By Emily Durning
As a senior undergraduate student nearing graduation, it can be jarring to think back to the state of AI, and more specifically Generative AI, my first semester of college. In late 2022, the first version of ChatGPT was gearing up to be released, meaning that while AI was still seeping its way into all walks of life, it had not yet hit the exponential increase of usage and integration that has been seen the past few years.
Taxonomy of Artificial Intelligence
Artificial intelligence is defined by Oxford languages as “the application of computer systems able to perform tasks or produce output normally requiring human intelligence.” Microsoft then classifies AI into traditional, predictive, conversational, and generative subcategories. There is overlap between these subcategories, and conflicting categorizations depending on what sources you’re looking at. For example, ChatGPT is a tool that overlaps the conversational and generative AI subcategories, as the user is able to converse with it as a chatbot, but its ability to generate new content makes it fall under the category of generative AI as well.
Confusingly, if instead of consulting Microsoft's website you look at IBM’s subcategory definitions, the terminology used is completely different. Generative AI is listed as an example within the umbrella of “Limited memory AI” which is one of their functionality-based AI categories. This overwhelming difference in language around these tools can be seen almost anywhere when reading about AI. There is a surplus of information on AI that is all slightly discordant with one another, producing the illusion that AI is too complicated to understand.
Why the introduction to AI terminology and its discrepancies? With the pace that AI has been moving the past few years, feelings of apprehension as well as strong motivations to adapt to everything AI are both understandable reactions. There is a lack of a strong foundation in understanding of the functionality of AI and its optimal uses. Additionally, this AI surge has caused many to feel very behind, very quickly. In an effort to “catch up”, AI has disjointedly been added throughout companies and institutions without a comprehensive plan of action to be understood by the organizations as a whole.
Pitt’s AI Ecosystem
Within the University, we at RDS@Pitt are in the process of documenting each of these individual instances of AI within courses, curriculums, research, workshops, and more. The goal behind this documentation is to be able to clearly understand what schools or offices within the university are already making steps towards integrating AI into their research and courses or are opening up new opportunities specifically to learn about AI. This understanding will increase the opportunities for collaboration across disciplines.
The figure below reflects the current state of this mapping. It is organized by the college within the university or other main organization that each record belongs to. Further, the points are colored based on the Type 1 categorization of each entity. These “Type 1” categories are Product, Research, Courses, Programs, and Degrees. Hovering over each point will give you the name of the resource and the main organization it is tied to.
For a more fine-grained look at each Type 1 category, there are additional figures (fig. 2-6) at the end of this blog, one for each of the 5 Type 1 categories. Within these smaller visualizations, each point is still linked to the main organization it can be found in within the university but they are now colored by a Type 2 categorization. The Type 2 categorizations are contingent on the Type 1 mapping.
From the information we have today, the main takeaway is that there are a few main silos of AI initiatives, with little overlap between parent organizations.
What’s Next
This is a part of an ongoing effort to create a resource that is a true representation of all AI projects, courses, research efforts, and institutional initiatives within the university. In an effort to keep our database as accurate as possible, we ask you to fill out our form if you are aware of anything that should be added. We appreciate your contribution to the documentation of Pitt’s AI Ecosystem!
References