What Happens When Intelligence Gets Cheap?

There was some big news in the world of math­e­mat­ics this week, when Ope­nAI announced one of their inter­nal mod­els had just pub­lished 722 maths papers. Now that may mean noth­ing to you and I but to math­e­mati­cians it seems to have been a big deal. The Econ­o­mist wrote “Tues­day, Octo­ber 6th of 2026, may go down as one of the most con­se­quen­tial days in the his­to­ry of math­e­mat­ics.” There’s prob­a­bly been quite a few con­se­quen­tial math­e­mat­ics days in the last 2500 years, so to make it “one of the most con­se­quen­tial”, real­ly says some­thing.

Accord­ing to Ope­nAI, near­ly every paper came from a sin­gle prompt to a sin­gle AI agent.

The response I’ve seen in online math­e­mat­ics forums is a com­bi­na­tion of shock, res­ig­na­tion and out­rage.

The cra­zi­est part of this is that a year ago, even the lead­ing AI mod­els could still stum­ble over basic maths.

I’m not going to pre­tend to under­stand what all these math­e­mat­i­cal papers mean. What I am inter­est­ed in is how this trend might play out across oth­er domains that could have an impact on invest­ing.

What hap­pens when intel­li­gence gets cheap enough to be applied to big prob­lems at an enor­mous scale?

When peo­ple in online AI com­mu­ni­ties talk about how AI is gob­bling up prob­lems, they call it being “solved”. Some of the hard­est domains of human thought have already been solved. Math­e­mat­ics is solved. Cod­ing is solved.

If the AI race con­tin­ues to progress at the rate is has been over the last four years, where we’ve gone from “wow Chat­G­PT 3.5 can have a con­ver­sa­tion” (Nov 2022) to “cod­ing and math­e­mat­ics are solved”, will we start to see every prob­lem in every domain solved?

Physics? Biol­o­gy? Med­i­cine? Cli­mate change? Eco­nom­ics?

I don’t know the answer, but we have to at least con­sid­er the pos­si­bil­i­ty that there will be extreme­ly fast progress in at least some of those domains.

So what does that mean for invest­ing?

I’ve thought a lot over the last few years about how we might use AI to improve QAV’s results or how to make QAV more effi­cient and, hon­est­ly, apart from using it to code the spread­sheet process, I haven’t real­ly come up with much. Every time I ask AI to come up with ways to improve it, it does­n’t have any­thing either. As I explored in a recent newslet­ter, prompt­ed by some of Tony’s thoughts on the pod­cast, it seems sta­tis­ti­cal­ly improb­a­ble to be able to con­sis­tent­ly return results greater than dou­ble mar­ket. Par­tic­u­lar­ly if you’re spend­ing an hour a week run­ning your port­fo­lio and don’t want to do it as a full-time job.

So I’m not sure if that’s even a direc­tion worth invest­ing time and ener­gy into.

A more inter­est­ing line of enquiry is how AI might affect the econ­o­my in gen­er­al and how the busi­ness­es that we can invest in might be impact­ed by AI in the next few years and whether or not we should be pay­ing atten­tion to that when we are scor­ing a com­pa­ny. The obvi­ous one every­one turns to is job loss­es and com­pa­nies prof­it­ing from low­er wages, but I think there will be short­er term impacts.

To be clear — philo­soph­i­cal­ly, QAV views fore­cast­ing as akin to witch­craft and home­opa­thy, so we don’t want to start scor­ing com­pa­nies based on what MIGHT hap­pen. If ear­ly adopters of AI are able to extract more prof­it, more cus­tomers, more effi­cien­cy, or more pro­duc­tiv­i­ty from their busi­ness, that will show up in their results, and look­ing at their results is how we already make most of our buy­ing deci­sions. But it’s fun to think about what the future might look like for some of the busi­ness­es that rou­tine­ly show up on our buy lists.

For a few years now I’ve been think­ing about busi­ness­es that prof­it from cus­tomers hav­ing lim­it­ed knowl­edge or lim­it­ed time in order to make buy­ing or switch­ing deci­sions, or busi­ness­es that rely on cus­tomer con­fu­sion and iner­tia. How might these busi­ness­es be affect­ed by cus­tomers sud­den­ly hav­ing access to unlim­it­ed intel­li­gence? Stud­ies show that sub­scrip­tion busi­ness­es ben­e­fit from cus­tomer iner­tia. We stay sub­scribed to ser­vices we either for­get that we’re pay­ing for or we don’t spend the time and effort to check to see if we can get a bet­ter deal some­where else.

In Robert Hein­lein’s first pub­lished nov­el, Life-Line, the basic plot is about a sci­en­tist who dis­cov­ers a way to accu­rate­ly pre­dict the time and date of some­body’s demise. At one point, he gets sued by life insur­ance com­pa­nies who say that cus­tomers are can­celling their poli­cies because this new inven­tion makes life insur­ance poli­cies point­less.

And no — I don’t think AI is going to be able to pre­dict the time and date of our death.

But one thing it can prob­a­bly already do is assess the sub­scrip­tions that I’m pay­ing and then help me make deci­sions about them.

In fact, while writ­ing this arti­cle, I down­loaded my recent bank state­ments, point­ed my AI assis­tant at them, and instruct­ed it to analyse all of my reg­u­lar pay­ments and sub­scrip­tions and to repeat that job on a month­ly basis. It quick­ly picked up that the New York Times had recent­ly increased my sub­scrip­tion fee from $2 a month to $24 a month, which had escaped my notice. So I imme­di­ate­ly act­ed on that and got my orig­i­nal NYT deal renewed. Boom! A $22 a month sav­ing, which coin­ci­den­tal­ly pays for a basic AI Pro sub­scrip­tion.

Most peo­ple prob­a­bly aren’t using AI to analyse their bank state­ments yet, but I’m pret­ty sure that will hap­pen fair­ly soon. It will creep up on them. Google already has AI scan­ning Gmail to flag what’s impor­tant and what’s like­ly to be spam. It won’t be long before it offers to act on your email on your behalf.

“Hey Cameron, I notice you haven’t opened this week­ly newslet­ter in months. Maybe you’d like me to unsub­scribe you?”

“Hey Cameron, I notice you’ve been pay­ing for a gym mem­ber­ship / Net­flix sub­scrip­tion that you haven’t been using. Want me to can­cel that for you?”

“Hey Cameron, I notice you just got an invoice from Youi Insur­ance. Do you want me to com­pare that to oth­er providers and see if I can get you a bet­ter deal?”

Right now I have my AI assis­tant get­ting me com­par­a­tive quotes for car insur­ance. It is find­ing enquiry con­tact forms on insur­ance web­sites, fill­ing out my car details, and then mon­i­tor­ing my email for respons­es, and com­par­ing them to find the best deal. I had to tell it to do that, but how long before it comes to me with those sorts of ideas? New ser­vices like Ope­nAI Dots are already mak­ing those kinds of proac­tive deci­sions but it’s some­thing only bleed­ing edge AI cus­tomers are explor­ing. Pret­ty soon it will go main­stream.

What per­cent­age of sub­scrip­tion rev­enue is at threat from AI assis­tants? Which busi­ness­es might take a hit from that in the next few years?

Some busi­ness­es prof­it from infor­ma­tion asym­me­try: they know more about what they’re sell­ing than their cus­tomers do. What hap­pens when the cus­tomer has an AI help­ing close that gap?

For exam­ple, imag­ine walk­ing into a car deal­er­ship (or real estate agent / insur­ance agent / bank man­ag­er / finance bro­ker / build­ing con­trac­tor ), and plonk­ing your phone on the table dur­ing your con­ver­sa­tion so it can fact check the sales per­son­’s claims in real time.

Today we tend to get three quotes on any piece of seri­ous work we need done but your AI assis­tant will be able to get you 300 quotes, read through all of the con­tracts, ask per­ti­nent ques­tions, clar­i­fy ambi­gu­i­ty and strike you the best deal.

So hav­ing an AI assis­tant will be great for cus­tomers — but what does that mean for us as investors?

Will we have to look for busi­ness­es that have rev­enue mod­els that can’t be white-anted by AI?

I don’t know the answer to that ques­tion yet — if you have any ideas, let me know.

---

QAV Myth Killers is a week­ly col­umn in the QAV newslet­ter, tak­ing apart a piece of
invest­ing con­ven­tion­al wis­dom. Read the series, or
get it by email every Fri­day.

Gen­er­al advice only. Space­craft Pub­lish­ing Pty Ltd
trad­ing as QAV is a Cor­po­rate Autho­rised Rep­re­sen­ta­tive (CAR 001292718) of MF & Co.
Asset Man­age­ment Pty Ltd (AFSL 520442). This is gen­er­al infor­ma­tion and does not take your
per­son­al cir­cum­stances into account. See our
dis­clo­sure.