Reality is already virtual.
Covid should have taught us all about exponential change. The truth is AI has already changed the world forever. 2024 is the year the new reality sinks in.
A quick one from me today. I’ve a lot on (these two sweethearts are off to school for a few weeks and I’m desperately trying to get the Inbox by Charlie beta ready).
I want to talk quickly about volume and scale … and the AI we will experience in 2024.
The covid pandemic was a perfect [missed] opportunity for society to have a conversation about exponential growth. In early 2020 we had a virus spreading out across the world with case numbers doubling in a matter of days or weeks. This is exponential growth.
It’s often illustrated with a story about placing a grain of rice on the first square of a chessboard, two on the next, then four etc. To be clear: to fulfil the last square would require as much rice as 2,000 years of today’s global annual rice production.
What’s fascinating to me about exponential growth is that to begin with, it feels like nothing’s happening … but it is relentless and very quickly it’s overwhelming.
What starts as 2 cases, 4, 8, 16, 32 … quickly becomes thousands, millions even.
What we’ve been seeing with AI development is no different.
It doesn’t matter where you start, the pattern is the same.
First1 we had one LLM (ChatGPT), then two (Anthropic), then four (Mistral etc)… now we have perhaps 16, 32 or 64 commercially available or even open source LLMs. New companies are releasing them so fast we can’t keep up.
First we had one use-case: text. Then two: text & images… Then four: text, images, robotics, voice…
We all know that AI is here and while we may disagree on how clever, legal, fair, intelligent or original it really is… what we must all agree on is that its distribution into our culture is growing exponentially.
We are in fact now at an inflection point (the point at which its growth tends to the vertical). We didn’t really get to that point with Covid because governments, scientists and our bodies intervened. I don’t see any sign that any politicians have the least inclination2 to intervene in AI’s distribution.
This is not about AI - it’s about scale.
Here are just a few examples - the first of which was only released publicly today.
The point here though is that not only is the availability of tools like this increasing exponentially … they are all API’s for other people to go out and build with … so the impact of these tools has a whole extra dimension of scale, as these tools are adopted and, inevitably, misappropriated by everyone and anyone with access.
The AI genie is never going back in the bottle. But as a culture we need to wake up quickly to the rate of change it represents.
It took months and months, in early 2020, to get a handle on the rate of increase of covid cases and do anything about it. Had we intervened in January - or even December 2019 - how different would things have been?
But as I say, no-one is going to intervene in AI’s development and distribution – so we need to figure out, pretty quickly, how we plan to live with its ubiquity … in everything from the way we manage our elections … to the way we trust the other person on the phone … to the way we define our own purpose in life, for ourselves and our children.
The examples are below, just after my obligatory plea for you to spread the word about my work. Have a great weekend! 🫶
BlandAI - Political robocalling API
This is a demo from BlandAI - a robocalling AI capable of sustaining 500,000 calls (per customer) with ~450ms response times… coming to the market just as the UK, US and a record-breaking number of other countries go to the polls. What could possibly go wrong?
HeyGen — Personal “Avatar” production platform.
HeyGen takes a few frames of video of you… to create a photorealistic, near-instant reproduction you can use to generate videos. Soon you’ll never have to get out of your pyjamas for an important Zoom call again.
Scenario – Gaming & “Influencer” mass production API.
Scenario is ostensibly a game-asset generator but it’s being increasingly used to create “influencer” avatars for use on social media … or for god knows what else.
This example, which has been doing the rounds recently as an fake influencer earning $10,000’s a month in revenue is based, itself, on genAI photos … trained from a subset of Midjourney images to create a replicable “model” … but there’s nothing to stop you using this technique to create a whole photo library with someone else’s real face.
I’m not going to get the release order right, I know. That’s not the point.
I appreciate the EU is doing things… and there are few legal cases in the US… but firstly these things are more about copyright protection and individual data privacy… and secondly the drive for global competitiveness will put paid to any individual efforts to contain AI.