Everyone is talking about Artificial Intelligence. Many are also doing something with it. Not just the likes of Klarna that claim to have replaced 700 human agents by ChatGPT. The AI boom also reaches our pets (monitoring of health), our children (education and mentoring) and even insects (pest control).
This technology wave is not unique
It seems like AI is overwhelming us like no other technology has before. While subjectively, this might be true, objectively also the AI trend follows clear patterns. To understand what happens to our economy and society, one must understand the core advantage of almost all foundational technologies: They reduce marginal costs of a good towards zero.
One impressive example of such a shift was the move away from classic data centers to the cloud. If you required an additional unit of computing in the pre-cloud era, you had to start a cumbersome process with your IT infrastructure provider. You had to specify your needs, order an individual set of hardware, wait until it is installed, and then pay a fixed monthly fee for the upcoming 3 years.
Now with the cloud, you can order an additional unit of computing for less than a penny. To be exact: You get 1 Gigabyte-Second (GBs) of computing power on Azure Serverless for US$ 0,000016. If you only need one additional GBs, this is what you pay.
The mechanism of marginal costs approaching zero can be observed manifold in the past: What now is a free email was a horse ride of the pony express before.What now is switching on the electric city lights with a button, were hords of lamplightersin the era of gas. still todayfor reasons of nostalgia.
What is this pattern applied to artificial intelligence? OpenAI’s ChatGPT, Google’s Vertex, and Meta´s Llama are hugely expensive in development and training, yet they bring down the marginal costs of human-like cognition. How much would a classic brainstorming cost with your human team? How much would you pay for a human translating a document into Korean? How much would you pay a graphic designer for a couple of drawings?
The principle is obvious: Without AI you would need to bring your team together, you organize the calendar invite, have everyone briefed before, hire a moderator, have everyone relaxed with fancy opening questions, and then go into brainstorming mode. Eventually, someone wraps everything up in a powerpoint. With AI? You write a prompt.
Shifting cost paradigms shift economies
Now you might say: Fine, I can do brainstormings cheaper with AI now. Hallucinations even seem to be a feature here. But what’s the big thing with brainstormings? Does big tech really invest billions into something that is good in brainstorming? Here again, our cloud analogy is of good help: in 2009, our savvy IT admins said: What`s the big thing that I can rent a minute of computing power by itself? It would actually be super-stupid to do so, because if you eventually utilize your cloud resources fully for a whole year, the cloud will be way more costly compared to your own IT infrastructure.
Well, what the IT admin did not see: Flexibly renting cloud-based IT-services allowed for totally new business models. Today´s world of free (advertising-based) or freemium online services like Instagram, Google, Miro, Zoom, Dropbox, Canva, Slack and Spotify would simply not be imaginable, if their providers had to go through a month-long ordering process for a linux server, some storage, and a firewall for every new customer.
Hence, cloud as a foundational technology reduced marginal costs of computing and lead to a big shift in our economy. Information technology became fast, accessible, and scalable and sparked a whole new era of innovation. Apple, Google, Microsoft, Amazon and Facebook became the most valuable, most profitable companies, leaving the good old oil industry behind.
Back to our analogy: Artificial Intelligence reduces the marginal costs of human-like cognition. Thinking, painting, talking, understanding, making concepts, all of this will become cheaper, accessible and technically scalable. We have reasons to expect this will shift paradigms of our economy, again.
Artificial intelligence replaces some human jobs
The first wave of innovation was easy to observe. Large language models (LLM) are good with language. Hence, Klarna replaced some human agents by bots. Also text-to-image models like Midjourney took over tasks done by photographers and artists.Consultatants,who do lots of things with words, can increasingly be replaced by LLMs: Analyzing texts, writing concepts, and spreading knowledge is nothing the AI can´t do. My condolences.
price elasticity function. Price elasticity function reflects the inherent inverse relationship."
I`d prefer to say it more bluntly: When prices go down, people buy more stuff.
The second wave of innovations isn’t obvious
What happens when goods are widely available and cheap, can be observed when looking how postal services developed over the last 30 years. The marginal costs of an analogue letter have never been on the level of a windows server. Still, 20-80ct plus the cumbersome act of printing, stamping, packaging and walking to the post office let you think twice, if you really wanted to write one.
With the advent of the email, not only the postal fee of sending the letter decreased to zero, also the process of packaging and sending it was reduced to a click of a button. The next chart showed the first wave of innovation: Many classic letters now turned into an Email.
But then, the second wave jumped accompagnied by the rise of mobile phones: Now as messages were free, every digital thing started to use them. Messaging services like WhatsApp popped up. Humans invented group messaging, software applications got connected via message queues. Then, social networks like Twitter and Facebook totally reframed communication leading to totally new markets. The new way of free messages even changed our language (LOL, 😀🤩😪, TLDR), started revolutions , and transformed our society. Can anyone imagine this in a world, where each of these messages would require 80ct for a stamp, and a cumbersome way to the letter box?
What happens when the marginal cost of thinking is zero?
Returning from the analogy to artificial intelligence: This new technology is not only way cheaper in perceiving, hearing, thinking, talking and writing than humans already, its marginal costs are even shrinking further. The follow-up question now is: Which kind of innovations will humans come up with tomorrow?
Will we now create messaging apps for AI models? Most probably not, but there will be many other use cases. Most of these applications will not replace human jobs, but will do stuff, humans never would have done before. Examples? AI is already particularly good in writing unit tests and documentation (in software development). This is a task that human developers of the past could easily have done, but seldomly did. With AI now, we likelihood rises, that the software we use will actually be tested and documented.
Source: https://www.deeplearning.ai/the-batch/falling-llm-token-prices-and-what-they-mean-for-ai-companies/
Another, more accessible example of the effect that people buy more stuff if its cheaper, is the youtuber Marius Quast. The creator of ski and biking videos now speaks 7 languages fluently in his videos. The quality of the underlying AI service(Dubly.AI)is incredible. Without Artificial Intelligence, Marius simply wouldn´t have done this.
Widely available new foundational technologies combined with human creativity are super powerful. What will mankind now do with ever cheaper, ever better artificial intelligence? I think the following 3 videos are giving a glimpse into our future:
Holoportation: Online-Meeting on Steroids
New television-like devices will render realistic representations to other places, requiring lots of additional AI infrastructure.
Metaverse: Fully 3D blended reality
New augmented reality devices like glasses or contact lenses will blend virtually anything with real life – requiring lots of additional AI infrastructure.
Personal Robots: Bringing AI to the edge
Physical robots will enter the market, building and requiring a kind of edge AI infrastructure.
More chips, more data centers, and more energy
To sum it up: Artificial intelligence is a foundational technology lowering the marginal costs of human-like cognition. In a first wave, this will replace some human jobs, but in a second wave, it will lead to an unimaginable wave of innovation.
What is imaginable though, are the consequences of this growth for the underlying infrastructure for Artificial Intelligence. The world in general, but also Europe and Germany, simply will require exponentially more of it. More efficiency, more chips, more data centers, more energy, better networks. In short, more of everything related to AI infrastructure.
Looking abroad, we can observe where Europe might need to invest into its future:
- Microsoft considers restarting the nuclear powerplant on Three Mile Island.
- Also AWS aims to invest into new forms of energy – in this casesmall modular reactors..
- The near east might develop into a hotspot of AI infrastructure, due to its cheap access to energy.
- NVIDIA has identified data networks as additional bottlenecks , especially in the training phase.
Together with the KI-Park, cloud ahead has started a series of blog posts validating the hypothesis that “Germany is not prepared for the AI boom”. We will dive deeper into the demand for AI infrastructure, discuss the challenges of chips, data centers and energy supply, understand physical and self-given constraints, and evaluate potential solutions for Europe and Germany.
So, stay tuned on this channel!
