Ethan Mollick, a outstanding voice in AI and innovation, simply printed an vital essay titled “Prophecies of the Flood.” His thesis? We’re seeing pressing new indicators that superintelligent AI will not be a long time away—it could be simply across the nook.
Mollick isn’t alone. A rising variety of researchers from high AI labs are sounding alarms (or ringing bells, relying in your perspective) that we’re nearing a basic breakthrough in AI. And these rumblings aren’t the everyday tech hype—many are coming from individuals inside main labs who seem genuinely satisfied we’re on the point of one thing large.
So, what precisely are these so-called prophecies? And the way ought to companies and people make sense of the frenzied dialog round superintelligence?
I obtained the solutions from Advertising and marketing AI Institute founder and CEO Paul Roetzer on Episode 130 ofThe Synthetic Intelligence Present.
Mollick’s “Alerts of the Flood”
Mollick calls out a number of indicators that superior AI could be arriving prior to most individuals count on:
OpenAI’s o3 mannequin. This new mannequin reportedly hit an 87% accuracy on a check that even PhDs, utilizing the web, solely averaged 34% on—in the event that they ventured past their specialty. It’s additionally fixing high-level math issues and crushing ARC AGI benchmarks.
Early AI brokers. Mollick factors to Google’s Gemini and its Deep Analysis function, highlighting how he requested it to investigate 173 web sites and generate a 17-page report in minutes. That’s principally a customized AI researcher at your beck and name.
Researchers sounding the alarm. The place typical tech hype may come from entrepreneurs or evangelists, Mollick notes these alerts are coming from inside AI labs—a serious distinction from the same old cycle of massive product guarantees.
Nevertheless, regardless of the drumbeat round superintelligent techniques, Mollick reminds us to maintain a degree head. Even when AI labs attain one thing resembling “AGI,” mass integration into companies and society typically lags behind the lab breakthroughs.
He additionally warns that whereas AI researchers deal with alignment and security, much less consideration appears to go towards how society ought to adapt to (or form) this highly effective expertise.
Why Most Individuals Don’t “Get It” But
Roetzer factors to a telling tweet by Vedant Mishra, a DeepMind researcher with previous AI stints at OpenAI and HubSpot. Mishra says solely a “few hundred individuals on the planet” really perceive the scope and velocity of what’s coming, as a result of it’s a must to concurrently grasp:
Fast algorithmic enhancements
Recursive self-improvement through reinforcement studying
Tens of billions of {dollars} dedicated to AI knowledge facilities (“intelligence factories”)
Both these consultants are all incorrect, Mishra says, “or every little thing is about to vary.”
The Math That Adjustments The whole lot
To get a way of how a lot is already altering—and the way unprepared companies are—Roetzer then laid out a easy however startling instance:
Utilizing at present’s AI capabilities—no future breakthroughs required—you possibly can possible unlock a ten% effectivity acquire in almost any data employee’s day proper now.
What does that imply? Some back-of-the-napkin math:
A single worker working 176 hours per thirty days saves about 18 hours at a ten% effectivity increase.
In a 10-person firm, that’s 180 hours a month complete—equal to a full additional worker.
In a 1,000-person group, you’re successfully liberating up 100 workers value of time every month.
Scale that to 10,000 workers, and its 1,000 full-time equivalents of month-to-month productiveness, with out hiring a single further individual.
And that’s in case you solely get 10%. Realistically, Roetzer sees 20–30% as utterly attainable with higher coaching and adoption—at present, proper now, with off-the-shelf AI instruments.
How do you try this? Roetzer recommends these steps:
Map out duties. Take one individual’s position, break it down into 3–5 main duties they carry out repeatedly.
Construct 3-5 customized GPTs. These AI “assistants” deal with chunks of the individual’s workload—like drafting content material, analyzing knowledge, or summarizing analysis. Construct 1 (no coding required) to help with every of the individual’s main duties.
Understand the financial savings. Even a modest 10% effectivity bounce interprets to main productiveness beneficial properties when multiplied throughout your complete firm.
Now, think about what turns into doable if superintelligent AI turns into a actuality.
Why This Issues (Even If You Don’t Care About “Superintelligence”)
However, on the finish of the day, this issues to what you are promoting proper now whether or not or not we attain superintelligence, says Roetzer.
Whether or not your group goals to scale back headcount or broaden output, ignoring AI’s effectivity beneficial properties is a large missed alternative. Opponents who undertake rapidly will acquire the higher hand.
The largest boundaries are lack of AI literacy and inside inertia, not technological limitations. The instruments exist already; utilizing them is the problem.
As Mollick and the labs chase superintelligence and superior AI brokers, plain-old generative AI is quietly reshaping firms from the within out.
“Everyone has this selection,” says Roetzer. “You may hold doing what you are doing, or you possibly can settle for that you may change issues and also you speed up your AI literacy and capabilities.”
Backside Line: Don’t Watch for the “Flood” to Hit
Even in case you’re not satisfied superintelligence is true across the nook, a metamorphosis of the workforce is already underway. As Mollick, Altman, and numerous insiders hold warning, ignoring AI now may very well be like ignoring the early indicators of a flood.
A ten% effectivity acquire at present—with the potential of doubling or tripling that within the close to future—might make or break organizations within the subsequent 1–2 years. Whether or not it results in workforce reductions or an explosion of recent initiatives, the consequence is identical: AI is altering the calculus of enterprise quicker than any expertise in dwelling reminiscence.
Mollick’s query isn’t if the flood is coming—it’s whether or not we’ll be prepared when the waters begin to rise.
