Today a new Large Language Model has seen the light of the day: Apertus (formely known as ‘the Swiss LLM’). It’s co-created by researchers at EPFL, ETH Zurich, and the Swiss National Supercomputing Centre (CSCS). Its fully open and anybody wanting to use it can have a go on Hugging Face and soon there will be API and chat access via Public AI. Let me share my thinking why Apertus, a large model...
The Command Line is finally back – thanks to AI
In 2008, I talked about how important it is for websites to work correctly in text browsers (Lynx at the time). And guess what? We’ve pretty much landed back there, thanks to AI! Act one The first time I stumbled across the rise of the command line was with the advent of AI agents. When doing their job, they try to work as efficiently as possible and therefore prefer easy to navigate...
Digitale Souveränität – was heisst das eigentlich?
Wer von uns will es nicht: Unabhängig sein! Doch was heisst das in der digitalen Welt genau? Die folgende Grafik von Matthias Stürmer bringt es auf den Punkt. Was digitale Souveränität nämlich nicht heisst ist: Es geht sowieso nicht ohne «fremde Hilfe» (wahrscheinlich ist einer der amerikanischen Hyperscaler gemeint). Worum es aber geht ist: Keine Abhängigkeit von einem Hersteller (Vendor Lock...
The age of single-use software
Large language models excel at language tasks that benefit from statistical patterns. So from at the very beginning of GenAI, generating program code was an obvious application. As in many other fields, people stated that there was much more to programming than just writing code. Yes, and then came AI agents. The speed of development of algorithms to generate code and managing all documentation...
How to scale AI by skipping human effort
The same wisdom again, but this time illustrated by (and mitigated with) AI: Humans don’t scale. With Seed-Coder, the people from ByteDance just released a family of LLMs focused on the generation of code. Aside an excellent performance and an open-source approach, they highlight their curation process of data used for pretraining. In short, LLMs played a central role in filtering and curating...
Web Accessibility nach 20 Jahren
Im Rahmen des UX Camps habe ich meine Dateiablage mit 333 Dokumenten zu Web Accessibility (passender: Universal Access) durchforstet. Ergebnis davon sind die folgenden Aussagen und die Ernüchterung, dass ich dieselben Slides wie in 2002 heute noch nutzen kann. Einiges hat sich verbessert, das Allermeiste aber nicht. Von historischen Interesse ist die im Jahr 2004 zur Einführung des Bundesgesetz...
Guardrails: I’m afraid I can’t do that
With «A practical guide to building agents» OpenAI provides us with an interesting weekend lecture. Let me focus on just one aspect: Guardrails. Their paper shows a fairly simple architecture of agents working together, but their approach to controlling risk is quite sophisticated. While they propose a single agent that validates output (preventing output that could damage the integrity of your...
Swiss AI sovereignty – market opportunity or really complicated?
The AI systems we are about to deploy are not only produced in the US or China, but in most cases also using a cloud scenario where the second endpoint is neither in Switzerland nor under our control. Since hosting models locally with tools like Ollama, GPT4All or OpenLLM locally is an quite an easy fix, let’s talk about the elephant in the room: Building and tuning models under your control. To...
The end of SEO and SEA as we know
Most companies have their playbook with a successful mix of owned, earned and paid media. You can safely dispose of it, because – as the traffic tells us – your prospects are changing the way they search. People start their online journey using an AI tool Google is pushing AI answers The new kids on the block We all know the beauty of generative AI and it’s ability to answer questions in plain...
Der Online-Shop sieht keinen Kunden mehr: AI-Shopping
An einer Tagung des Handelsverband spreche ich zu: AI Shopping – the Age of Agents. Online-Shops sehen keine Kunden mehr, können diese nicht beraten und steuern keinen Check-out. Die harmlose Version davon ist die Fähigkeit von KI-Agenten — beispielsweise OpenAI Operator — in meinem Auftrag einzukaufen. Knackiger ist aber Perplexity’s «Buy with Pro», der nicht mal ein Agent...