Okładka książki
Książka po angielsku
The Articulation Tax
Opis książki
Most debates about AI safety ask whether a model should refuse a dangerous request. This book asks the harder question: What happens after the refusal?When an AI system flags a private conversation, a chain of decisions may begin-automated classification, data retention, human review, identity linkage, account sanctions, or even referral to public authorities. These actions are often described as parts of a single "safety" process. But they are not the same act, and they should not require the same evidence, authority, or oversight.Beginning with public reporting surrounding the Darren Zhou case, The Articulation Tax examines how a private conversation can become a matter of institutional suspicion. It asks who defines dangerous speech, who is permitted to inspect it, how private companies acquire quasi-public judgment powers, and what remains of due process when the standards themselves are hidden, personalized, and constantly changing.At the center of the book is the concept of the articulation tax: the hidden cost of translating an intuitive sense of injustice into the precise language a system recognizes. A user who can name the relevant distinctions-between refusal and reporting, thought and action, detection and judgment, company policy and law-may receive a serious structural answer. Someone who cannot may receive only reassurance that the system is working as intended.This difference is not merely conversational. It can determine who is able to recognize power, contest a decision, or even understand what has happened to them.Joseph Lee explores: - The difference between model refusal and external referral- Data sovereignty as a question of technical control, not promises- The privatization of suspicion through AI safety systems- The danger of collapsing "dangerous ideas" into dangerous actions- Why personalized explanations can weaken public reason- Whether separation of powers can exist inside a black-box model- How local and open AI may preserve technological pluralism- Where democratic friction must be placed before consequences become irreversibleThis is not an argument against AI safety, nor a defense of genuine threats. It is an argument that safety systems must be treated as fallible sensors-not sovereign judges.As AI becomes an intermediary between private thought and public authority, the essential question is no longer simply whether the system can detect danger.It is who gets to decide what counts as dangerous-and who has the power to act on that decision.
- ISBN-13
- 9798192832097
- Język
- Angielski
- Waga
- 95 g
- Oprawa
- Miękka
- Wydawnictwo
- Amazon Digital Services LLC - Kdp
- Liczba stron
- 60
- Rok wydania
- 2026
- Kategoria
- Nauka, technika i medycyna
Termin wysyłki jest obecnie niedostępny.
Wszystkie podane terminy wysyłki liczymy w dniach roboczych.