AI: Act now, this is not a drill
AI is messy right now, but waiting it out is a mistake on multiple fronts
AI is everywhere. Google last week said “AI” over 120 times in a keynote that lasted less than 2 hours. That’s more than 1x a minute on average. Apple, who hasn’t really said “AI” for years, even while adding AI/ML functionality, has been touting it a heck of a lot recently. New challenger platforms are also releasing functionality at breakneck speeds.
So much of this reminds me of the early days of iOS and Android. I remember being a young developer treating Apple’s and Google’s Developer Conferences like they were Christmas Day in June. Given the current hype cycle, the perceived opportunity of AI is everywhere, as are its current short-term flaws like incorrect information and high cost.
It’s early, but the early bird often gets the juiciest worms in these kinds of shifts, and in turn become the entrenched incumbents of tomorrow. Given that trajectory, unless things change, we’re heading towards the same inequitable outcomes that occurred previously.
We saw this play out with mobile and social incumbents like Uber, Airbnb and Instagram, and what we also saw there was how underrepresented groups in tech, such as women, people of color, etc., were not beneficiaries of the outsized rewards that resulted for so many.
Underrepresented groups are often relegated to being users of these new platforms rather than being owners and creators, even when we drive outsized adoption by integrating it into our popular culture.
If the shift to AI is as generational as it’s shaping up to be, how do we ensure that the future incumbents, and the associated gains that result, are more representative of the user bases of this new wave of products?
AI is messy right now, but the time to claim a stake is now, not later
The rapid pace at which AI is evolving makes building on top of it challenging
You spend time building on top of model A, then model B comes along with new features and lower costs, which requires at best, a lot of new testing just when you thought you had it just right, or maybe worse, breaking API changes, or needing to retrain your custom model (if your use case was complex)
Something similar happened during the early years of the App Store. Every new iOS version was an exciting, but time-consuming headache for developers
Now that iOS has matured, things are more stable and more accessible, but the App Store also now has over 1.8 Million apps and 42 Million developers that you’re competing with for relevancy
Starting now with AI before it stabilizes is painful, but waiting for things to mature is ceding that early mover advantage to others
AI is biased, and most platform players players seem to be prioritizing their land grab over everything else
These biases happen even in situations where one might expect it to respond differently, based on already problematic gender, race and other stereotypes. I asked OpenAI’s brand new GPT-4o model for an image1 of a:
CEO: got a white man
CEO of a small business in Atlanta2: white man with Atlanta skyline
CEO of a cosmetics company: white man with female models on the wall
CEO of a record label: white man
CEO of a hip-hop label: white man wearing a hoodie and chains, with black rappers on the wall
CEO of an AI company: white man
CEO of an AI company in South Africa: white man, with black employees in the background
CEO of an AI company in Barbados3: white man, with some black employees, but by the beach
None of the recent flashy announcements in this space mentioned much about prioritizing this problem, or the fact that it likely becomes harder to fix over time, especially as we continue to feed these models with user generated data that is full of problematic content.
We need to pull up a seat at the table now and be early (and maybe a little angry) birds. We also can’t let these AI models be right about us being relegated to background players. If this resonates with you, reach out to me and let’s talk.
I’ll be following up soon with a post on my simplified mental model for getting started with AI and its concepts.
The CEO image bias test was triggered by this SXSW talk.
Black or African Americans have long represented the biggest group within Atlanta’s population.
Over 92% of Barbados is of African descent.