While the age of sentient robot assistants isn’t quite here yet, AI is fast making a bid to be your next co-worker.
More than half of U.S. workers are now using some form of AI in their jobs. According to an international survey of 5,000 employees by the Organisation for Economic Co-operation and Development (OECD), around 80 percent of AI users reported that AI had improved their performance at work, largely pointing to increased automation. For some, the ethical integration of AI is the top workplace concern of 2024.
But while proponents note how much potential there is for AI technologies to improve and streamline more equitable workplaces — and there’s probably examples of AI already at play in your job, as well — that doesn’t mean we should all rush to bring AI into our work.
That same OECD survey also documented continued fear of job loss and wage decrease as AI digs its heels deeper into the employment landscape. A different survey of U.S. workers by CNBC and SurveyMonkey reported 42 percent of employees were concerned about AI’s impact on their job, skewing higher for those with lower incomes and for workers of color.
And with the rise of AI-based scams, ongoing debate over government regulation, and worries about online privacy (not to mention the sheer over-saturation of “new” AI releases), there’s still a lot of unknowns when it comes to AI’s future.
It’s best to tread into the world of AI at work with a bit of trepidation — or at least with some questions in your back pocket.
What kind of AI are we talking about, exactly?
First step: Familiarize yourself with artificial intelligence at large. As the term has grown in popular use, “Artificial Intelligence” has evolved into a catchall phrase referring more to a variety of technologies and services than a specific noun.
Mashable’s Cecily Mauran defines artificial intelligence as a “blanket term for technology that can automate or execute certain tasks designed by a human.” She notes that what many are now referring to as AI is actually something more specific, known as generative AI or artificial general intelligence. Generative AI, Mauran explains, is able to “create text, images, video, audio, and code based on prompts from a user.” This use has recently come under fire for producing hallucinations (or made up facts), spreading misinformation, and facilitating scams and deep fakes.
Other forms of AI include simple recommendation algorithms, more complex algorithms known as neural networks, and broader machine learning.
As Saira Meuller reports for Mashable, AI has already integrated itself into the workplace (and your life) in a multitude of ways, including Gmail’s predictive features, LinkedIn’s recommendation system, and Microsoft ‘s range of Office tools.
Things as simple as live transcripts or captions turned on during video meetings rely on AI. You could also encounter it in the form of algorithms that facilitate data gathering, within voice assistants on your personal devices or office software, or even as machine learning that offers spelling suggestions or language translations.
Does your company have an AI policy?
Once you’ve established that the AI tool falls outside of a use case already employed in your day-to-day work, and thus might need some further oversight, it’s time to reach out to management. Better safe than sorry!
Your company will hopefully have guidelines in place for exactly what kind of AI services can be pulled into your work and how they should be used, but there’s a high chance it won’t — a 2023 survey from The Conference Board found that three-quarters of companies still lacked organizational AI policy. If there are no rules, get clarity from your manager, and potentially even legal or human resources teams, depending on what tech you’re using.
Only use generative AI tools pre-approved by your place of work.
In a global survey of workers by business management platform Salesforce, 28 percent of workers said they were incorporating generative AI tools in their work, but only 30 percent had received any training on using the tool appropriately and ethically. A startling 64 percent of the 7,000 workers reported passing off generative AI work as their own.
Based on the response rate of unsupervised use, the survey team recommended that employees only use company-approved generative AI tools and programs, and that they never use confidential company data or personally identifiable customer data in prompts for generative AI.
Even big companies like Apple and Google have banned generative AI use in the past.
Things to consider before using a generative AI tool:
Data privacy. If you are using generative AI, what kind of information are you plugging into the tool, such as a chatbot or other LLM? Is this information sensitive to individuals you work with or proprietary to your work? Is the data encrypted or protected in any way when it is used by the AI?Copyright issues. If you are using a generative AI system to design creative concepts, where is the tech sourcing the artistic data needed to train its model? Do you have a legal right to use the images, video, or audio the AI generates? Accuracy. Have you fact-checked the information provided by the AI tool or spotted any hallucinations? Does the tech have a reputation for inaccuracy?Who would the AI serve?
It’s also important to distinguish where AI fits in your daily work-flow, and who will be interacting with any generative AI outputs. There is a difference between incorporating AI tools like chatbots or assistants within your own daily tasks, and replacing an entire job task with it. Who will be affected by your use of AI, and could it be a risk to you or your clients? The disclosure of AI use is a question even law firms lack clear answers to, but a majority of Americans believe companies should be mandated to do so.
Thing to consider:
Are you using an AI tool to generate ideas solely for your own brainstorming process?Does your use of AI result in any decision-making for you, your coworkers, or your clients? Is it used to track, monitor, or evaluate employees?Will the AI-generated content be seen by clients or anyone outside of the company? Should that be disclosed to them, and how?Who is in charge of the AI?
You’ve gotten the go-ahead from your company and you understand the type of AI you’re using, but now you’ve got some larger ethical matters to consider.
Many AI watchdogs point out that the quick rush to innovate in the field has led to the conglomeration of a few Big Tech players funding and controlling the majority of AI development.
AI policy and research institute AI Now points out that this could be a problem when those companies have their own conflicts and controversies. “Large-scale AI models are still largely controlled by Big Tech firms because of the enormous computing and data resources they require, and also present well-documented concerns around discrimination, privacy and security vulnerabilities, and negative environmental impacts,” the institute wrote in an April 2023 report.
AI Now also notes that a lot of so-called open source generative AI products — a designation that means the source code of a software program is available and free to be used or modified by the public — actually operate more like black boxes, which means that users and third-party developers are blocked from seeing the actual inner workings of the AI and its algorithms. AI Now calls this a conflation of open-source programs with open-access policies.
At the same time, a lack of federal regulation and unclear data privacy policies have prompted worries about unmonitored AI development. Following an executive order on AI from President Joe Biden, several software companies have agreed to submit safety tests for federal oversight before release, part of a push to monitor foreign influence. But standard regulatory guidelines are still in development.
So you may want to take into account what line of your work you’re in, your company’s partnerships (and even its mission statement), and any conflicts of interest that may overlap with using products made by specific AI developers.
Things to consider:
Who built the AI?Does it source from another company’s work or utilize an API, such as OpenAI’s Large Language Models (LLMs)?Does your company have any conflicting business with the AI’s owner?Do you know the company’s privacy policies and how it stores data given to generative AI tools?Is the AI developer agreeing to any kind of oversight?Could the AI have any relevant biases?
Even the smartest AI’s can reflect the inherent biases of their creators, the algorithms they build, and the data they source from. In the same April report, AI Now reports that intentional human oversight often reiterates this trend, rather than preventing it.
“There is no clear definition of what would constitute ‘meaningful’ oversight, and research indicates that people presented with the advice of automated tools tend to exhibit automation bias, or deference to automated systems without scrutiny,” the organization has found.
In an article for The Conversation, technology ethics and education researcher Casey Fielder writes that many tech companies are ignoring the social repercussions of AI’s utilization in favor of a technological revolution.
Rather than a “technical debt” — a phrase used in software development to refer to the future costs of rushing solutions and thus releases — AI solutions may come with what she calls an “ethical debt.” Fielder explains that wariness about AI systems focuses less on bugs and more on its potential to amplify “harmful biases and stereotypes and students using AI deceptively. We hear about privacy concerns, people being fooled by misinformation, labor exploitation and fears about how quickly human jobs may be replaced, to name a few. These problems are not software glitches. Realizing that a technology reinforces oppression or bias is very different from learning that a button on a website doesn’t work.”
Some companies that have automated services using AI systems, like health insurance providers who use algorithms to determine care or coverage for patients, have dealt with both social and legal ramifications. Responding to patient-led lawsuits alleging that the use of an AI system constituted a scam, the federal government clarified that the technology couldn’t be used to determine coverage without human oversight.
In educational settings, both students and teachers have been accused of utilizing AI in ethically-gray ways, either to plagiarize assignments or to unfairly punish students based on algorithmic biases. These mistakes have professional consequences, as well.
“Just as technical debt can result from limited testing during the development process, ethical debt results from not considering possible negative consequences or societal harm,” Fielder writes. “And with ethical debt in particular, the people who incur it are rarely the people who pay for it in the end.”
While your workplace might appear to be much lower stakes than a federal health insurance schema or the education of future generations, it still matters what ethical debt you may be taking on when using AI.