Why we need Local AI

The European Union has taken a historic step by moving to enact the world's first AI law. This legislation, from a regulatory standpoint, is as much a milestone as the evolution of the technology itself, aiming to regulate the development and use of artificial intelligence in a way that fosters innovation while protecting the fundamental rights of all citizens. The initial drafts of the document look promising. It differentiates between "high-risk AI systems" and systems with "limited risk." It's important to understand that it's not the AI models themselves that are deemed risky, but rather the context and the area in which they are to be deployed; for example, systems intended for use in law enforcement or in the management and operation of critical infrastructures are considered high-risk. Models capable of creating deepfakes or similar image and video manipulations, however, are not categorized in this high-risk group. This comes despite recent warnings in the World Risk Report 2024 about the dangers of disinformation, even placing it above extreme weather events due to climate change. "AI-generated fake news and cyberattacks are considered the top immediate global risk - especially with upcoming elections in major countries like the USA, the UK, and India.“ (1).

AI is making its way into more and more areas of our lives. While it's still a long way to the ultimate "Little Helper" made of flesh (silicon?) and blood (motor oil?), software-wise, we are getting closer to this reality. The idea of an artificial jack-of-all-trades for practical home use is ancient, long just the product of entertaining books, series, and science fiction movies. Those days are over, Skynet (wink, wink) is practically live, and the year 2029 is not so far off...

Jokes aside.

During our visit last week to the World Economic Forum in Davos, we did not get the impression that anyone is worried about the use of artificial intelligence - on the contrary - doors and floodgates of support (in the form of money but also a general acceptance of a new technology we yet have to fully understand) are willingly opened, hooray, finally innovation again, the motto being: it’s the game of the first-movers. Sam Altman is received like a rock star. What was considered the next big thing one or two years ago; at this year's edition of the WEF is only a side note: Crypto and blockchain have moved to the background, with everyone talking only about AI.

However, the combination of these two technologies can be very exciting. A brief impulse:

Fundamentally, digital, centrally organized systems always carry a certain inherent risk: first, comprehensive data security cannot be achieved (no system is secure), and second, they repeatedly raise the question of who actually owns our data, and what happens to it. Sure - even a blockchain-based, encrypted a hundred thousand times over, decentralized database does not offer one hundred percent protection against cyber attacks, but then again - such protection will never exist anywhere.

The same logically applies to AI systems. Here, another risk is added, that of a "biased AI"; i.e., AI that develops prejudices against, for example, ethnic minorities based on its training data. The procurement, management, and feeding of high-quality and balanced training data for AIs is a science in itself - and, therefore, shall not be the topic for this article.

Let's mentally return for a moment to our fantasy of the "Little Helper," our own AI that we can specifically train according to our requirements, desires, and needs. Wouldn't that be mega cool? "Don't forget to buy tissues," "remember your mother's birthday the day after tomorrow, a suitable gift for her could be a yellow phone case," "do you need help with your tax return?," "I found the product cheaper in another shop," and such things, are the kind of assistance that come to mind instantly. Surely, everyone has their own personal idea of how their digital assistant should function.

To make this a reality, we would need to train our private AI assistant with very personal and sensitive data. The system would need access to my location, my browser history, the metadata from my photo library, maybe even content from my chats. Would you just throw such data in large quantities onto some server without any real guarantee that no mischief will be made with it?

Well, jokes on us, I would say, after all, we are already doing it on a large scale by using smartphones, smartwatches, smart fridges, smart-whatever-else-there-might-be. We produce boatloads of data every day. This needs to change urgently, especially in dealing with AI, which can not only access, but automatically compare, and link, and cross-reference these data.

We need local AIs.

Local AI systems work directly on the user's end device, without the need to send data to external servers or cloud services. Furthermore, local AI systems can also function when no internet connection is available (not entirely unlikely in the digital desert of Germany), making them ideal for applications in remote and poorly connected areas. Parallel to this, blockchain technology has established itself as a revolutionary tool for creating a decentralized and transparent digital infrastructure. From finance and healthcare to education and art, blockchain provides a secure platform where transactions can be verified, recorded, and made transparent for all participants. Its ability to ensure integrity and trust in digital processes makes it a valuable tool in today's connected world. Blockchain technically enables us to truly own the data we produce and to decide autonomously on their use. I deliberately say „use“ because it should be everyone's right to monetize their data. Who knows, maybe one day the tables will turn and the big platform operators of today will pay us to provide some of our highly sought-after data for some advertising. Future music. Back to the topic.

By providing a decentralized and transparent platform for data storage and exchange, blockchain can significantly enhance the security and privacy of local AI systems. The combination of local AI with blockchain technology opens up new possibilities for securityand autonomy in the digital space. Local AI can provide fast, efficient, and personalized services directly on the user's device, while blockchain enables a trustworthy record of these processes. This synergy can be particularly revolutionary in areas such as the Internet of Things (IoT), digital identities, and secure data traffic. The last point is arguably the most important when it comes to sensitive, private data. To cultivate a personal assistant that really benefits you, it will be crucial to link the various devices that produce data (location data from the car, information from the smart fridge, browsing histories from laptop, smartphone, tablet, TV). Ultimately, blockchain can establish a secure sync between these various devices, thus realizing true "data ownership." And even in data traffic with third parties—whom I may not necessarily trust because I do not know them—blockchain ensures that only the data I have authorized is shared. Managing my entire digital identity becomes significantly easier.

As we move into 2024 and beyond, developments in these areas will undoubtedly continue to be at the forefront of technological innovation and discussion. EU legislation and similar regulatory efforts worldwide will play a key role in ensuring that these technologies are used for the benefit of society and in accordance with the fundamental rights of all. At the same time, we have the opportunity to go even further. I would argue AI and blockchain are a perfect match, Adam and Eve, lid on pot, made for each other. The AI revolution and the blockchain revolution should no longer be viewed separately. It's about time to combine the best of both worlds. One day, we will thank ourselves for it.

(1): Weltrisikobericht 2024: Größtes globales Risiko: Desinformation
(2): KI-Gesetz: erste Regulierung der künstlichen Intelligenz