(Edge) AI: From Science Fiction to Reality

Today, when it comes to artificial intelligence (AI), we inevitably think of generative AI and cloud services such as ChatGPT. But in the future, numerous AI algorithms will do their work directly at the data source: “Edge AI” will shape our everyday lives – and impose stricter requirements on system design and hardware components.

Only five years ago, a chatbot as powerful as ChatGPT seemed like science fiction, at least to people outside the AI industry. Today, however, AI in all its variations is finding its way into countless processes and products – be it generative AI such as ChatGPT or Midjourney, machine learning (ML) for highly scalable data analysis, or AI-based image recognition (computer vision). And suddenly science fiction turns into everyday life.

Going one step further and thinking about the future, I am wondering how AI will influence our lives five to six years from now. Which things that seem like science fiction to us today will become a reality? Also, as we are entering the vacation season, I couldn’t help but wonder in which ways AI will change our vacation – or so-called “workation” – experience.

“Good morning, a double espresso for breakfast, as usual?”, the coffee machine in the Airbnb apartment might ask. Will it recognize new guests by the sound of their footsteps after just one day thanks to a built-in AI chip and learn their coffee-drinking habits just as quickly?

When we dive into downtown traffic for sightseeing, will AI help us proceed more swiftly: will the autonomous streetcar stop automatically whenever a left-turning driver blocks the track. My guess is that there will still be no autonomous driving, as there still is too much hustle and bustle on city streets, but the cars will keep a good distance from the vehicle in front and automatically avoid bicyclists. But will traffic lights dynamically calculate the optimum switchover time to ensure that traffic flows as smoothly as possible? Will they autonomously synchronize?

Drones with surveillance cameras will probably keep an eye on beaches – as has been the case in Barcelona since 2022 – and report crowd density to the control center. But will apps guide us to the least crowded beach wherever we travel?

Will shopping lists pop up on our smartphones, AI-generated by the bot from the local supermarket we visited the day before? And when we collect our orders later, will the supermarket pick-up station recognize us by facial scan and automatically debit our purchases from our debit or credit cards?

Will the Airbnb be set automatically to 21°C because the smart home app knows your schedule and your preferred room temperature? After another brief chat with the coffee machine, will we sit at the desk and let the voice bot summarize emails, chats, and voice messages we've missed, sorted by relevance?

I have no doubt that AI will make our lives more convenient. Helping us to make the most out of our vacations, but also speed up our work, allowing us to be more efficient and focus on what really matters.

New Advantages come with New Challenges

All the possibilities I just described have one thing in common: they will be AI-assisted, but they won’t necessarily rely on data and data analysis in the cloud. While some aspects such as model training, trend analysis, and archiving might be still cloud-based, others will use edge computers embedded in the coffee machine, streetcar, traffic light, surveillance camera, etc.

There are many reasons for that. One cause is latency: an autonomous streetcar simply cannot first check the cloud to determine whether the shadow ahead might be a pedestrian. There is also the issue of data sovereignty: manufacturing companies, to name one example, are reluctant to hand confidential data to cloud providers. In other cases, it simply is more cost-effective to carry out AI tasks directly on-site instead of moving them to the cloud – if an internet connection is available at all.

Edge AI scenarios require devices whose entire hardware – from the processor to the storage modules – is designed to meet the challenges of the network edge. The systems must be as compact as possible, e.g. to fit into a coffee machine or traffic light. They must be able to withstand both very high and very low temperatures. Low power consumption is important to avoid shutdown due to overheating. And deployments in cars or streetcars require components that can deal with vibration and unfavorable weather conditions.

Another important factor: in accordance with the GDPR and AI Act, the storage components must guarantee the secure, encrypted storage of operational data, which might include personally identifiable information. Access must be restricted to authorized persons or applications. This makes secure storage a key building block for edge AI.

Finally, the compute and storage components must be designed for maximum reliability. Otherwise, we would risk a standstill in the (remote) office, in city traffic, in the supermarket, or, in a worst-case scenario, even of the coffee machine.

Ensuring Security and Reliability in Edge AI

Secure storage components, be they off-the-shelf or individually tailored to manufacturer specifications, will allow us to optimize countless aspects of our lives in our edge AI-assisted future. It will be critically important that these components meet the highest security standards, guaranteeing the enforcement of strict information security and privacy regulations. At the same time, they must be extremely reliable, as they provide the foundation for uninterrupted, resilient AI usage in all kinds of edge scenarios. If these criteria are met, secure storage components will help to make sure that these scenarios don’t remain science fiction.

This article was previously released on Matthias’ LinkedIn Channel.