Andrea Endara: "Streaming algorithms now function as invisible geopolitics editors. Youth learn more international relations from TikTok than from school." https://www.youtube.com/watch?v=sfcSRRQG6Mk

State bans are ultima ratio, not ideal policy. But Australia's data is striking: kids upgrade to real social contact instead of seeking workarounds. Even imperfect age verification shifts behavior massively. The real problem: platforms that consistently do too little.

The World Wide Web has reached a critical threshold. First, the open, decentralized web shifted toward platform-controlled silos, now increasingly mined to feed AI models. What remains of the web in ten years? Perhaps only private communities sharing human-only content?

YouTube is crafted video production. TikTok is everyone's spontaneous video diary. Two worlds apart—in content, creativity, misinformation. Yet fully surrendering to TikTok's entertainment-first algorithm redefines what it means to be a digital couch potato: quietly, willingly machine-controlled.

At least one good effect of this scandal. Ppl get more privacy aware & drop bigtech solutions. "SignalGate, as it's come to be called, may be the biggest scandal to hit the Trump administration in its first months in power. But it's been great for Signal." cxd.li/pBrgcA

Another way to view the crisis of the West is through Life Long Learning and its investment, which are perceived quite differently in the Global South.

Working with KI chatbots meaningfully is the hardest challenge, especially when moving beyond simple queries toward true Human Enhancement. Current AI models fall short in enabling this deeper collaboration.

Why is user adoption of AI chatbots actually not that breathtaking? Tool fatigue & preference for search, disappointment with mediocre results & privacy concerns, lack of clear use cases beyond hype, and the need to learn new skills for experimentation and effective prompting. cxd.li/4M3eA9

Why is it so challenging for companies to make use of AI chatbots? First, it reverses traditional IT practices - from specialized tools with clear rules to one tool serving all purposes without rules. Second, most organizations are too risk-averse for the experimentation LLMs require. cxd.li/4M3eA9

Best way to cut through the AI hype? Ask what hasn't happened yet. Despite all the buzz, we still see almost no fine-tuned language models providing quality information about niche topics or local knowledge. Zero such services exist. Just an obsession with big models.

It is crucial to understand the vast differences in AI energy usage when using AI chatbots. For text generation tasks alone, energy consumption can differ easily up to 50 times between models. That's why smaller language models are becoming increasingly important. cxd.li/D2SO4C

I love the new AI Energy Efficiency Score dashboard by Huggingface, because it rationalizes the AI energy consumption debate. It matters which Large Language Models you use - their energy consumption varies significantly. huggingface.co/spaces/AIEn...

Browsing through my reading list of strategy documents, I can skip many after recent developments. Digital development & foreign aid have changed dramatically with USAID's decline & UK FCDO's massive cuts.

Small steps indicate profound change. Only an administrative request was enough to block Ukraine’s access to Maxar satellite data—crucial for military & open-source intelligence communities alike. cxd.li/N3510y

Despite the hype, AI adoption is still lagging. ChatGPT's 400 million users in 2 years sounds less impressive when you consider Google Drive had 3 billion users in 2021. What's holding AI back? My analysis here: cxd.li/4M3eA9

Big Data is increasingly replacing official statistics. Combining these two worlds could be valuable - if platforms would share data, which they mostly don't. But that's not the only challenge: Big Data can often be non-representative compared to carefully gathered official statistics. cxd.li/9kA3W6

Can blending AI and human input amplify creativity? By combining multiple AI brainstorming sessions, a single person can gather a trove of ideas. Yet with current large language models, it's still hard work to achieve genuine results. cxd.li/k8oGDs

Recent events have made it crystal clear: Europe needs its own digital public infrastructure, independent from US tech giants. We must reclaim control of our digital public sphere, e-government systems, and critical platforms. The time for digital European sovereignty is now.

Hiring is becoming an AI vs AI game: companies deploy automated interviews while job seekers develop AI tools to beat them. We're drifting toward a future where machines unknowingly talk to other machines in recruiting. Check out this hilarious technical test cheating tool! www.interviewcoder.co/

It is hilarious! The Stargate PR stunt of supposedly investing $500 billion in AI by Altman and Trump was crushed by Chinese Deepseek's shock revelation that good models can cost just a fraction of that cost.

Funny, for many years, most of my time programming was spent debugging code. Now, I have to deal with a black box machine and find ways to see if the code makes sense. Well-functioning code does not mean it is free of various flaws.

Improvisation skills are crucial in complex environments—something Western countries often lack. Take a German street crossing with broken traffic lights. Chaos often ensues as Germans struggle to improvise. Improvisation: A skill increasingly vital for the future.

The Dual Role of Robots in Enhancing elderly Care in Japan: Ethnographic observations evidence finds in cases even more work for care workers through robots. A statistical analysis tells a different tale. Why am I not surprised? cxd.li/QKwJwX

Will research papers become obsolete? According to The Economist, 10% of all research paper abstracts are now created with AI assistance. Imagine if results could be rewritten for every audience, from experts to kids. cxd.li/Yo4XbP

The 50 Mbps threshold is sufficient: The push for faster internet speeds is largely unnecessary for private households and primarily driven by telecom marketing. 50 Mbps easily covers most family needs. cxd.li/ydWvrV

I wonder about the ratio of people writing AI policy papers versus those actually building meaningful AI applications. I bet the numbers are surprisingly close.

Hearing about China's massive Co2 consumption these days, right? But let's not skip the history lesson. This revealing chart says it all. The USA has consumed twice as much as China, and the EU has so far used seven times more than India since industrialization began. cxd.li/kP9iBj

Large Language Models' impact on Artificial Intelligence reminds me of what the World Wide Web did to the Internet: Both gave access to groundbreaking technology to the masses, leading to vast experimentation with an open-ended future.

after years of avoiding it, i had to use a windows pc for a project. noticed a new skill i've developed: multitasking while i wait for it to respond! 🐢 windows: proving that you can still be the biggest company in the world and still producing crap software.

AI's ability to prioritize information effectively is questionable. Language models lack insight into our personal knowledge bases and fail to summarize what's most relevant. Yet, they can help analyze large text volumes for patterns. cxd.li/Rk70Dl

when you think things can't get any more "efficient", faster, or shorter: from fast fashion to ultra-fast fashion, chinese platforms like pinduoduo are now directly "connecting" consumers with factories through their platform, temu. cxd.li/oh74rj

Was LinkedIn ever truly about knowledge sharing? It has evolved into a showcase of ever-changing job titles and achievements. Nowadays, people seem to talk less about what they know and more about grabbing inflationary and ever-expanding pseudo roles.

Robots in caregiving: a boon or a burden? Jame Wright research "Are Robots the Solution to Japan's Care Crisis?" unveils unexpected challenges faced by care staff when integrating robots. cxd.li/Fvk5oO

the 'pig butchering' cybercrime scam epitomizes this shift, merging emotional manipulation with high-tech deceit in a multi-layered exploitation unlike the simpler scams of the old days. it reveals a disturbing reality: victims are found at both ends. cxd.li/wthtbk

the introduction of chatgpt has led to shifts in freelancing platforms, affecting job demand & earnings according to two studies. the effect was swift & unsurprisingly led to a decrease. the more competent freelancers were supposedly more affected. cxd.li/tnecra

staggering number: china consumed more cement between 2011 and 2013 than the u.s. did in the entire 20th century. source le monde diplomatique german edition (march).

With big tech & aid sectors increasingly cooperating, I ponder: How does big tech's monopoly on cloud services restrict solutions and innovation in humanitarian tech? My thoughts on the Access Now report: Mapping Humanitarian Tech. cxd.li/TK0qg3

Prophet Song by Paul Lynch is a great read! It paints a chilling picture of what it would look like if a country like Ireland were to decay into an authoritarian regime.

The impact of technology on our carbon emissions is undeniable, from CPU-draining websites to energy-hungry data centers. I wrote an intro about the different areas of green ICT and a resource list to help get started with green IT and green software. cxd.li/ficZku.

The UK Government Digital Service discovered that 70% of users found chatbots useful. However, is usefulness enough when reliability is at stake? It's certain that chatbots will outperform most search engines, particularly government ones. cxd.li/9m0biS

South Korea's 56-hour digital government outage left 240,000 citizens in limbo, highlighting a crucial fact: Even the world's third-best e-government can falter. The digital government lesson: Perfection is a myth, anticipation of failure is key. cxd.li/oNCsQo

Hurray! Completed my first digital service job from home in Germany with an electronic ID after 8 attempts and 26 swipes over the phone, taking 45 minutes. A classic case of a rule-bound German Digital Government overlooking user needs and design best practices.

generative ai for feedback processing is like using a sledgehammer for surgery. reading about its use in employee surveys & for citizen participation, i couldn't help but laugh. imagine ai prioritize human feedback. what could possibly go wrong? cxd.li/lghzjn

Data leaks & cybersecurity breaches are becoming all too common. One reason: The reputation effect on compromised companies remains low. Stay vigilant with alerts from HaveIBeenPwned cxd.li/PKfVyp

When AI Took Over the ILO: A Hilarious Tale of Predicting Job Impacts The ILO has recently conducted a study examining the potential impact of generative AI on the job market. Instead of relying solely on their own research, they turned to chatGPT for its prediction of automation. bit.ly/3tVWXRD

Positive news from Germany against hate speech. German court mandates Meta to seek and remove such content, shifting duty from users to company. Turning the principles of the Communications Decency Act of 1996 (USA) that brought us into this mess. cxd.li/fd8FMc

Ventured into using two automated delivery services, only to find algorithms faltering at real-world exceptions due to rigid machine protocols. A testament to the limitations of automation in adapting to life's unpredictabilities and challenges when humans are out of the loop. bit.ly/3Hw6eCX

exploring the evolving world of digital-ready legislation. it's a fascinating mix of ambition and limitation, from denmark to the eu. making laws digital-friendly is one thing, but these principles barely focus on citizens, their impact, and feedback loops www.crisscrossed.net/en/blog/prin...

spent a day unraveling the complex web of eco-tech terms. green tech, clean tech, green it - they're not all the same. my takeaway? it's a mix of hope and hard truths about our environmental impact. check out my latest blog post www.crisscrossed.net/en/blog/deco...