Pre-Enshittified Tech, or How I Learned to Stop Worrying About The AI-pocalpyse and Love ChatGPT
Some of us are old enough to remember when the web was the thing all the pundits, analysts and “experts” were freaking out about 24x7.
Then it was mobile. No, not smartphones, I’m talking about Palm Query Apps, WAP (no, not THAT, it meant something very different back then), and ringtones. Smartphones wouldn’t be big for close to 10 years from the time I’m talking about, the Treo was the state of the art.
Then it was Web 2.0, social networking, smartphones, crypto, etc. that investors rained money on and the media went into histrionics over while the reality of the tech was far different than the perception.
It is hard to overstate how disruptive AI will be or how misunderstood the tech, societal, legal, and economic issues are, but it isn’t the end of the world just yet, and ChatGPT in particular is a tiger made of far thinner paper than current level of media coverage might indicate.
As is the case with tools like DALL-E and MidJourney, the tendency seems to be for people to assume the new tech works the same way as things they’re familiar with.
In the case of image generators it seems like certain people believe that they’re like magazine collages or vision boards where the imagery comes straight out of chopping up and combining a small set of pre-existing work. (There are even some dumb lawsuits on this subject.)
For ChatGPT, people assume that it works like a search engine, or a natural language interface like Siri or Alexa. Those technologies do what they do by searching large databases, putting them through a series of algorithms that weight and sort the results into something useful, and then giving us output specifically relevant to what we asked for.
That is not even close to what ChatGPT actually does.
ChatGPT is a language model designed to do exactly one thing: Take natural language input and output a reply that makes sense… to itself.
It doesn’t even think past the next word in a sentence. The fact that you get the words one at a time in the UI isn’t theatric or skeuomorphic, it’s literally creating that text based on the words that came before it. (Even on the new Bing this reality means that it takes maybe 10 times as long to render the chat results as it does traditional search.)
The GPT3/3.5 model that powers ChatGPT is then augmented by some specific input and output filters that do things like prevent it from using certain words, restrict content from the internet deemed unsavory, and try to constrain it from performing certain tasks.
Beyond those filters, in terms of actual intelligence ChatGPT is closer to a parrot than a person. It makes no attempt to weigh or even understand the information it reads or writes, verify the accuracy, or in many cases even get basic facts right.
You can test this very easily by doing math. Try this one, for example:
Assuming you didn’t memorize multiplication tables if you’re my age or ace the common core if you’re younger, you may need to head to your phone’s calculator for this one.
Turns out, 26 x 78 is 2028.
Google, Siri, and Alexa wouldn’t make this mistake because they’d recognize it as a math problem and shunt it off to a specific calculator routine before even trying to answer it.
ChatGPT doesn’t have a calculator routine and as such can’t actually do math, even simple math like this.
It only knows that when the question has numbers and uses that “x” symbol between them, the answer usually kinda looks like 2008. Got a character wrong? Oh well, 3 out of 4 ain’t bad.
It’s the equivalent of MidJourney having trouble with fingers or when StableDiffusion approximates a Getty Images watermark from its training data. We call this tech intelligent but as you can hopefully see, that isn’t really the case.
As such, I don’t see ChatGPT replacing humans doing anything but closely supervised busywork in the near future… though by that I really only mean a year or two.
Combining what ChatGPT does with other applications (as Microsoft is doing with Bing) could yield some interesting results, but I don’t expect these tools will be unleashed upon a helpless population out of the gate.
Make no mistake, the last tech that was disruptive at this level is the iPhone, but I see this rollout going differently.
Here’s 5 reasons why I’m not panicking about the coming AI-pocalypse:
- There are very few companies in the world that have the hardware infrastructure, development expertise, massive training dataset, and deep enough pockets to even play in this space. Those companies are Microsoft, Google, Amazon, Meta, and (maybe) Apple.
- Of those companies, Apple and Meta bet on VR/AR as the next big thing and focused their resources on that. (Best of luck with that, I’m sure the Metaverse won’t be at all like the Facebook Portal or Apple Ping and I can’t wait to see that $3,000 VR/AR headset Apple will announce this fall. I’ll put it next to Google Glass and Microsoft HoloLens on my shelf).
- Amazon has the data and the hardware, but direct to consumer AI probably isn’t high on their list of things to do unless they’re wanting to goose Alexa. (Doesn’t seem like it) They’ll want to add AI and GPT powered language models as an AWS service to compete with MS and Google’s, but they don’t have consumer facing products other than their store, whose enshittified search is powered by payola. I’m counting them out in the near term.
- That leaves us with Microsoft and Google in an AI arms race around a tech that… doesn’t quite do what they need it to, and might also accidentally destroy their existing businesses.
- Neither Microsoft nor Google are super great and understanding what their users want and tailoring their products to deliver it. They’re far more likely to ham-handedly bash users over the head with features they may or may not find useful.
Combining ChatGPT’s language interface with something like Bing or Google Search is a no brainer, but it’s going to get a bit dicey making it work with live search data, answer math problems, and you know, not advocate for eugenics or genocide.
I’d expected Microsoft would be putting ChatGPT into Cortana vs. just stopping with Bing, but in retrospect it seems this week’s announcement was more about one-upping Google than releasing a consumer-ready product, and Google’s response was similarly half-baked.
Despite their urgency to be the one who moves first, both companies’ primary goals are not going to be making ChatGPT the best, most disruptive, and/or most useful thing possible. Rather, it will be protecting their existing revenue streams. That means this tech is going to buck the cycle and roll out Pre-Enshittified.
There is plenty of historical precedent here. Legacy players rarely fully commit to new tech that kills them and if they don’t try to quash it outright they roll out crippled versions that serve them instead of users. Early taxi app competitors to Uber, early label responses to file sharing and MP3’s, and virtually everything traditional media companies have done since the web began come to mind as examples.
The thing is, in those cases the internet evened the playing field so the legacy players couldn’t just turtle up and continue to run the table. Once the web disintermediated cab companies, record labels, and media companies they were vulnerable to upstart players that understood the new landscape better.
What’s different here is that with no meaningful competition outside of these two AI superpowers, they’ll be free to make decisions to incrementally improve their existing product lines and take shots at each other, and that’s about it.
Academics and kids in garages will still innovate at a small scale, but they’ll be acquired and incorporated into these giant players or caught and killed by them before they get big enough to cause a major disruption.
Maybe Amazon, Meta, or Apple will enter the contest via acquisition or an internal initiative, but even if they do they’ll still be playing defense on their core business vs really trying to change the game the way Uber and the iPhone did.
Early reviews on GPT powered Bing are positive, but not world changing. They’re doing just enough to take on Google Search but they’re only rocking the boat a little. It’s not the Wild West scenario we got in previous generations of tech.
Truly disruptive tech at this stage would combine the search engine and virtual assistant along with a voice synthesizer and a speech recognizer into a Jarvis type virtual intelligence, but that product would utterly gut the search ad revenue both Google and Microsoft depend on and potentially destroy the market for corporate IT solutions, desktop computers, and enterprise SaaS contracts.
I’d envisioned Microsoft tricking out the Office 365 suite with AI… except that if they bring a product to market smart enough to allow corporations to radically downsize, those employee-less companies will also cancel their Office 365 seats and that won’t look good on the ol’ quarterly SaaS revenue report.
That means that while yes, MS will most definitely give O365 enough of a boost to ensure they stay ahead of other productivity suites (notably Google’s) they’ll likely stop short of completely decimating the white collar workforce the way some articles have predicted.
Even with the latest Teams enhancements I’m wondering if Microsoft might be a bit over their skis. Teams competes directly with both Slack and Zoom, and I’ll admit the new features give it a compelling edge it lacked previously.
But… as someone who works on developing SaaS applications myself, I’ll think long and hard about using Azure as my cloud platform if Microsoft is going to make a habit of directly competing with SaaS vendors. I’m hoping AWS and Google will remain Switzerland in terms of doing things like that, otherwise Digital Ocean et al will suddenly start sounding a lot better.
If Microsoft doesn’t pull their punches and NOT go all out with AI, they’ll kill their cloud and SaaS businesses. Can you imagine that earnings call?
So what happens now?
Until quantum computers REALLY disrupt things in a few years, AI will remain a two way space race with the Enshittification baked in from day one.
Microsoft has traditionally pursued an “All things to all people locked into a closed ecosystem” model like it’s Russia going after a warm-water port. This new Bing rollout pushes people using Windows to use the Edge browser, set Bing as default search, and set MSN as their homepage in order to skip the line, so it seems like it’s a return to form for them.
They’ll roll out features but probably demand annoying concessions to use them. I’ve held off on Windows 11 despite constant pressure from my own damn computer to “upgrade” to bloated adware that doesn’t actually do anything I can’t do already. I’m sure at some point they’ll only run the AI powered stuff on Win11 and deliver electric shocks through the keyboard if I don’t use Edge and Teams and I’ll finally cave.
Google is suddenly looking like they’re in second place and honestly hasn’t truly innovated much since rolling out AdWords and GMail. Other than acquiring YouTube, Android, and Motorola Mobility all they’ve done in recent memory is reshuffle the deck chairs on Google Hosted Ap…, no sorry, gSuite, wait I mean Google Workspace and Hango… no wait sorry is it called gChat now?
Anyway, product launches and product management aren’t really Google’s core competency. It’s anyone’s guess what they’ll do now that their search business is credibly challenged by someone for the first time since they launched it, I suspect we’ll see a bit more “unpolished” moves like we did this week while they find their footing.
Despite the fact that they’re both lumbering giants that trip and sometimes fall hilariously when they try to innovate, the scale and expense necessary to play in the AI space will keep third parties from seriously challenging them, at least for a little while.
And… honestly I’m fine with that scenario. AI needs to be rolled out carefully. We can’t let it just smash into an established industry like Uber or Craigslist did, there’s too much at stake.
Maybe Pre-Enshittification will end up saving the world, who knows? (ChatGPT certainly doesn’t.)
