We're all just training data

Through the AI Looking Glass

Welcome to 50 or so new readers who have joined since last week! If we don’t know each other, reply to this email and say hi. Keeps me out of spam and I make a new friend.

Turning on a few light paid growth levers for the first time next week to make this thing a real venture so this felt like my last week to really let the colorful prose fly. Begrudgingly PG-13 from here on out.

A robot kindergarten class photo by DALL-E

In the immortal words of Tallahassee, 2024 is the year for AI to nut up or shut up. After a year of fawning coverage, the headlines around AI are starting to read a little more like this: 

As Benedict Evans says, the vibe in tech right now is still AI and everything else. With that lens, it’s hard not to view every single little innocuous move and product update from large tech companies as castling on the AI grand strategy chessboard.  

Case in point: Last week, Amazon quietly increased its backend keyword limit from 200 to 500 words, a move largely celebrated by sellers and other operators on the platform. The obvious interpretation to this is that with every passing day, Amazon becomes an increasingly commoditized marketplace and opening a longer tail of keywords helps sophisticated operators find pockets of SEO juice. But dig one level deeper and you see it’s just another training da(y)ta source for AI.  

While the AI battleground was initially fought in the more esoteric corners of tech where Silicon Valley sells to Silicon Valley, both Open AI and Anthropic are starting to sign FORTUNE 500 businesses that drill deep into mainstream America. OpenAI counts Estee Lauder, Carlyle and PwC on its client roster while Anthropic is servicing 70% of the largest banks and insurance companies in America along with LexisNexis and Pfizer. 

What companies that touch the mainstream American consumer do with the training data we unwittingly have been feeding them will define where AI lands on the hype cycle by the end of this year. Like Swifties becoming experts in Cover 2, AI impacting our daily lives will happen gradually, then suddenly. 

In that vein, the most visceral reaction I had to a product launch in 2023 was when Shopify announced a new AI tool called Sidekick. At the time I wrote the following (in a separate newsletter I write for a trade publication called Martech Record)

One of the most abused quips in Silicon Valley is the Arthur Clarke line that “any sufficiently advanced technology is indistinguishable from magic.” But every once in a while, there’s a product launch that truly evokes that kind of emotion. For me, it was Tobi Lutke’s video debuting Sidekick.

To date, most AI applications have basically claimed to essentially be automation. There’s no shame in this–hundreds of billions of dollars in enterprise value come from automating mundane shit or corporate theater. Despite the cutesy name and framing, at the core, Sidekick teases something very different. With the level of functionality they’ve built in, Shopify is insinuating that AI can be better than individual operators at the core competences of growing an eCommerce business. If they’re right, it will change the fundamental trajectory of eCommerce. 

Sidekick knocked me on my ass because it wasn’t promising to just automate the backend tasks of eCommerce. It was promising to fully run the key functions that separate the good brands from the great. Looking back now 6+ months from launch, I don’t really know any commerce operators leveraging Sidekick for core business functions. Whether or not that changes is a big theme that I’ll be watching closely. 

When it comes to desperately trying to make AI happen, it is my old friends at LinkedIn who set the standard. 

From a grand strategy perspective, LinkedIn is staring down the barrel of a fascinating opportunity. With Twitter engagement cratering and Elon Musk potentially six months away from getting bored and dumping this thing on private equity, LinkedIn has a real chance to become the place that actually drives the professional narrative in tech. As recently as 18 months ago, that sentence would have read like comedy. For all its cringe, LinkedIn possesses a secret weapon here– a large team of highly talented reporters and editors who, if deployed correctly, can both curate and evoke conversation. 

With this golden goose in front of it, LinkedIn is choosing to focus on…checks notes….. force feeding AI into the core member experience via weird prompts to hit up recruiters with AI-generated messages. And in one of the most shamelessly egregious ways to build up an AI archive, the company has introduced collaborative articles. 

I love LinkedIn. All told, I think it’s the most unequivocally net positive big tech product out there. But holy hell, collaborative articles is putting LinkedIn on a path to becoming the undisputed thirsty bitch of the AI wars. 

If we are to have an AI dominated future, I believe it is our patriotic duty to make it as weird and fun as possible. To that end, I treated the LinkedIn AI to a little training data when it interrupted my day to ask me “how to best track customer behavior” as an eCommerce website?”  

Chuck Schumer be damned, the future runs on wintergreen ZYN.   

Amazonia 

My quick take on the week’s most interesting story in the Amazon ecosystem

Wanted: Attorneys to litigate against Amazon: When Lina Khan is on Twitter still trying to build the team that is going to litigate FTC vs. Amazon, I gotta say that reads like advantage Amazon.

Look, I’ve always given Khan a lot of the benefit of the doubt. As a writer, how do you not love someone who made a cult classic out of a 50 page law school paper and parlayed that fame into one of the most quietly powerful posts in America. If nothing else, Khan is living proof that if you spit out a little fire content, good things can happen.

My take on Khan has been that many of her perceived missteps have been chess moves in a decade long battle to win hearts and minds before FTC vs. Amazon actually goes to trial. Take some Ls but get some perceived overreach and bullying behavior from Amazon on the record.

But for me, blasting out the FTC job posting on main just feels kinda weak and desperate? In any event, FTC vs. Amazon isn’t expected to see a courtroom until 2026 at the earliest so there’s plenty of time to assemble a squad

Dispatches from America

A potpurri of vibes from across the land