AI in the Workplace

Generative AI and Assistive AI Concepts

Generative AI and Assistive AI represent two distinct concepts within the field of artificial intelligence, each with its own focus and applications. The following descriptions were created with the help of Generative AI tools and then edited by a human for improved clarity and tone.

Generative AI

Purpose: Generative AI is designed to create new content, such as text, images, videos, music, or even code. It generates outputs that did not exist before, based on the patterns and data it has been trained on.

How It Works: Generative AI utilizes models that have been trained on vast amounts of data to understand and mimic the structure, style, or logic of that data. For instance, it can generate realistic images from textual descriptions or write essays on given topics.

Applications: Generative AI tools can be used to create art, generate realistic computer graphics, compose music, write stories or articles, develop new designs, and more. Tools like DALL-E for image creation or GPT (Generative Pre-trained Transformer) for text are examples of Generative AI.

Assistive AI

Purpose: Assistive AI is intended to aid humans with their tasks, improving efficiency, accuracy, or accessibility. It is not necessarily about creating new content but about enhancing human capabilities or making tasks easier.

How It Works: Assistive AI involves AI systems that can understand human requests, provide recommendations, automate repetitive tasks, offer intelligent insights, or support decision-making. Assistive AI is often interactive and responsive to human inputs.

Applications: Assistive AI includes virtual assistants (like Siri or Alexa), predictive typing and text completion, AI in healthcare for diagnosis support, tools for accessibility (such as voice-to-text for those unable to type), and AI in customer service (like chatbots).

Key Differences

Creation vs. Support: Generative AI focuses on creating new outputs based on learned data, while Assistive AI focuses on supporting humans in their tasks, enhancing productivity, or improving accessibility.

Application Areas: Generative AI is prominent in creative fields and content generation, whereas Assistive AI is widespread in service-oriented sectors, including healthcare, customer service, and personal assistance.

Interaction: While Generative AI’s primary interaction is through the generation of new content based on input prompts, Assistive AI interacts by understanding and responding to human needs, often in a collaborative manner.

Thus, Generative AI is about creating new, original content, whereas Assistive AI aims to aid, enhance, or streamline human tasks. Both are powerful, but they serve different purposes and are used in different contexts.

Is AI “New Technology”?

Not exactly! AI tools, including Generative and Assistive AI, have been around for decades. What’s new, and why AI is getting so much attention now, is that companies have figured out how to add natural human-like language (called natural language processing) to the data, which allows us to interact with large complex data sets as if we were speaking to a friend. This means that you no longer need to be a data scientist or have specialized knowledge to access this data. Anyone with a computer can now participate.

Is AI “Disruptive Technology”?

AI—like the Internet, mobile phones, and SaaS (software as a solution) before it—will impact the way we work for generations to come. It’s called “disruptive technology,” but that doesn’t mean it’s a bad thing. There are, though, a lot of unanswered questions as we navigate this new space, and we need to be thoughtful and intentional in our approach. Here in ITS, we’ve dealt with many disruptive technologies over the years, and we have well established policies and practices for analyzing, evaluating, and operationalizing these new offerings.

How do I use AI safely?

Community members that use AI technologies are expected to adhere to Middlebury’s existing Responsible Use and Information Security policies to engage in safe, ethical, and law-abiding behavior, conserving common resources and with respect for others. ITS will continue to support the Middlebury community in its effort to adhere to these policies by communicating specifically what this means for AI technologies.

Some basic guidelines to follow are:

1. Always consider the data security, privacy, and the ethics of the work. If you wouldn’t want that information published on a public website, then do not use that data within an AI tool.
2. Never enter your account credentials within an AI chatbot. This includes usernames, passwords, and special access tokens or keys.
3. Never use AI on “auto-pilot”. AI features are not a “set it and forget it” toolset. Human judgement and guidance is need to properly instruct an AI agent and to review the output of any AI agent.

How does ITS evaluate AI features?

Large companies like Microsoft, Google, Zoom, Adobe, and others, are already introducing AI functionality into their software. In some cases, these features are implemented without an option to shut them off. However, where we have more control over that functionality, ITS uses a formal evaluation process to determine if that software is safe, secure, accessible, and sustainable for long-term use.

A high-level overview of our process can be summarized as:

1. An initial business architecture review is conducted and our vendor questionnaire (along with any supplemental document requests) are sent to the vendor. Any requirements for an NDA are established. 

2. Once the questionnaire is returned back to use from the vendor, ITS takes particular care to evaluate the information from the lens of security, data privacy, and accessibility. 

3. Follow-up information is often required, and after a cycle of discovery with the vendor, summary reports with scoring are created for security, data privacy, and accessibility. Summary reports are shared and reviewed with the Information Security Officer

4. Recommendations are communicated to the requesting Middlebury partner and a decision is made to either continue moving forward with the vendor, or to disengage. 

5. If required, an entry into the risk registry is made. 

Which services already have AI features available?

The following AI tools are available to our community.

What opportunities exist for the workplace?

Search the internet for AI and you will find thousands of articles, videos, and examples of amazing work being done with these new tools. Some are inspiring, and some raise questions around the ethics of certain activities. 

For our administrative offices, we’re focused on three main areas where we see opportunities for leveraging AI tools to improve organizational nimbleness and productivity.

I. Making existing systems easier to use.

Reviewing ways to Introduce AI tools into our existing platforms to make using that platform, or using the data within that platform, easier.

II. Using AI as a Creative Partner

Collaborating with an AI tool as a thought partner to brainstorm, review, or create first draft material where appropriate.

As a rule, AI generated content is should never be used “as-is” but rather within the H-A-H model—Human-AI-Human. 

• A human guides and instructs the AI tool in a thoughtful, ethical, and safe manner.
• The AI agent generates its output.
• A human reviews, corrects, and approves the output.

III. Realizing Efficiency Gains

With an eye towards efficiency, examining existing processes that are ripe for automation or where we engage in repeated tasks with minor differences (we call these serial processes).

This is Just the Beginning

We will be updating and expanding this resource with new information as we all explore and grow in our understanding of what AI means for our administrative offices here at Middlebury. ITS is committed to supporting you. If you have any questions, please contact us to get a conversation going.