At this year’s General Assembly of German Influencers (digital trade fair OMR with 70,000 visitors), Pit Philipp Klöckner gave the key speech on the topic par excellence: Artificial Intelligence. On 150 pages of PowerPoint, he formulated what the talk in Germany is on the topic.
What an excellent presentation. It is worth watching it from start to finish (in German only. Thanks @Ole Harms, for the recommendation.)
For my English audience, I wrote this newsletter in English.
Pit asked:
Does AI have a “moat”?
The concept comes from the analysis of companies. How strong their competitive advantage is. Gartner speaks of “Peak AI” and expects the hype to cool down. In this context, the implicit question is: is AI a hype or will AI remain?
The answer is trivial: Of course. It’s both.
We know that there are still 15 years to go before AI has really moved into all aspects of life. Why do we, like little children before Christmas, always have everything instantly? The following generations will also make breakthroughs, perhaps even better, safer, and more sustainable than the generations living today.
For adults, this question is more exciting:
How well is Germany positioned for AI? Will Germany’s competitiveness benefit? And thus the car industry?
This answer is much more exciting.
| Brain power: do we have enough developers? | ++/0 |
| Do we have training data? | – |
| Do we have enough chips? | – |
| Is there enough water to cool the chips? | 0 |
| Do we have enough energy? | — |
| Are we producing quality AI (hallucination)? | (+) |
| Will we remain independent? | – |
In short, under today’s conditions, Germany has no competitive advantage, even though we have the technical expertise. If everyone concentrates and creates the conditions together, then it can still be something.
With the competitive advantage.
With the automotive industry,
The next few years are decisive.
Germany has brain power
Pit cites a study (Stanford Developer Index) according to which Germany is No. 3 in the world for the quality of AI developers. I couldn’t verify this, but it seems plausible to me. Our training infrastructure continues to produce sufficient technically trained engineers.
However, the question is whether Germany will create attractive conditions for developers: Pawel Durow, CEO of Telegram, a messenger service with almost 1 billion users worldwide, initially tried to locate his company in Berlin. He had founded the “Russian Facebook” VK in Russia and had sold his company and left the country due to disagreements with the local authorities on the subject of data protection.
Germany can produce brain power. We have good universities and research institutions. When it comes to attracting and retaining top talent, Germany clearly has a structural disadvantage.
There is not enough data
As the number of Large Language Models (LLM) increases, a paradox arises at some point: the data generated by LLM exceeds the originally available data.
Conversely, this means that less and less “natural” data is available for training the LLM. One model invents facts, the other processes the other, to put it bluntly.
Germany has limitations in terms of data.
Because Europe has the strictest data protection regulations.
The logic is: the self-determination of each person must not be restricted. But most data is created as a result of human behavior, for example when we work, shop, pay, move or watch a video and surf the Internet in everyday life. All this data may only be used with the consent of the respective person.
That is the law. And that’s right.
In addition, there is copyright.
And the EU AI Act.
However, this limits the data available for training.
If we think of competitors such as the USA, China, India, Russia, etc., then we know that there are different cultural backgrounds and legal systems there. And much more data uncritical for the training and development of LLM
A clear, unchangeable disadvantage for Germany when training LLM.
Control over microchips
Germany is generally weak in the production of microchips.
This is not news. The German government is currently counteracting this with controversial subsidies for the construction of chip factories .
But it takes time to build a self-sufficient supply chain. The construction of a large chip factory can take 5 to 7 years. Not every chip produced can be used for AI applications.
In the medium term, this will not enable Germany or its European partners to produce microchips in sufficient quantities for the expected growth in the required computing power.
Europe has critical technologies in the global chip supply chain, but will also remain dependent on supplies, especially from the USA and Taiwan.
How great is the influence of German developers on chip design? Can Germany achieve performance advantages in artificial intelligence through specially designed training or interference chips? Or do we have to make do with what others have developed?
This appears to be a clear disadvantage for Germany, which is unlikely to change significantly in the next 10 to 15 years.
Not enough water
Water is needed to cool the data centers. No water, no data centers, no artificial intelligence.
Pit speaks of “water consumption”: Imagine a water meter hanging at the inflow to a data center that measures how much water is taken for cooling. And there would not be enough water to supply data centers.
However, the word “water consumption” as Pit uses it is misleading.
On the one hand, a coolant similar to that in the freezer or air conditioner cools the chips.
On the other hand, water is not consumed. The amount of water in the “Planet Earth” system is constant. Water cannot be “consumed” in the same way that we use gasoline or diesel when driving. Because it doesn’t react and it doesn’t disappear.
It can be dirty or in the wrong place.
But it cannot “disappear” from the planet.
Accordingly, “water consumption” indicates a transport or treatment problem. The problem in certain populated places can be that there is not enough water to supply the population with the simultaneous cooling requirements of a data center.
Enough water can be provided anywhere on earth if it is either transported there or processed by, e.g., reverse osmosis. Both require a lot of energy.
The simplified formula for water scarcity is: water = energy. For example, in Spain, due to the drought of the last 2 years, two seawater desalination plants are being built near Barcelona, powered by solar power plants. Ships will bring in water from the Ebro Delta, 100 km away, until these are operational.
In Germany, the operation of large plants with high water consumption (power plants, chemical plants, etc.) has not been a problem in recent decades, except for periods of drought. In addition, we have the technology, infrastructure, and operational experience for the industrial treatment and transport of water in Germany.
Compared to other regions of the world, Germany has a clear structural advantage here.
Although there is no such thing as “water consumption,”. So there is no competitive advantage here.
Energy
The demand for computing capacity will increase exponentially in the coming decades with the spread of artificial intelligence into all industries and areas of life. The demand for electrical energy is also increasing.
A ChatGPT prompt consumes about ten times as much energy as a Google search, and the trend is rising.
Pit states the growth of electricity demand through AI by 2030 at +10% worldwide. But it is much more and more complicated:
- The prediction for the US is that the energy demand for AI will increase FOURFOLD by 2030 (see chart). Looking at the world’s needs is not particularly meaningful.
- In addition, local concentrations of data centers are emerging in geographically favorable locations: a place that promises sufficient water and cooling can so far have little generation infrastructure.
- It can be assumed that data centers that have an electricity requirement of 1 GW will require new construction projects or will be located next to planned new buildings.
- Intermittent generation modes such as solar or wind are unsuitable for data centers. Load or frequency fluctuations lead to damage to data centers. Accordingly, data centers need their own, cost-effective energy supply.
- This requires base-load power plants, powered by hydropower, nuclear power, or cheap fossil fuels, which are then compensated for by CO2 certificates. Storage technologies are only sporadically available and make electricity even more expensive.
Both Sam Altman and Bill Gates are emerging as private investors in nuclear fusion startups and nuclear fusion startups , respectively. Small Modular Reactor (SMR). They seem to have recognized an impending global energy shortage, especially in the industrialized nations, as a problem.
Germany, on the other hand, with its sluggish energy transition and its many unanswered technical, administrative, legal and economic questions and high energy production costs, currently seems to be very poorly prepared for this additional demand for base load.
The topic of “Energy for AI” is by far the greatest need for action. If we want to play along with AI, we need cheap electricity in abundance. At the beginning of the last century, Germany was a leader in the production of energy infrastructure: AEG, KWU and Siemens were once the world’s suppliers of electricity, power plants and grids.
Today, the suppliers of the Energiewende in Germany, which supply solar cells and wind power plants, are from China.
Reliability
Pit states that the medical AI assistant medGemini achieves an accuracy of 91% in diagnosing patients. Doctors would achieve an accuracy of 89%.
That is not enough.
While doctors can never achieve more than 90% accuracy, machines in the medical field with less than 99.9% accuracy are unacceptable. Because humans have “common sense” and can make good decisions in a fraction of a second. And everyone makes mistakes. This is not a “bug,” but making mistakes makes us human. A machine that makes mistakes is faulty and must be taken out of service.
In any case, I would always give preference to the doctor when it comes to my health.
The real problem with today’s LLMs is that they do not yet specify a “truth vector”. In addition to the output from a prompt, there is no indication of how far the statement has “drifted”. This is because there are slight deviations along the causal chains (the vectors) that an AI processes: the logic path is correct, but the result is not quite correct.
I imagine it like a ship sailing across the ocean. The distance covered is entered on the map as a course and distance. This is called “pairing”. What I don’t take into account, for example, are wind drift or ocean currents that take the ship off the coupled course. With paddocks, I arrive across the ocean, but maybe not in the planned “real” port.
It works similarly with AI. Here, too, there is “drift”.
In AI, this leads to a “hallucination”.
Logical but incorrect results are output.
That’s why we need a measure of how likely a false, hallucinated statement of the AI is.
I would like to see a technology and a legal requirement for the accuracy of AI developed in Germany. And, of course, the problems of privacy and copyright protection will be solved.
This would make AI usable for many more areas of application, such as medicine, science, law, production or administration.
This could be a locational advantage for Germany: “Made in Germany”.
Because we can.
And the world would buy it.
Who else?
Independence
Meta had first changed its name to build the “metaverse”. This is obviously not yet technically possible. Now Meta has released one of the best LLMs and made it available for free (not yet available in Germany).
This puts a lot of pressure on the economic models of other competitors, such as ChatGPT, Google, etc. After all, who can still charge $20/month when the same service is free around the corner?
But even for German providers, such as ALEPH ALPHA, the competition is overwhelming. The supremacy of US tech companies is overwhelming. They are very integrated into the German and European markets. Competition supervision seems too weak in view of the overwhelming technical and economic superiority of the American technology companies.
In May 2024, it seems impossible for a German tech company to establish itself in the international AI industry without the EU competition authority emancipating itself, asserting itself and creating space for real competition.
What Germany should do
The situation for AI in Germany is not good, but it is not hopeless.
Our most important asset is the people, the educational infrastructure and our culture people. We can do things. We can also use AI.
But we need energy. Infinite, cheap energy. Unfortunately, energy from solar and wind is not enough. We have to acknowledge this fact and then react.
Nor can we build competitiveness in a more multipolar world with laws alone. The size and power of the EU is not enough to regulate the market.
Competition is strong and especially outside of Germany and the EU. AI is too important for us to become dependent on this technology in the long term.
