Watercooler
March 31, 2023
10
min read

How AI Will Eat The World 🤖

Adrien Book

Of generative-AI-for-everything and synthetic pleasures

Remember the web3 hype? Tech bros with easy access to cheap liquidity wanted to create a decentralised, peer-to-peer internet powered by blockchain technology. Spoiler alert, it did not work. And it will not work. It simply cannot scale due to technical complexities ruining “decentralization”, while scams and thefts gave crypto and the entire web3 concept a bad name.

Artificial Intelligence experts, meanwhile, have been steadily releasing real tools and solutions, honing their science. In fact, they’ve been doing so since the 70s. This steady drip of innovation has come to a momentary culmination with the release of generative AIs (ChatGPT, DALL-E, MidJourney, etc). As is the mark of truly transformative tech, not only have these solutions proven to be immediately useful, they are also easy to understand and use by all — to the displeasure of many.


Just like the iPhone cemented the change from Web1 to Web2, Artificial Intelligence is creating what will undoubtedly be known as the “real” Web3: a new internet paradigm. Here’s how.

We are moving towards “Generative-AI-for-Everything”…

Today, we tend to talk about generative AIs like ChatGPT as a way to reinvent or even replace the traditional internet search engine. Indeed, why would one need to type a query, then click on multiple links — and avoid adverts — when a chatbot can easily provide the most statistically relevant information? LLM users can even adjust prompts as they go — a “co-creation process” between user and bot — which is impossible in browsers as we know them today.

But the threat to the established digital world is much broader than replacing Search. As generative AIs progress, they are learning to code websites from drawings, create images based on text, book holidays, understand verbal queues, etc. There’s been plenty of such examples over the past few weeks, with more to come. Even more impressive, many generative AIs remember past interactions — though that still requires some work.

Multiple players are naturally vying for dominance in the AI space. Amazon, Google, Apple, Microsoft, Meta… they’re all at it. They are creating tools that are right now unrelated and un-interoperable: you can’t use ChatGPT with MidJourney, and Bing’s attempts are not yet convincing.

This is unlikely to last long given the efforts and resources at play. Not only will a few actors manage to master Images, Speech, Text, and Video, but digital flywheels/network effects required to operate successful AI models at scale will ensure that only one or two players emerge as winners (see — Microsoft taking a stake in OpenAI). As Instagram founder Kevin Systrom recently said, “There’s a very real possibility that using artificial intelligence — some — or generally one of these companies will walk away with the everything app.”

Interconnecting services come naturally to companies that aim to create a moat and increase customer stickiness. Mentioning the iPhone was no random choice: what gives Apple much of its value is the way its hardware inter-connects, creating a seamless experience for the user. Today, your iPhone links to your Apple Watch and AirPods… and tomorrow, your vocal assistant (à l’Alexa) will link to your work assistant (à la CoPilot), your browser (à la ChatGPT), and your creative image-generating outlet (à la Midjourney). In fact, it will hopefully be one and the same. Simultaneously, AI memory is fast improving as increased resources are poured in cloud computing.

This will enable one lucky company to create a multi-modal personal assistant. Let’s call it — him? — Jarvis.

Picture Jarvis. It’s been trained on the entirety of the internet-based human experience. Using him is literally the equivalent of humanity hive-experiencing itself on an emotional level. But there’s more; Jarvis will also be trained on your specific data. How you write based on text, how you speak based on calls, where you’ve been, and with whom based on photos. When you shop for a toaster, Jarvis won’t find the best toaster for the average person. It will find the best toaster for you.

It will remember you across your channels (phone, watch, home speaker, laptop, TV, mobile…) and will be your one-stop shop any time you have a question or a request. We will no longer need browsers. We will no longer need apps. “Jarvis” will become the one and only Operating System for the internet, which will become one personalized web portal.

… and it will change everything

In the beginning, we had millions of websites. They were created by professionals. Operated by professionals. Populated by professionals. That was Web1. Maybe we should not have gone further.

We then federated, aggregated, and platformised the web. Facebook, TikTok, Amazon, Medium, and Youtube allowed us to consume user-generated content (web2). So much content, in fact, that it had to be optimised by AI to cater to our preferences (Web2.5?). This gave us the content needed to train large AI models.

When algorithmically-catered content is no longer unique or specific enough, we will increasingly interact with AI-generated content (Web2+). In fact, it’s already begun with the likes of AI-generated streams, AI-generated comics, and AI-generated pornography. Those are like… the Internet’s Top 3 Things!

But we won’t need existing platforms as “Jarvis” evolves. “Generative-AI-for-Everything” needs but ONE point of entry. ONE platform where “Jarvis” aggregates the aggregators to give unique and tailored content, answers, and recommendations. This is why historical Internet actors are freaking out. They will either become the Basilisk or be punished by it. They see no other way out. They cannot slow down, because slowing down means dying a slow, painful capitalistic death. As goes the local library, so goes the tech billionaire, right?

What then? We essentially have a god at our fingertips who expertly knows how to activate our dopamine centers — a science perfected by Facebook and TikTok — so that we continue to use it and gradually remove ourselves from day-to-day life. And then come the mattresses and the sunglasses adverts? That would be the expected Pavlov… — sorry —  Capitalistic urge.

And that would be fine if we only had to care for the entertainment side of the world. But as we make great strides in Augmented Reality and Robotics, we’ll soon have to discuss “AI-generated Elderly Care”, “AI-generated Education”, “AI-generated Mental Health”, etc.

NB: I would have also added AI-generated Romance, but at this point, I’m fairly sure Tinder is just 4 chatbots asking each other what they like to do in their free times.

Except that we’re already discussing them. And not in the best way.

“AI-generated War” is also a worry. If you think that the US, the EU, Russia, and China will agree to have the same “Jarvis”, you have not been paying attention. What happens when US Jarvis predicts Russian Jarvis is about to conduct a cyber-attack? What happens When Hallucinations Become a Feature and not a bug?

There are hundreds of such questions to ask, and little to answer them all. Marc Andreessen wrote in 2011 that “Software Is Eating the World”. He was right then. But today, AI is eating software… and everything else. Artificial Intelligence needs to be FED to work properly. And it feeds on us.

A final word

This vision is — mildly — hyperbolic. The Artificial Intelligence vocabulary has always been a phantasmagorical entanglement of messianic dreams and apocalyptic visions. Elon Musk went as far as to say in 2014 that “with artificial intelligence we are summoning the demon”, later speaking of an AI that would "rise in status to become more like a god, something that can write its own bible and draw humans to worship it”.

Maybe we’re at a loss for words in front of something we struggle to understand. Maybe we understand exactly what’s about to happen. We are entering a period of incredible unknowns, and the ball is rolling. In the 2020s, the internet will feel more alive than ever… but will be a place where we are increasingly walled off from other human beings. In the 2030s, we may all together abandon the idea of speaking to each other online. Who knows what will happen next?

Forget Web3. AI will give us Web0. Or WebNo. Maybe it’s for the best.

Good luck out there.

How AI Will Eat The World 🤖

March 31, 2023
10
min read

Subscribe to DevDigest

Get a weekly, curated and easy to digest email with everything that matters in the developer world.

Learn more

From developers. For developers.