the requirement of human intervention to organize the dataset.

Large language models (LLMs) – an AI model that processes immense amounts of text (usually collected from publicly available online sources) and uses deep learning algorithms to detect and understand language patterns, common usages, and contexts. LLMs are uniquely capable of recognizing patterns of human communication and are typically enhanced through training by humans.

GPT-3 – generative pretrained transformer 3, an LLM created by OpenAI that uses deep learning to produce (generate) human-like text based on user prompts. GPT-3 can perform a range of related language tasks and is the third iteration of the service. It is uniquely able to do things it has not been explicitly trained to do.

Generative AI – AI specifically developed to produce an output. These typically aim to produce text, audio, or video, but are expanding into other fields such as music as well. ChatGPT is an example of text-to-text generative AI, while tools like Stable Diffusion and DALL-E 2 are examples of text-to-image generative AI.

Artificial General Intelligence (AGI) – AI fully capable of human-like thinking and communication, including creative and original ideas and self-awareness (sentience). There are no current forms of AGI yet developed.

ChatGPT is a chatbot, much like you might use for questions or support on a website, but more capable and powerful. You can ask ChatGPT any question, and it will reply using everyday language and an immense database of knowledge (Figure 1). It can write essays on a particular topic, poetry in a particular style, or process guides for accomplishing a task. ChatGPT can generate working code, chemistry compositions, and a variety of text formats from a few simple user inputs. Give it a few words or a sample lyric, and it can write you a song. Give it a few words of copy, and it can write a half-decent ad.

You can use ChatGPT at no cost – at least for now.  However, please know you will need to create an account, and it will require two-factor authentication using a phone number. It will also store and claim ownership over any inputs or outputs given to the system.

Figure 1: A prompt invites ChatGPT to explain its own capabilities.

In more technical terms, ChatGPT is a specialized usage of GPT-3 (Generative Pre-trained Transformer 3), a large language model (LLM) released in 2020 by OpenAI.  An LLM uses billions of examples of text communications to detect patterns that help it understand word relationships and contexts. (GPT-3 refers to over 175 billion parameters as its database – not the largest LLM, but certainly a big one!) Conversational AI products, like ChatGPT and other chatbots, are the main applications of LLMs like GPT-3.

ChatGPT has been trained specifically on how it should interact with users and what constitutes an appropriate output. It uses these rules alongside predictive analytics to create replies to questions that a user asks, in everyday language (Devaney, 2022). The chatbot also remembers questions and directives within a conversation and uses these to continue to adapt and refine its responses to you. The below video offers some additional details about how natural language processing works.

These conversations around what ChatGPT can do, what AI looks like in the world and other considerations around the technology behind this will be covered in the workshop on The Tech Behind ChatGPT and the Generative AI Landscape – (EDV0044).

OpenAI, the organization behind ChatGPT, was originally founded in 2015 as a not-for-profit research and development venture for ethical AI (OpenAI, 2018). Initial investors and their original board included notable tech leaders such as Reid Hoffman (founder of LinkedIn), Elon Musk (founder of Tesla and SpaceX), and Sam Altman (now current CEO of OpenAI). OpenAI’s Chief Science Officer, Ilya Sutskever, is a renowned researcher and innovator in the field of artificial intelligence and neural networks, working formerly on Google’s DeepMind project.

OpenAI has a charter that commits to creating AI which will benefit all of humanity through collaboration with other organizations. Their charter explicitly states that they believe AI should benefit all of humanity, and identifies the goals of cooperation with research and policy institutions. CEO Sam Altman suggests that OpenAI’s approach of early release of AI innovations to the public will make AI more transparent by preparing society for the introduction of AI into daily life, and allow legislation and regulation to keep up with the fast pace of AI developments (Loizos, 2023). However, others have questioned whether their actions are truly in alignment with their charter (Hao, 2020).

Because LLMs are so large, they require massive and rapidly expanding computing capabilities. Estimates suggest that AI computing has a doubling time of every 3.5 months, requiring larger storage capacities and more powerful supercomputers (Hao, 2020). Considering these high costs, in 2019 OpenAI established a for-profit branch which allowed them to seek business and funding partnerships.

Shortly thereafter, OpenAI began a partnership with Microsoft. In 2019, Microsoft invested one billion dollars in OpenAI (OpenAI, 2019). The deal saw Microsoft provide an Azure supercomputer and seed funding, and in return Microsoft benefitted from a multi-year non-exclusive partnership (Nellis, 2019). After the release of ChatGPT in November 2022, reports indicate that Microsoft and OpenAI began conversations around a new proposed $10 billion investment deal (Kiran, 2023). This arrangement may help Microsoft recover from the very negative public launch of their first efforts in AI in 2016 (Kraft, 2016), and would be a step forward in catching up to Google’s AI product LaMDA, which is already so advanced it has its own controversies (Maruf, 2022).

ChatGPT isn’t the only tool of its kind. There are many different types of generative AI tools already available, which produce different forms of media. Many of these are paid services. A Canadian software engineer, Aaron Sim (@aaronsiim), posted a helpful visual (Figure 2) of some of the generative AI tools available (current as of 10/22, so already a bit outdated). This list shows us that generative AI can be used for many forms of media-to-media production, particularly generating text and images.

Figure 2: A non-comprehensive list of generative AI tools already on the market, typically as paid services.

It’s important to keep in mind that the limitations of these tools are consistently evolving. One major limitation is that ChatGPT is not actively connected to the internet. Unlike a search engine, it is not capable of scraping the web for current information, meaning it also can’t analyze webpages in real time. Information and events after 2021 are currently unknown to it. The chatbot will also make up facts, particularly when asked for references. Recent updates have sought to improve these limitations, but most LLMs currently lack the capacity to accurately triangulate or synthesize from several sources.

Additionally, ChatGPT is currently unable to:

• Consistently associate correct details or characteristics to different characters or situations.
• Make predictions about the future.
• Summarize, contrast and compare content accurately from multiple sources.
• Draw connections between texts and visuals or videos.
• Output accurate or suitable graphic, process or flow charts.
• Consistently follow prompts that contradict its own rules and restrictions.

These limitations should not be taken as faults or flaws of ChatGPT. They are limitations reasonable to what it is – an automated knowledge base capable of doing a good job of guessing what words belong together in a particular context (Bender et al., 2021). We should keep in mind, however, that as a learning machine, ChatGPT will likely continue to quickly evolve (Vincent, 2023).

Figure 3: A prompt asks ChatGPT to describe how it organizes information.

There are many uncertainties surrounding AI but one of the most significant is copyright ownership. Who owns the input and output from these systems?

ChatGPT collects a record of all inputs and outputs used on the platform. The current terms and conditions of use establish OpenAI as the owner of this data, and they grant the end user (us) the rights to sell or redistribute that work. Similarly, OpenAI reserves the right to grant other people the rights to also sell and redistribute the same or similar works. Some AI products, like Stable Diffusion, establish outputs as belonging to the public domain.

In addition, ChatGPT and other LLMs are trained using publicly available information from individual users, content producers and corporations, including conversations, images, collaboratively co-edited information, websites, blogs and more. Since AI systems are trained on copyright-protected works, it is possible that output from these systems could produce a work that is a copy of an original work. A copy of a protected work could be copyright infringement.

Ownership of AI content will eventually be determined by a legislative amendment. Until ownership is determined by this amendment, it is not advisable to use ChatGPT in activities related to course materials production or assessment generation. Conestoga College’s Intellectual Property Policy clarifies that the college owns all material “developed by employees within the scope of their employment” (Conestoga College, 2018, p. 5). Using a system like ChatGPT could pass ownership of college intellectual property to another entity. In addition, using output created by an AI system could put Conestoga College at risk of litigation if an output is considered a copy of another work.

Further questions can be referred to the Copyright Coordinator – James Yochem.

Students may attempt to use AI-generated content as answers to prompts for assignments. While organizations like Turnitin, GPTZero, and OpenAI have released or proposed AI content detectors, their accuracy has not yet been adequately tested beyond the vendors’ claims, and there is little transparency about how their algorithms “detect” algorithmic writing outputs. On top of this, it is unclear whether AI content detectors will be consistently and reliably available, or how they will retain and use the student-created works when input into their systems.

There is also some evidence that AI content detectors can be easily evaded if suitable prompts are used. Students who create more sophisticated prompts may be more able to circumvent AI content detectors. Similarly, the students who are most likely to be caught using them are likely those who may be least successful in a course.

To consider further the implications of generative AI on academic integrity, consider attending the workshop Generative AI and ChatGPT: Future Considerations and Potential Challenges in Assessing Student Learning (EDV0043) by Cory Scurr (Manager – Academic Integrity) and the Teaching and Learning team.

While it can be exciting and engaging to use an innovative tool like ChatGPT in teaching and learning activities, it should be remembered that LLMs have been criticized over several risks (Bender et al., 2021). Further, it is recommended not to require students’ use of generative AI tools, and support choice and opting out if deciding to allow. Below, you’ll find some of the most notable areas of caution.

Figure 5: One of ChatGPT’s first messages to new users informs them of the intent of the service and the possibility of problematic or misleading outputs.

It may be inaccessible to some students.

It can be confidently inaccurate.

While the writing tone and style of ChatGPT sounds confident, it can present inaccurate information that appears to be factual and sourced, a phenomenon known in AI development as “hallucination”. If prompted to provide references or quotes, it may often make them up. The service also often does a poor job of accurately summarizing the information it does have. This may improve rapidly but illustrates the importance of fact-checking and verifying information when it is presented. Users of the service will need to still evaluate the trustworthiness of any information generated.

It is built on and can produce harmful content.

LLMs like GPT-3 are usually trained on publicly available information, like social media sites and public opinion forums. This means that they are vulnerable to reflecting biases exhibited in these spaces. Despite measures to train ChatGPT on appropriate responses and restrict inappropriate outputs, it is capable of expressing bias, sexism or racism towards cultures or groups. It is also capable of providing outputs with harmful potential, such as the code for potential malware.  While OpenAI has tried to implement barriers to these types of malicious uses, these are not completely failsafe. Users have already been able to circumvent security protocols to produce malicious outputs.

Students may develop cognitive overreliance on the tool.

Learners may turn to generative AI tools as a solution to “looking it up” and begin to over-rely on these tools as sources of information. This may interfere with important mental processes of triangulation (Zhao, 2023), which occurs when we compare this source to others, gather our own inferences, and make conclusions. Enhancing teaching and learning activities that prioritize thinking processes of analysis, comparison, or synthesis may help support this problem.

Over the past 10 years, some organizations have begun integrating chatbots into various aspects of education. Primarily, chatbots are used in teaching and learning, administration, assessment, advisory, and research and development capacities (Okonkwo & Ade-Ibijola, 2021). Zeide (2019) summarized their uses into three main branches of applications – institutional, student support, and instructional.

Figure 6: Three main branches of AI use in education, primarily focused around chatbot use.

Institutional Supports

Institutional applications for learners might look like improved predictions about recruitment and improved processes for application and acceptance. It could also provide supports for resource planning and growth of the organization.

Student Supports

Student supports might look like improving systems that help students access advisors, and which make predictions about at-risk students, offering suggestions for how to support them more effectively.

Instructional Supports

Chatbots are primarily used in education for teaching and learning purposes, supporting the delivery of course content and student engagement and retention (Okonkwo & Ade-Ibijola, 2021). Their use typically aligns best with pedagogical and instructional design principles such as personalized and experiential learning, social dialog, collaborative learning, affective learning, and learning by teaching (Kuhail et al., 2023). In particular, AI tools can support faculty and learners with:

• identifying opportunities for customized instruction and feedback,
• supporting reflective activities,
• providing social dialog to engage users,
• collaborative techniques to improve learners’ critical thinking and argumentation,
• providing empathetic feedback to keep students motivated,
• and allowing students to learn through teaching.

The instructional branch of AI for education also applies learner analytics, visualizations of trends in student performance at course and program levels. Learner analytics inform instruction and provide guidance from which to better adapt and refine courses supportive of learners’ behaviours and needs. Learner analytics should be supported by consideration of the ethical implications of using the data. The DELICATE checklist offers a framework for evaluating ethical use of learner analytics (Draschsler & Greller, 2016). Effective LLMs may continue to broaden the scope of what is possible with chatbots in education over the next few years.

Generative AI is not new and has been gaining momentum in recent years. Below, we have collected a timeline of notable advancements in AI since the 1940’s.

The language-based AI field has several key branches, each with their own trajectories of development. These key branches are primarily handwriting, image and speech recognition, reading comprehension, and language understanding. In Figure 1, the rapid improvements in each of these AI branches are outlined, showing just how quickly they have evolved to surpass human performance.

Figure 7: The rapid improvement of language and image-based AI systems.

These conversations around what ChatGPT can do, what AI looks like in the world and other considerations around the technology behind this will be covered in the workshop on The Tech Behind ChatGPT and the Generative AI Landscape – (EDV0044).

The release of ChatGPT has spurred significant movement from both collaborators and competitors. Microsoft has already incorporated GPT-3 into Azure, and is already incorporating elements of it into Word, Outlook, and Bing to increase their competition with Google. It had already been integrated in 2022 into GitHub Copilot, a coding production service (and another Microsoft product). Soon, we’ll also see the release of Microsoft Designer, an alternative to design tools like Canva and Visme. We can also expect other tools to develop from the GPT-3 (or GPT-4) protocol, as OpenAI offers subscription-based API access, allowing developers to create apps using the LLM as a backbone.

Google, Amazon, and other large tech organizations have all released similar managed-development platforms for their LLMs over the past three years (see the timeline in How did the technology develop? for details).  Google has begun releasing consumer-level AI tools, including a chatbot like ChatGPT named Bard (Milmo, 2023). Facebook released a similar chatbot a few months prior to ChatGPT. AI chatbot growth is also international – for example, many organizations in China are also rapidly moving forward with the release of similar tools (Yang, 2023). Considering the initial steps being taken, these tools will likely have a significant impact, particularly on how we use search engines to find information.

Businesses will begin to look to incorporate these tools into services or processes. Many industries, such as healthcare, agriculture, insurance, food production and customer service already have forms of AI built into systems which could be enhanced or complemented with what LLMs offer (Wallis and Mac, 2018). In all, the rapidly changing field of AI will likely be a prominent element of research and development projects for both businesses and research initiatives over the coming years.

Consider attending some of the workshops available from Teaching and Learning on AI, ChatGPT and a variety of teaching considerations.

If you’re looking for additional context on how AI is developing in Canada, and what it may look like in various industries, The AI Effect is a two-season podcast investigating “themes around the Canadian AI ecosystem”. (Found on Apple, Spotify, or any other major podcast platform.)

Professor Toby Walsh, a leading researcher on AI’s impacts on society, hosted an AMA on Reddit on January 31^st. (Mildly NSFW language).

The DELICATE checklist offers a framework for evaluating ethical use of learner analytics, one form of AI (Draschsler & Greller, 2016).

The Université de Montreal has established the Montreal Artificial Intelligence Ethics Institute (MAIEI), which has established the Montreal Declaration of Responsible AI, which offers 10 guiding principles to lead AI development. Review the signatories to see the impact on Canadian and global AI partners.