ChatGPT Issues Sudden Ban, Cutting Off Global AI Writers Overnight

marsbitОпубліковано о 2026-07-31Востаннє оновлено о 2026-07-31

Анотація

The Golden Age of AI writing is showing signs of collapse. While AI models grow more powerful, the quality of their generated text is perceived to be declining. Research notes a divergence: models excel in structured domains like coding and math due to clear feedback signals, but their language expression and reasoning plateau or even regress, as there's no reliable automated metric to score "good writing." Simultaneously, platforms are imposing new restrictions. ChatGPT has reportedly begun refusing prompts that explicitly instruct it to mimic the style of specific famous authors like Stephen King or J.K. Rowling, though requests based on abstract narrative features remain possible. This shift undermines user expectations. AI writing tools are becoming less reliable for end-to-end content creation, as model capabilities may not consistently improve and prompt efficacy can change overnight due to platform policy updates. Consequently, AI is receding to a more auxiliary role—a tool for research, structuring, and editing—while human judgment, experience, and creative intent regain paramount importance.

The golden age of AI writing is collapsing!

Entering it, each generation of models grows stronger, yet the articles become increasingly worse.

It knows when to list several suggestions, where to insert subheadings, and never forgets to add a concluding sentence at the end.

But after reading the entire piece, not a single sentence remains in your mind.

The Stronger the Model, the Less Human the Writing Sounds

Recently, Cambridge University postdoctoral researcher Adam Hunt also noticed this phenomenon. His research focuses on evolutionary psychiatry and human evolution.

Hunt was previously optimistic about AI. However, on July 28, he frankly admitted in a long post: he is growing increasingly pessimistic now.

There was a popular metaphor before, depicting model capability as a circle with a sharp spike.

First, code and mathematics surpass humans, then as scale increases, every direction gradually grows, ultimately achieving AGI.

But in reality, the spike for code and mathematics is indeed growing longer, while language expression and simple reasoning have not improved synchronously.

From a reinforcement learning perspective, this outcome is almost inevitable.

Earlier generations of LLMs appeared to become "comprehensively smarter" because the training corpus itself covered all domains, from poetry to theses, encompassing the entire internet.

However, that was a byproduct of the corpus, not the model's genuine "comprehension."

Chain-of-thought and web search were subsequently introduced, giving models a temporary boost. But ultimately, this path hit a wall.

Following that, AI companies eager for a turnaround turned their focus to code.

The reason is simple: there is a constant stream of training data available.

GitHub commits, reviews, and PRs are all readily available. Moreover, similar to mathematics, code provides clear feedback signals: either it runs or it doesn't.

What about writing articles? How do you automatically score "how well a piece of prose is written"? Without a reward, RL won't optimize in that direction.

Thus, all training efforts shifted to code, while language expression had to rely on the foundation laid during the pre-training phase.

This foundational knowledge can only diminish over time. To accommodate improvements in other areas, it might even degrade.

This is precisely where Goodhart's law tends to manifest.

When a metric becomes the training target, it no longer accurately reflects true capability. Labs optimize models using benchmark tests, then prove their improvement with similar tests.

Benchmark accuracy scores will continue to rise. Yet, the actual user experience may stagnate or even decline.

This Time, Imitation Is Also Restricted

While capabilities are degrading, another development is occurring simultaneously.

A Reddit user complained in a post that they had been using ChatGPT to write a book for several months, subscribing to the advanced model. They took a break in between.

When they returned, the prompt that had always worked suddenly returned a refusal.

The prompt itself wasn't complex: specify an author, request to increase dialogue, enrich details, and avoid fragmented sentences.

In the past, simply writing the author's name allowed the model to generate content approximating their style, without needing additional explanations on sentence length, rhythm, or perspective.

Now, this type of prompt no longer works.

Ars Technica tested with Stephen King, J.K. Rowling, Amy Tan, Dickens, Hemingway, and Engadget tested with Agatha Christie. ChatGPT refused to imitate all of them.

If you don't name names and only request abstract features like "suspense intensity," "narrative pacing," or "dialogue density," it can still output. But the flavor is no longer the same.

The Golden Age of AI Writing Has Collapsed

Today, this controversy has shaken users' fundamental expectations regarding AI writing.

Models may not necessarily write better, and prompts are no longer guaranteed to remain effective long-term.

When platforms change their rules, entire writing pipelines built around specific models can collapse overnight.

However, AI writing won't disappear; it has merely retreated to its proper place: as a tool.

It can research information, structure content, and revise sentences, but it cannot replace the author in deciding what to write and why to write it.

In other words, human experience, judgment, and desire to express are becoming increasingly valuable.

Reference: https://arstechnica.com/ai/2026/07/chatgpt-stops-cloning-famous-writers-voices-but-may-capture-a-similar-feeling/

This article is from WeChat public account "New Zhiyuan," author: ASI Apocalypse

Трендові криптовалюти

Пов'язані питання

QAccording to the article, what is the main reason for the decline in the quality of AI-generated writing?

AThe article attributes the decline to reinforcement learning (RL) optimization focusing on domains with clear, automated feedback like code, while creative writing lacks such objective metrics, causing language expression capabilities to stagnate or degrade as models prioritize other benchmarks.

QWhat specific change did ChatGPT make regarding prompts that imitate famous writers?

AChatGPT now refuses to generate content in the style of specific, named famous authors like Stephen King or J.K. Rowling when prompted directly. Prompts using only abstract stylistic features are still allowed, but the output is less authentic.

QWhat is Goodhart's law, as mentioned in the article, and how does it relate to AI model training?

AGoodhart's law states that when a measure becomes a target, it ceases to be a good measure. The article relates this to AI companies optimizing models for specific benchmark tests, which improves scores on those tests but doesn't necessarily translate to better real-world performance, potentially leading to a decline in user experience.

QWhat does the article suggest is the future role of AI in writing?

AThe article suggests AI writing will recede from attempting to replace authors and settle into its proper role as a tool. It will be used for tasks like researching, structuring content, and editing sentences, but the human author's experience, judgment, and desire to express remain central and increasingly valuable.

QWhat analogy does researcher Adam Hunt use to describe the unrealistic expectation of AI model development?

AHe references a popular analogy depicting model capability as a spiky circle, where proficiency in areas like code and math grows into a sharp peak, expecting all other abilities like language to grow uniformly toward AGI. The article argues that in reality, the 'spike' for code has grown while language and simple reasoning have not kept pace.

Пов'язані матеріали

a16z: From Companies to DAOs, DUNA May Become the Next Generation Organizational Form

This article, "From Companies to DAOs: How DUNA Could Become the Next Organizational Form," traces the 500-year evolution of business collaboration. It begins with medieval structures like the *commenda* and Florentine *compagnia*, which exposed partners to personal risk. The modern corporation, exemplified by the Dutch East India Company (VOC), was a revolutionary leap, enabling large-scale, capital-intensive ventures by offering limited liability and reducing coordination costs. However, corporations introduced new challenges like principal-agent problems and bureaucratic overhead. The piece argues that software and internet-native protocols are now reducing these traditional overheads. Decentralized Autonomous Organizations (DAOs) emerged as a new model for coordination without centralized management. Yet, DAOs face a significant legal vacuum: they lack legal recognition, leaving members exposed to unlimited personal liability, and their tokens are vulnerable to being classified as securities under unclear regulations (e.g., the Howey Test). This has forced projects into suboptimal workarounds like offshore foundations. The article identifies the Decentralized Unincorporated Nonprofit Association (DUNA) as a potential solution. Recently legalized in states like Wyoming, the DUNA provides a legal wrapper for decentralized networks. It grants key protections—legal personality, limited liability, and perpetual existence—to a group without imposing a traditional hierarchical management structure. This allows token-holder communities to govern, hold assets, and contract as a single legal entity, aligning with their decentralized nature. While DUNA doesn't solve all governance challenges or magically resolve securities law questions, it represents a crucial step. It fills the legal recognition gap, offering a native legal form for internet-scale, decentralized collaboration and extending the separation of personal risk from organizational venture into a new domain.

marsbit6 хв тому

a16z: From Companies to DAOs, DUNA May Become the Next Generation Organizational Form

marsbit6 хв тому

2026 Mid-Year Report On-Chain RWA: Tokenized Stock Market Cap Doubles in a Year, But 90% of Rights Are Hollow Shells

The 2026 Mid-Year Report on On-Chain RWA highlights a significant growth in tokenized stock market capitalization, which nearly doubled from $951 million in March to $1.89 billion by July. However, the report reveals a fundamental contradiction in this "layer 2.5" ecosystem: products with the strongest legal foundation (like regulated U.S. infrastructure) lack liquidity and distribution, while freely tradable offshored wrapper products often lack substantive ownership rights. The increase is driven largely by a few products (SECZ, FGRS, STRCx) and platforms (Ondo, xStocks, Securitize collectively hold over 85% share). While distributed value across networks like Ethereum, Solana, and BNB Chain has grown, the market remains fragmented. Products referencing the same underlying asset (e.g., Apple stock) are distinct legal liabilities with different intermediaries and jurisdictional rules, offering varying degrees of legal claim. The report cautions that headline numbers are misleading, as they reflect changes in distributed token value—driven by issuance, conversions, and price movements—not pure investor inflows. True "canonical shares" with legal ownership, wide wallet distribution, institutional liquidity, and independent on-chain price discovery do not yet exist at scale. Tokenized treasuries show stronger product-market fit, and ETFs may be easier to scale than single stocks. The core takeaway is a trade-off: legal certainty versus liquidity and composability.

marsbit55 хв тому

2026 Mid-Year Report On-Chain RWA: Tokenized Stock Market Cap Doubles in a Year, But 90% of Rights Are Hollow Shells

marsbit55 хв тому

Coldcard Hardware Wallet Hacked: 594 Bitcoin Withdrawn in 25 Minutes

The Coldcard hardware wallet has been compromised, with hackers stealing approximately 594.5 Bitcoin (~$40 million) from 500 addresses in just 25 minutes. The root cause was a critical software bug, undetected for five years, which disabled the device's secure chip for generating true random numbers. This led to the creation of private keys based on predictable data like the processor's serial number, drastically reducing cryptographic security. The attackers exploited this offline by brute-forcing possible seed phrases, finding active addresses on the public ledger, and signing transactions. Initially, Coinkite (Coldcard's maker) claimed only older models were at risk but later admitted all devices running the compromised firmware were vulnerable. CEO Rodolphe Novak (NVK) apologized but ruled out financial compensation for affected users. To secure funds, owners must urgently update their firmware to specific safe versions, generate a completely new seed phrase on the updated device, and transfer all assets to new addresses created with that new seed. While a BIP-39 passphrase can help, it does not replace this migration process. Other Coinkite products like TAPSIGNER were not affected. This incident underscores that even specialized hardware requires rigorous, independent code audits, especially for cryptographic functions. It parallels past failures, like a 2006 OpenSSL bug in Debian, and raises questions about whether automated code analysis can ever fully replace human scrutiny in critical security areas.

cryptonews.ru3 год тому

Coldcard Hardware Wallet Hacked: 594 Bitcoin Withdrawn in 25 Minutes

cryptonews.ru3 год тому

Торгівля

Спот

Популярні статті

Як купити BAN

Ласкаво просимо до HTX.com! Ми зробили покупку Comedian (BAN) простою та зручною. Дотримуйтесь нашої покрокової інструкції, щоб розпочати свою криптовалютну подорож.Крок 1: Створіть обліковий запис на HTXВикористовуйте свою електронну пошту або номер телефону, щоб зареєструвати обліковий запис на HTX безплатно. Пройдіть безпроблемну реєстрацію й отримайте доступ до всіх функцій.ЗареєструватисьКрок 2: Перейдіть до розділу Купити крипту і виберіть спосіб оплатиКредитна/дебетова картка: використовуйте вашу картку Visa або Mastercard, щоб миттєво купити Comedian (BAN).Баланс: використовуйте кошти з балансу вашого рахунку HTX для безперешкодної торгівлі.Треті особи: ми додали популярні способи оплати, такі як Google Pay та Apple Pay, щоб підвищити зручність.P2P: Торгуйте безпосередньо з іншими користувачами на HTX.Позабіржова торгівля (OTC): ми пропонуємо індивідуальні послуги та конкурентні обмінні курси для трейдерів.Крок 3: Зберігайте свої Comedian (BAN)Після придбання Comedian (BAN) збережіть його у своєму обліковому записі на HTX. Крім того, ви можете відправити його в інше місце за допомогою блокчейн-переказу або використовувати його для торгівлі іншими криптовалютами.Крок 4: Торгівля Comedian (BAN)Легко торгуйте Comedian (BAN) на спотовому ринку HTX. Просто увійдіть до свого облікового запису, виберіть торгову пару, укладайте угоди та спостерігайте за ними в режимі реального часу. Ми пропонуємо зручний досвід як для початківців, так і для досвідчених трейдерів.

560 переглядів усьогоОпубліковано 2024.12.11Оновлено 2026.06.02

Як купити BAN

Обговорення

Ласкаво просимо до спільноти HTX. Тут ви можете бути в курсі останніх подій розвитку платформи та отримати доступ до професійної ринкової інформації. Нижче представлені думки користувачів щодо ціни BAN (BAN).

活动图片