Marie Kilg
I’m a journalist, speaker, consultant and former Chief AI Officer.
My experience includes product management for Amazon Alexa, creating and consulting automation projects and AI strategies for all sorts of companies and media institutions, hosting Germany’s most popular AI-podcast and launching the first ever non-human columnist in a German print newspaper.
Here’s a list of some more things I’m passionate about or interested in: intersectional feminism, trans rights, worker’s rights, time travel, kung fu, cabbage, sci fi, books, dogs, hip hop, swimming pools, eggs, potatoes, butter, whisky, rum, Zelda BOTW, TOTK, LGBTQ+, XR, lemons, sleep
Resume
Work experience
Consultant, Trainer and Speaker
2017 – present
Among others: LMU Munich, Tech Show Frankfurt, re:publica, University of Bamberg, German School of Journalism (DJS), University of Applied Sciences Munich, Sueddeutsche Zeitung Publishing, Zurich University of the Arts, Mediaschool Bayern, ifp Institute Munich, Netzwerk Recherche, Zündfunk Netzkongress, Media Innovation Camp Munich
Read more here: Teacher & Speaker
Deutsche Welle, Berlin, Germany
2022 – 2026
I served as technical lead at DW Lab and later as Chief AI Officer, advising on tech strategy up to Director General level and building the AI team and governance from the ground up.
Amazon Alexa, Munich and Seattle
2017 – 2022
I joined Amazon as an editor and subsequently worked in a few different roles in both Munich and Seattle.
As Content Program Manager, I was responsible for Alexa’s German personality and the overall content strategy in Germany and Austria.
In Seattle, I managed international improvements for Alexa Music in 14+ locales.
By 2020, I was Senior Product Manager responsible for the launch roadmap for all innovations in German-speaking Alexa, including Echo devices and software features.
I also helped launch Alexa in France and Brazil.
Reporter and Editor, print/radio/online
2014 – 2017
Munich and Berlin
Bavarian Public Broadcasting (Bayerischer Rundfunk), taz.die tageszeitung, Süddeutsche Zeitung, M94.5 and others
I authored and edited articles and radio programs in media – , science -, economics -, and investigative feature departments, and served as engagement manager for the podcast and Youtube series “Die Frage”.
I also conceptualized new formats and led innovative online projects, i.a. the experimental twitter bot @JudithBotler, nominated for the German Alternative Media Award
Education
Media Entrepreneurship Program
2017
Design Thinking fellowship at digital media incubation hub “Medialab Bayern”
German School of Journalism (Deutsche Journalistenschule)
2015 – 2017
15 month program for print, online, tv and radio journalism. Managing editor for final radio feature and final online project
University of Munich (Ludwig-Maximilians-Universität München)
2011 – 2015
Bachelor’s Degree, American Studies and Communication Studies
University of Costa Rica
2013
Exchange Semester, Journalism Studies & History
National Taiwan University
2012
Summer University, Mandarin and History & Culture of Taiwan
Skills, Qualifications, Awards
Product Management
Operational efficiency, Automation, International cooperation, Design Thinking, Quality Assurance, Software Launch Management, Device Launch Management, Consensus Building, Coaching and Mentoring
Journalism
Content Strategy, Digital Storytelling, Data Analysis, Public Speaking, University Teaching, Social Media, Community Management, Audio Editing, Radio Scripting, Mobile Reporting, Writing & Editing (print/radio/online)
Tech
Generative Media, Machine Learning Application in Journalism, Generative AI, Voice Assistant Technology, Natural Language Understanding, Automatic Speech Recognition, Chatbots, Twitter Bots, R, VBA, Microsoft Office, Atlassian Jira, Confluence, SIM Simple Issue Manager
Languages
German, English, French, Spanish
Honors & awards
Datenschutzmedienpreis 2025 – Hauptpreis
Stuttgarter Moderationspreis 2025 – Nominierung Kategorie Journalistische Qualität (Presentation Award – Shortlisted in the category “Journalistic Quality”)
Grimme Online Award 2024 – Shortlisted
Journalist of the Year 2023 – Science Category #2
Deutscher Sozialpreis 2018 (German Social Award) – Category Online: “The Question: What’s it like to be trans*?”, YouTube Series
17 Best Stories of 2017 – Sueddeutsche Zeitung: “Injured, burned, lost”, data-driven investigation about injuries in the German national soccer league
Alternative Media Award 2017 (Alternativer Medienpreis)- Shortlist: @JudithBotler, a feminist twitter bot
Track record
What she said,
and when she said it.
A documented record of public calls — published in ZEIT ONLINE’s Natürlich intelligent newsletter and elsewhere — benchmarked against when the broader discourse caught up.
Before the hype
2023
AI is not a black box. The black-box claim is an excuse. +12 mo
When “AI can’t be controlled because even its makers don’t understand it” was conventional wisdom, Kilg argued the opposite: explainability is technically achievable, and those who claim otherwise want to avoid accountability.
What happened: Mechanistic interpretability became a mainstream research priority in 2024–25. The EU AI Act’s explainability requirements — once seen as technically naive — are now standard practice in enterprise deployments. The “we don’t understand it” defense has largely collapsed as a credible regulatory argument.
2023
AI agents: the self-evaluation problem will limit them. +6 mo
Six months before the mainstream agent hype peak, Kilg described autonomous agents and their structural flaw: they evaluate their own work with the same systems that produced it.
What happened: Self-evaluation became one of the defining research problems of 2024. OpenAI’s o1 and o3 reasoning models are a direct architectural response to this failure mode. Devin — the autonomous coding agent that shocked developers in March 2024 — was independently tested weeks later and solved roughly 14% of the tasks it claimed to complete.
2023
The data wall is real. Scaling will plateau. +9 mo
When Bill Gates floated a scaling plateau and was largely dismissed, Kilg grounded the argument in specifics: the supply of high-quality training text is finite, and models consume it faster than humans can produce it.
What happened: The “data wall” became a consensus concern by mid-2024, documented by Epoch AI. European and Chinese open-weight models closed the gap faster than expected — consistent with Kilg’s argument that the data advantage of large US labs was never infinite.
2023
Full automation is not coming. The physical economy sets the ceiling. +8 mo
At the peak of OpenAI-driven techno-utopianism, Kilg argued that exponential AI progress does not automatically transfer to the broader economy. The slowest-moving sector sets the limit.
What happened: Goldman Sachs published “Gen AI: Too much spend, too little benefit?” in July 2024. Sequoia’s “AI’s $600B Question” followed shortly after. Both made the same structural argument — eight months later, with far larger audiences.
2023
Google’s real problem is its business model, not the technology. +6 mo
When most coverage asked “will ChatGPT kill Google?”, Kilg reframed it: Google cannot cannibalize its own ad revenue aggressively enough to win. The threat is structural, not technical.
What happened: Google’s “AI Overviews” launch in May 2024 produced embarrassing errors — a direct consequence of the incentive misalignment Kilg described. Google has since retreated and continued to lose ground to AI-native search interfaces.
2024
Altman’s energy optimism is evasion, not strategy. +4 mo
At the height of Altman’s Davos moment, Kilg wrote that his fusion-will-save-us framing avoided honest accounting of AI’s current energy costs — costs effectively subsidized by the absence of carbon pricing.
What happened: IEA projections confirmed data centers would consume as much electricity as France by 2026. Microsoft’s carbon emissions rose 30% against its own climate targets. The fusion breakthrough has not arrived.
2023
AI music democratises expression — the art debate asks the wrong question. +4 mo
Before Suno and Udio became household names, Kilg reframed the AI-art debate: alphabetisation waves were never about producing more Shakespeares, but about giving ordinary people a new means of expression.
What happened: The RIAA’s 2024 lawsuits against Suno and Udio played out exactly as framed — not a philosophical debate about artistic authenticity, but a legal fight over who controls access to a new expressive medium.
Craft, not prediction
Germany’s first AI-authored newspaper column — before ChatGPT.
In mid-2022, when GPT-3 was known only in specialist circles, Kilg conceived and launched Anic T. Wae — the first non-human columnist in a German print newspaper — working with large language models before they were a public phenomenon.
Why it matters: This wasn’t a tech experiment. It was applied editorial judgment about what these models could and couldn’t do — six months before the rest of the world got access to ChatGPT. The learnings from that project directly inform her consulting practice.
“The best AI expert in Germany — because she loved AI back when AI was still pretty bad.”
— Gregor Schmalzried, co-host of ARD’s Der KI-Podcast
Work with Marie →