The AI Flash Boys
Maybe the very first catchphrase about AI I remember was early 2023, shortly after ChatGPT had launched. The company we had used to build an AI abstract tool for SREOs repeated this phrase at every meeting.
“You won’t be replaced by AI, but you might be replaced by someone that uses AI.” It has a certain ring to it, right? Catchy, memorable, a not too subtle threat to improve your skills or maybe…just maybe…get replaced.
This week, news broke of an Apple and Google partnership to replace Siri’s backend model with Google’s Gemini.
“Hey Siri, you won’t be replaced by AI…but you might be replaced by an AI that uses AI…”
Siri sucks so badly that Apple decided to pay a competitor to use their model. Apple is so concerned about continuing to fall behind that it is willing to let the fox in the henhouse…and to pay the fox $1B a year…rather than to risk more iPhone users moving to Android because Siri is so frustrating to use.
One of Siri’s weaknesses is that the model is limited to 150 billion parameters. Gemini’s is nearly 10x that - 1.2 trillion. This translates to better performance on harder, multi-step tasks (scheduling, planning, etc), and more cushion on complex tasks before the model taps out. This partnership is a bandaid meant to buy Apple time until it can catch up, but that is likely years away. Being so far behind meant they needed to immediately rent capabilities to stop the bleeding.
Nearly everyone I speak to about AI is worried that they are falling behind the competition, just like Apple. Our clients sell properties, Apple sells phones. There are similarities, I get it.
But Apple doesn’t really sell phones, they sell a personal command center. If Siri lags behind ChatGPT and Google Assistant, the entire ecosystem loses value instantly.
I know a lot of you have websites that say something to the effect of “we don’t sell real estate, we sell community” but let’s be real…you sell real estate.
Part of the reason AI is so hard to implement in real estate is that the industry has been built on a patchwork of systems that work pretty well. Not everything needs to be AI - your Excel models are already precise, AI isn’t going to make them math better.
Contrast that with some other industries whose very existence depends on keeping up:
- Apple - core to the ecosystem
- Social media - algorithms
- Military - duh
- Autonomous vehicles - safer vehicles save lives
A lot of CRE firms talk about needing to implement AI to move faster, but do they really? Our industry timelines are measured in months, not seconds.
In Michael Lewis’ 2014 book about high frequency trading, Flash Boys, a hedge fund paid over $300mm to lay 827 miles of fiber optic cable to reduce order relay times from Chicago to NYC from 17 milliseconds to 13 milliseconds. That 4 milliseconds of savings is 75x faster than the blink of an eye. These firms paid millions of dollars a year to move their servers a few feet closer to the NYSE servers. Is real estate really that time sensitive?
There is no bigger AI fanboy than me, but we are fortunate to work in an industry whose survival isn’t dependent on keeping up with every AI advancement. Chasing a 1 trillion parameter model. Improving underwriting decisions by milliseconds.
AI will be seamlessly woven into everything we do someday, but I don’t believe individual real estate firms face extinction risk because they aren’t chasing AI the way a competitor might.
Apple might be the greatest technology company of all time and has spent billions on AI…and they just threw in the towel. Let the AI flash boys spend billions of dollars in the AI arms race and just license something from them for $20 a month. Your competition isn’t going to put you out of business just because they spent a few million bucks on an AI model they can brag to investors about.
And we’re lucky that’s the case.
