Computers in the corporate environment are nothing new. Most people are familiar with using a computer for a variety of activities: email, spreadsheets, web browsing, and programming, to name a few.
Traditionally, computers served as an instrument of our minds to express thought into a digital world. Computers assisted in handling data by serving as recorders, displayers, and transmitters in a useful format and manner. As computing power advanced, new capabilities came along that still required human thought and action to enact. Such examples include highlighting text in a document, making graphs, or emailing department-wide announcements.
No longer is computing driven exclusively by humans. With artificial intelligence (AI), we are seeing computers start to act more like their human operators. To summarize, “for decades, machines had to be taught everything. With artificial intelligence, they learn.” Change is no longer coming; it is here.
Large language models (LLMs), such as ChatGPT made by OpenAI and Microsoft, have captured the imagination of hundreds of millions around the world. In fact, ChatGPT broke the record for the “fastest growing user consumer application in history” with “100 million monthly active users in January [2023].” Trained on billions of pieces of text generated by humans and spread throughout the internet and literature, ChatGPT version 4 (GPT-4 henceforth) can output text that is useful and convincing to many human readers, at least at first glance.
While still in its early stages, GPT-4 and subsequent versions of the system continue to improve based on analysis of user input. Just like other technology, it is always at its weakest in the present timeframe we are using it in.
Thanks to the powerful sentiment analysis that GPT-4 has enabled, new systems such as AgentGPT, Auto-GPT, and others have been built to chain together GPT-4 and other task-specific AI platforms to achieve goals that any of them separately wouldn’t be able to fulfill.
GPT-4 may be used to decipher the meaning of a user’s request and generate a text prompt for Midjourney (an AI image generation tool).
Midjourney then creates a picture.
A third system reviews the Midjourney output for accuracy and, upon approval, sends the picture on to the next step.
Finally, another AI system handles posting the photo to a merchandise sales account such as Redbubble. It then coordinates a social media campaign promoting mugs and t-shirts with the image on them.
As one AI consultant and tool builder stated, “we now have the power and responsibility of managing a coordinated team of AI agents at our fingertips without much effort.” This leads well into our next consideration: ethics.
When bringing AI into the workplace, careful considerations must be made regarding ethics. The corporate world is entering uncharted territory with AI, with a workforce that has heterogeneity in experience with computers and individual moral compasses.
Global AI Adoption Rates [IBM, 2022]
According to research by IBM, “two in three companies say they lack the skills and training to develop and manage trustworthy AI.” When computers first became mainstream in the office, they were relegated to simple tasks such as note-keeping, accounting, and inventory tracking. Now, with the entrance of advanced AI, even an inexperienced operator can wield force beyond human imagination. If utilized thoughtlessly, these tools can, in a worst-case scenario, be recklessly destructive or, more mildly, reputation-damaging.
Every person employed within a company needs to have training specific to their role and to each AI tool they are going to be using under the company’s banner. If your human resources department has been using an AI to sift through resumes and that AI was only trained on “traditionally” male-sounding names, are you certain that the tool won’t miss a potential excellent hire, nay, leave your company open for potential legal action? The maker of ChatGPT, OpenAI, has published warnings about AI-generated output containing “harmful stereotypical and demeaning associations for certain marginalized groups.”
Every company must review how they will utilize these new tools with an understanding of where the training data is sourced, the quality of the data, the limitations of the tool, and where legal or revenue exposure may occur.
The next few years are going to see rapid shifts in how we work with AI competencies and capabilities continuing to grow at breakneck speeds. As companies grapple with using the new technologies, we have responsibilities to not only use the systems well but also responsibly. As you and your team implement AI in your workflows, work together to remember the human at every step. “Compute unto others as you would want them to compute unto you.”
IBM – “IBM Global AI Adoption Index 2022” by IBM and Morning Consult. Published May 2022. https://www.ibm.com/downloads/cas/GVAGA3JP
OpenAI – “GPT-4 System Card”, no author credited. Published March 23rd, 2023. https://cdn.openai.com/papers/gpt-4-system-card.pdf
Reuters – “ChatGPT Sets Record for Fastest-Growing User Base - Analyst Note” by Krystal Hu. Published February 2nd, 2023. https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01/
Shaurette - “Information Security Management Handbook, Fourth Edition, Volume II” ISBN-10 0849308003. Book edited by Harold Tipton and Micki Krause, quote by Ken Shaurette. Published October 20th, 2000.
Thomson Reuters – “Understanding the Key Machine Learning Terms for AI”, no author credited. Published May 23rd, 2023. https://legal.thomsonreuters.com/blog/understanding-the-key-machine-learning-terms-for-ai/
Venture Beat – “As AI Agents Like Auto-GPT Speed Up Generative AI Race, We All Need to Buckle Up” by Sharon Goldman. Published April 17th, 2023. https://venturebeat.com/ai/as-ai-agents-like-auto-gpt-speed-up-generative-ai-race-we-all-need-to-buckle-up-the-ai-beat/