Why We Need to Help AI Help Us

By James Kupernik, CTO at VidMob

AI has reached milestones that everyone can understand; beating the world’s best chess player, winning at Jeopardy, and today, creating art and articles that can pass for something created by a professional.

For people in creative professions, from designers to writers to computer engineers, apps like ChatGPT and Midjourney can be really scary. While the inputs that combine to deliver the content they create is up for debate, the speed and quality of the output can’t be denied. The obvious question for a creative becomes, “Will AI eliminate my job?”

Certainly, AI will eliminate some jobs, but not for the reasons you might think. The very best creatives will interact with the AI – collaborating, training and molding the output to push their work further and reduce manual work. Whatever AI can deliver, a creative can shape and mold output to make things faster, more scalable, more unique, more thoughtful, and more effective. What’s more, the more that people interact with AI tools, the better and more useful they become. In the creative economy, AI probably won’t take your job, but those that use it better will.

What to Do With Unwieldy AI

An executive at my company recently asked if I could help with a particularly thorny query of our data. It required a query that would unite several tables in a complex way that could easily bring back faulty results if it wasn’t structured correctly. I decided to feed the parameters of the query into ChatGPT  to help speed my process, and got back a incorrect results on my first try. My second try wasn’t much better.

I’ve talked to a number of engineers who have encountered the same issue. There are plenty of these tools out there – GitHub has one, AWS another, Chinchilla, Bloom, the list is growing by the day. The thing is, each one of these tools sounds just as confident whether or not the answer is correct, so the user needs to have a certain level of institutional knowledge to scrutinize the results, and experts on a particular topic can get frustrated when results are repeatedly unusable. Super smart engineers often have a sort of knee-jerk reaction to roll their eyes and throw out an example or two of how the tools don’t work well for “real engineering problems” or fail to see how this might work into their existing workflow. To this, I counter that we need to put in the time to learn how to write better prompts and help improve the tools.

Right now, AI tools are great for building out the foundational code that can be a time suck for engineers. Using rules, best practices and templates to build out the basic structure to get a project started poses little threat to ambitious engineers and provides efficiency. I’m all for it. This is just the first step. The more we push AI, the more it can do for us in the future, and the more time we can spend on truly tough problems

AI Will Get Better. Will You?

We shouldn’t be dismissive of where AI will go. Time savings, manual labor – it’s great that AI will help us with these things. But, that’s not enough. We should be excited, encouraging, and collaborative to push AI further and push ourselves further.

In this spirit, I didn’t stop after my first two attempts to get a good query from ChatGPT. I refined my instructions a couple more times, and got back something that…didn’t work. But just like engaging in a pair-programming session, I provided the tool feedback and it did make a suggestion that transformed my thinking and allowed me to put a query together that worked.

I think about it as an apprentice sitting with me as I work. Today, that apprentice is young and a bit naive, but they have promise. AI for complex engineering won’t always disappoint super smart engineers. Tomorrow, that apprentice gains knowledge and becomes a collaborator. Forward-thinking engineers won’t fear this inevitability, but rather prepare for it and help shape it to be as valuable as possible. .

Already, designers are feeding client requirements into AI systems to get mood boards back that not only save time, but foster new ideas that they might not have had themselves. This kind of human-computer collaboration is right around the corner across many more disciplines. When AI can suggest really smart approaches, we shouldn’t be scared, we should be inspired. We can take that information and work with it.

When the Student Becomes The Teacher

Ben Thompson recently noted on his January 12th Sharp Tech podcast that people often take a new technology at face value based on how it works today with little foresight into how it may work in the future. In the Febuary 14th Master of Scale episode, Reid Hoffman said, “Entrepreneurs have to be skating to where the puck is going, not to where the puck is.” If we look ahead a year or three, AI will become significantly more powerful. One day, our apprentice will become a teacher, coach and/or collaborator. This still does not mean that we are no longer valuable to the process – if we rise to the occasion by leaning into what it can do for us today and prepare for the future by evolving the way we work.

GET IN TOUCH
Got a Question? We’ve Got Answers.