I like to think myself as the business equivalent of a polyglot - proficient at doing a lot of things! Which has given me a lot of pause in online and in-person conversations of late.
There’s so much discourse about the functions people offload to AI (ok… let’s be clear when we say AI, we mean the new wave of LLMs or traditional data science/ML solutions masquerading as cutting edge AI). My feed is littered with posts from people sharing the AI tools that changed their life. I have had so many discussions with business leaders I know about the constant stream of new product pitches they get centered around AI. But I’m a skeptic by nature!
The AI Skeptic’s Question: what are the spaces where the state of AI can perform better than a human?
And I truly mean “better”, rather than just augment.
I’m a bit rusty at programming (hey, it really hasn’t been a focus of my day-to-day professional life), but there’s a litany of AI coding tools that could make me a coding superhero, right?
That meme, accurately portrays the new normal for AI-enhanced programming (at least according to a friend that used Claude to help him build the prototype he’s exploring for a startup).
Learning from technology enhancements in our past
As a product person, I gave some thought to what are the truly under-solved challenges, that we as humans (biological computers!) fundamentally struggle with, where the technology can bring a force multiplier. And digging into my past I found an analogue - that I think applies.
Even if you know me well, you probably don’t know that in the many lives I’ve had, one of them was as a vaccine researcher. It was literally decades ago, at the start of my career…
I was a young, whip-smart college kid, with a lusciously full head of hair (yup it was a LONG time ago) - and finagled myself a job working with some absolutely brilliant scientific researchers at the Institute of Human Virology at the University of Maryland School of Medicine (IHV). To place it - the Human Genome Project had just hit major milestones, and these brilliant scientific mind’s at the IHV had the idea to use computational biology to design a novel HIV vaccine, that could be synthesized, tested, iterated, and ultimately brought to market. Designing and synthesizing a protein vaccine was an expensive proposition - especially for a largely grant-funded institution - but in hiring an inexpensive, albeit brilliant student with experience in bioinformatics and applied science, and modeling everything in a computer - this could be done amazingly efficiently. Laboring literally my entire undergrad career, we took several vaccine candidate through the pipeline - each rigorously validated with the best computer models we had at the time, before the expensive process of taking them to the lab.
And weeks before we were set to publish our findings… we were scooped by a pharma!
Rather than approach the vaccine development problem in the creative, cautious, funding-restricted manner, the pharma (flush with cash) had synthesized every permutation, and tested them all. They didn’t rely on experience or creativity - they used their super power (of having money) and made it a big data processing problem that they could excel at.
Where Humans Struggle, AI Excels?
As I hope my anecdote conveys, I believe AI will be a force multiplier when it comes to drug discovery - where relying on big data and massive processing capability enables the production of a solution space that a human might otherwise have to “luck into”.
And in candor, while this may be coming back to first principles in a discussion on AI, I think we ought to segment problem spaces and how we apply tools - those where creativity and experience are a key asset to “generating a new solution” and those where the ability to process massive amounts of data can yield novel insights and avenues of solution.
In that vein, I’ll leave you with this question - next time you consider an LLM or AI product to solve your problem, will you be asking, how important is human creativity in finding the best solution?
