Why You Should Define Your AI Goals Before Using A Service Like ChatGPT

Last July, the United States State Department published the Risk Management Profile for Artificial Intelligence and Human Rights. This profile applied processes from the U.S. National Institute of Standards and Technology AI Risk Management Framework (AI RMF), and suggested actions that government agencies and private businesses could take to implement the AI RMF. When recommending actions to implement the AI RMF's "Map" function, the State Department recommends that "the intended purpose for the AI system should be well-specified and finite." In other words, an organization - or, as I will I argue, an individual - should know why they want to use AI before they start seriously using it.

Artificial Intelligence is the hottest tech idea of the moment. Everyone from Fortune 500 companies to my mother seem to want to get in on the trend. Why, though? Talk around AI resonates with vague ideas about how it will make you more productive, more creative, even less anxious. "AI" can help you hire employees (never mind that it could filter out the best applicants), or develop a company-wide knowledge base (never mind that it could let your employees read the CEO's emails). "AI" - as though "AI" were one thing, one singular anthropomorphized resource: my office assistant, Mr. AI. My teacher, Ms. AI. My friend, AI.

But "AI" means many things, and the various AI products can do many things - many positive, productive, creative, and useful things, and also many inadvertently or intentionally negative and harmful things. Or, relative to the software tools you already use, it might do nothing at all, while costing you or your employer a tremendous sum each month. Rather than succumb to shareholder pressure or social influence, organizations and individuals should dispassionately assess whether an AI tool is the right choice for them. And the only way to assess whether AI works is to first figure out what you want AI to do.

Do you want help organizing your schedule? You could use a handwritten calendar. Or, as I tried earlier this month, you could use AI to help you figure out the best times to schedule appointments. Both have worked for me – with upsides and downsides to each. Need creative ideas for your next personal writing project? You could find inspiration in analogue tools, like Brian Eno’s Oblique Strategies, or you could use a chatbot as a sounding board to help you brainstorm. There are certain tasks that AI products have been particularly good at – identifying patterns in data, or automating simple repetitive tasks. But there are other situations where the utility of an AI product is less certain.

First, define your problem. What issues do you, your business, or your organization face that have not been solved by existing tools? Or, what existing solutions have not been fully satisfying, that you think could be more efficient, or more thorough? Write that down. Next, take a look at the AI products that are available. The current AI landscape is not a monolith – there are a lot of options out there. If you want a Chatbot to help you draft documents or brainstorm ideas, you could consider OpenAI’s ChatGPT, Anthropic’s Claude, or Google’s Gemini (née Bard). Looking for assistance editing an existing document? Maybe try Claude as a standalone solution, or Microsoft’s Copilot if you use Microsoft 365 and want something integrated into your existing work environment. Trying to create images (an AI skill that, while very cool, is rife with potential copyright and ethical issues – but that’s a subject for another article)? Try OpenAI’s Dall-E, Stability AI’s Stable Diffusion, or Midjourney (by a research lab of the same name). Trying to search for answers a la Google? Well, Gemini will pop up at the top of your more traditional search results, or you could use Perplexity AI, which integrates a few different AI engines.

There are a lot of products out there, and they each excel at different tasks. Once you’ve chosen your tool, the next – and most important – step is to assess. How well does your AI product solve your problem, or improve your pursuit? Does it improve anything at all? Or, does it leave you worse off? Could it be harmful – to you, your mental wellbeing, your scholarly independence, your true creativity? This assessment should be ongoing, as AI products and technologies are developing at breakneck speeds, and one tool might be better than another a mere few months from now. Artificial Intelligence, generally, has great potential to improve society and all of our lives, but you cannot know how much it helps – or hurts – unless you first figure out why you want to use it. Don’t just jump in on the trend. Define your goals first – then leap.