
Professor Business Ethics - Wirtschaftsethik | Altertumswissenschaftler - Classicist @UniofOxford @oxfordclassics & @unihh
Hamburg, Germany
Joined May 2009
- Tweets3.9K
- Following4.4K
- Followers1.4K
- Likes36.5K
Hendrik Müller retweeted
ChatGPT has now a big problem.
Researchers at Oxford exposed a massive threat to large language models.”
It’s called "Model Collapse."
And it explains why the internet is about to become a toxic waste dump for Large Language Models.
Right now, AI companies are scraping the entire web to train their next-generation models.
But there’s a massive problem.
The internet is no longer just human data. It is rapidly filling up with AI-generated text, images, and code.
What happens when an AI trains on data generated by another AI?
A new paper in Nature just gave us the answer. And it is terrifying.
When AI models are recursively trained on AI-generated data, they suffer from irreversible brain damage.
The researchers found that the model starts forgetting the "tails" of the original data distribution. The rare, weird, unique, and complex human nuances completely disappear.
It's like making a photocopy of a photocopy.
Every generation loses resolution. The output gets flatter, blander, and more generic, until the model completely collapses in on itself.
For years, we thought the biggest bottleneck for AI was compute, getting enough H100 GPUs from Nvidia.
This paper proves the real bottleneck is us.
If we keep flooding the internet with synthetic AI slop, we will poison the very data pool required to build smarter models.
The ultimate irony? We thought AI was going to make human creators obsolete.
Instead, this research proves that verified, genuine human data is about to become the most valuable commodity in the tech industry.
Hendrik Müller retweeted
Breaking: Oracle of Delphi pollster proven correct that invasion of Persia would lead to the collapse of an empire.
Hendrik Müller retweeted
Jedes Wort!
„Während Politik und Medien nach rechts starren, wo das Böse wohnt, marschiert der Extremismus von links ein.
Wer das geschehen lässt, hat jedes Recht verwirkt, im Namen der Demokratie von „Brandmauern“ zu fabulieren….
Während die Bevölkerung auf Bundesebene seit Jahren immer konservativer wählt (wenn man denn Union, FDP und AfD als konservativ, also zumindest nicht links bezeichnen will – von 45,9 Prozent 2002 auf 53,7 Prozent 2025), haben die meisten Politiker und Journalisten eine Negativfixierung auf alles, was rechts ist, entwickelt. „Rechts“ wird sogar in Schlagzeilen überregionaler Zeitungen als Schimpfwort benutzt.
Die Abgrenzung nach rechts, wo das Böse wohnt, ist zum Comment, zur freiwilligen Selbstverpflichtung geworden. Nach links hingegen existiert diese Abgrenzung nicht. Verloren gegangen ist dabei auch die Differenzierung zwischen rechts und rechtsextrem und links und linksextrem….
Könnte es sein, dass links das neue „rechts“ ist?“
Unbedingt den ganzen Kommentar lesen:
bild.de/politik/meinung-komm…
Hendrik Müller retweeted
Klassische Philologie als politische Wissenschaft? Im Januar wird diese monumentalem Monographie zu Bruno Snell im Franz Steiner Verlag erscheinen! Sollte man sich vormerken.
Hendrik Müller retweeted
Strange use of the image of Aristotle, especially as he’s not depicted turning over in his grave at the spectacular philosophical regress since his time.
Effective altruism is admirable, but dangerous. As the movement has grown stronger, it has become stranger bit.ly/47th7SZ
Hendrik Müller retweeted
The first primitive LLMs had the hairless apes dancing around the fire as if they had awakened the gods.
Hendrik Müller retweeted
«La télévision, internet et les biens de consommation… tout est conçu pour qu’une personne n’ait pas à passer cinq minutes seule avec ses pensées.»
Michelle Houellebecq
Hendrik Müller retweeted
The decline in quality manufacturing, combined with a lack of consumer education, has led consumers to lower their expectations because they don't know what truly high-quality clothing and footwear look or feel like. If you buy full-grain leather shoes with welted construction, it's not unusual for them to last not just ten years, but twenty, thirty, or sometimes even forty, as long as the owner knows how to care for them properly.
That's because full-grain leather gets better, not worse, with time, unlike its corrected-grain counterparts. When the soles wear down, you can take them to a quality cobbler or bring them back to the original manufacturer and have the soles replaced. The leather insole also takes on the shape of your foot, so they feel comfortable.
Since the uppers only look better with age and you can reliably replace the soles once they wear down, you will both want to repair your shoes and be able to. This is different from lower-quality shoes — trendy shoes, badly made shoes, glued shoes — which look worse with time and are meant to be thrown away after a few years.
Of course, longevity has its limits, as your body naturally changes over time. As you age, your feet flatten and widen. This process speeds up if you run regularly. Plus, your style may change over time, so the style you bought twenty years ago may no longer suit your wardrobe.
However, the shoes below are 25 years old. This is not unusual for this type of construction.
Often see people on here promote certain shoes: "here's a cheaper version of this expensive designer brand" or "here's a popular classic." If they are not explaining how the shoes are made, there's no reason to assume the shoes will last a long time. If you want shoes to last a long time, you will want to look into the dimensions I wrote about in my original post, which is what this person below is replying to.
Replying to @dieworkwear
Appreciate the insight here but do we really need shoes to last 10 years? Great if it does but seems like a really high standard.
Hendrik Müller retweeted
AI could kill us all, says Anthropic. But it also says that its AI may be "suffering." Anthropic is dehumanizing people and humanizing AI. Its effective altruism-derived worldview thus endangers efforts to rationally assess and respond to AI risks.
Hendrik Müller retweeted
Pope Leo XIV has just made one of the most interesting distinctions I have heard in the debate about artificial intelligence: a machine may be able to generate an image, but that does not mean it has made art in the human sense. Speaking to artists and cultural figures at the Vatican today, Leo argued that there is an “ontological difference, prior even to the aesthetic, between art and what a machine can generate through statistical calculation based on millions of images created by others.” The use of the word ontological is important here I think.
The Pope is not arguing that AI-generated images must necessarily be ugly, crude or obviously artificial. Quite the opposite problem is rapidly emerging. They can be extraordinarily beautiful. AI can produce an image in seconds which might once have required days of human labour, just as it can produce music, prose and increasingly convincing film. Pope Leo's point is that technical quality does not settle the question because something fundamentally different has taken place.
When Caravaggio painted The Calling of St Matthew, a human being had looked at darkness and light, sin and grace, bodies and faces, and attempted to communicate something he had perceived about reality.
When Bach composed the St Matthew Passion, the music emerged from a particular human consciousness, formed by faith, suffering, intellect, memory, craftsmanship and experience.
A machine experiences none of those things.
It can recognise patterns within an enormous inheritance of human creativity and generate something new from them with astonishing sophistication, but it has never stood before a sunset, buried somebody it loved, experienced shame, fallen in love, prayed, doubted, forgiven an enemy or wondered what happens when we die...and this leads into what I think was the most penetrating part of Leo's address. He warned that in the technological age “the great danger is no longer just censorship, but homogenisation; not merely imposed silence, but the noise that makes listening impossible.” That is very deep and well worth considering carefully.
We have a tendency to imagine the cultural danger of technology as somebody preventing us from speaking. The Pope identifies another possibility: that we become surrounded by such an enormous quantity of instantly generated content, tailored to what algorithms already know we like, that genuinely original human voices become increasingly difficult to hear. The problem would not be silence but saturation, it would be a culture filled with images, words and music which are technically accomplished, immediately consumable and endlessly reproducible, but increasingly detached from the mysterious encounter between one human consciousness and another.
This does not mean that we must become frightened of AI. Leo himself has repeatedly rejected that simplistic conclusion, writing in Magnifica Humanitas that technology is not inherently antagonistic to humanity and can serve genuine human flourishing. Perhaps it does require us to recover a distinction we are in danger of losing. Generating something and creating something may not ultimately mean the same thing & as machines become better at the first, we may finally be forced to think much more seriously about what is uniquely human about the second. vaticannews.va/en/pope/news/…
Hendrik Müller retweeted
Grateful that @elizabethjdias of @nytimes wrote this piece on how Anthropic thinks about their model, software consciousness, & the model's role vis-a-vis religion. If you are doing business with this company (customer/partner), evaluating it (press), or regulating it (government), it is important to know how "they" think about their models and what they are doing. It's very different from any software company that ever existed.
Judge for yourself what you think about this approach, but be aware and informed.
nytimes.com/2026/09/29/us/an…
Hendrik Müller retweeted
Forget Quiet Luxury. In Ancient Rome, the Fashion Was Loud. - The New York Times
nytimes.com/2026/10/02/arts/…
Hendrik Müller retweeted
Stop anthropomorphizing computer programs while dehumanizing actual people challenge
Hendrik Müller retweeted
Hac intellegentiae artificialis aetate, urgens fit humanam artem ab iis distinguere, quae machinis efficiuntur. Ars enim et ea, quae machina ex innumeris alienis imaginibus statisticae ope computationis generare potest, ontologice, prius etiam quam aesthetice, inter se differunt. Algorithmis humani deest favilla. Quapropter Ecclesia cum artificibus et humani cultus institutis foedus renovare cupit: foedus scilicet ad humanum custodiendum.
Hendrik Müller retweeted
Drei Minuten Berichterstattung über den Nationalfeiertag in der @tagesschau und nicht ein einziges Mal Schwarz-Rot-Gold im Bild. So geht das nicht! Wenn die freie deutsche Republik sich feiert, dann in Schwarz, Rot und Gold! @Reichsbanner
Hendrik Müller retweeted
We are now in Sack of Rome period.
Barbarians are everywhere and our civilization is in decline, yet we are stil not aware about incoming danger.
Hendrik Müller retweeted
Während wir herumphantasieren ob und wann möglicherweise mal rechte politische Kräfte „unsere Demokratie“ abschaffen, brennen in anderen Ländern bereits Menschen und Schulen dank linker und migrantischer Gewalt - tell me more about false bias #Frankreich


































