Unless you’ve been living under a rock for the last 6+ months, you’ve probably heard a thing or two about large language models.
Regardless, there’s a good chance that a relative gave you a demo of ChatGPT last week at Christmas lunch and it blew your mind.
And rightly so.
The potential applications for generative AI are endless, and OpenAI is gunning for top spot. GPT models for language, DALLE-2 for images, and almost certainly some audio and video models in the works as well. But they’ve also got competition, and it’s starting to look an awful lot like the smartphone space in 2007.
Even though I was 13 at the time, hear me out.
Stage 1: New tech that causes a social shift
It was 2007 when Apple released the iPhone, and suddenly everyone wanted a smartphone. Gone were the vestigial parts of competing smartphones, like physical keyboards or scrollwheels. While other brands tried their best to compete, Apple’s choice of a capacitive screen opened up new ways to interact with a small display.
But it wasn’t until the AppStore was released in 2008 when the real magic happened. It opened the floodgates for developers, revolutionised the way we interacted with our phones via 3rd party apps, and has since transformed the way we do everyday stuff.
OpenAI are now experiencing a similar level of recognition after releasing ChatGPT in November.
It’s their iPhone moment.
It’s their turn to receive “this-feels-like-magic” feedback from the general public, not just from a tiny sample of fringe enthusiasts on the corner of the internet.
This has been a long time coming, as language models and chatbot technology have been around for quite some time. But the level of innovation that OpenAI has achieved with GPT-3 and beyond has been mindblowing to say the least.
So if 2008 is the App Store, then 2023 is ______?
I’d say it’s GPT-4, which could be released as early as Q2 this year.
It was initially rumoured to be weighing in at a whopping 100T parameters, making it magnitudes more powerful than GPT-3 on paper and could very likely pass the Turing test. This has since been shot down by Sam Altman, but any improvements over the next few years will still feel like icing on the cake.
Given the level of adoption that ChatGPT has driven in only a few short months, the release of GPT-4 would probably be the tipping point for significant change to humanity. The only difference to the smartphone race is the massive potential risk (and toxic engagement farming), but more on that another time.
Stage 2: An open-source competitor enters the chat
The first Android phone, the HTC Dream, was released a couple months after the iPhone 3G and iOS 2.0 in 2008. While Android wasn’t nearly as popular as the iPhone at first, it eventually caught up and went on to become the most popular mobile operating system in the world. Being open-source quickly led to a thriving ecosystem of developers who could now tap into a global market of potential users.
In the world of AI, Stability is shaping up to be Android’s equivalent (loosely).
As the creators of Stable Diffusion say in their own words:
“AI by the people, for the people. Designing and implementing solutions using collective intelligence and augmented technology.”
The company functions on the core goal of democratising AI and putting it in the hands of as many people as possible. This is in contrast to OpenAI, who despite their name, are far from open. As a result, I’m betting on a similar Android vs iOS level of innovation pingpong over the next 1–2 years.
Stage 3: An immovable object gets moved
Now I’m not saying that Google in 2023 will suffer the same fate as Nokia in the 2010s. But if anything were to trouble a titan like Google then it would be something like this.
With Microsoft and OpenAI teaming up to incorporate ChatGPT functionality into Bing, a true underdog comeback might just be brewing. By combining the real-time-relevance of search with the highly specific answers from the AI model, we’ll get a glimpse at the next era of the information age.
As Google’s primary revenue stream is from search ads, anything that challenges the current model would raise red flags internally. There’s a rumoured “code red” in Mountain View at the moment, which is probably not much different to the scenes at Nokia HQ 15 years ago.
If there’s a time to use the D word in this post, it’s probably now. Google may well become disrupted and it may just trigger a chain of events — just as Nokia was seemingly untouchable in 2007.
I wouldn’t bet against Google just yet. But it’s definitely possible.
Stage 4: The Wild West
Flashback to July 2008 when we could do so many incredible things on our phones. Remember the app that showed a beer on the screen and used the accelerometer to make it look like you’re drinking it? It was truly the pinnacle of technology. Surely nothing would be able to top that, except maybe a game where you fling cartoon birds at green pigs.
Flash forward to January 2023 and realise it won’t be long till we refer to moments as being pre-GPT-4 or post-GPT-4.
Call me a sensationalist, but things are about to get hectic.