What does information cost? – On the price of an AI Chabot's work
Fuente:
Zenodo
Guardado en:
| Autor principal: | |
|---|---|
| Formato: | Recurso digital |
| Lenguaje: | inglés |
| Publicado: |
Zenodo
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866901644086607872 |
|---|---|
| author | Weinmann, Siegfried |
| author_facet | Weinmann, Siegfried |
| contents | <p class="AbstractLabel">Abstract</p> <p>This paper examines the information provided by an AI Chatbot. The starting point is Shannon’s technical concept of information, which treats the message <em>y</em> ∈ as a formal selection from a set of possible responses to a prompt <em>x</em> ∈ . The relevant statistical quantities are the conditional self-information <em>i</em>(<em>y|x</em>) and, on average, the conditional entropy <em>H</em>(<em>Y|X </em>= <em>x</em>), where <em>X </em>and <em>Y </em>denote the random variables of the prompt and the response. Landauer’s principle links information theory with thermodynamics; through the concept of entropy, it forms a bridge between information and energy in the form of a thermodynamic lower bound on heat for logically irreversible processes. This thermodynamic minimum energy is vanishingly small compared to the actual production energy <em>E</em><sub>prod</sub>(<em>x</em>,<em>y</em>) of the response <em>y </em>on the provider’s side, which in turn represents only one component of the production costs <em>K</em><sub>prod</sub>(<em>x</em>,<em>y</em>) alongside computing time, storage, infrastructure and proportionate system costs. Using the example <em>x </em>= ‘Name a prime number less than 10’, the transition from the theoretical information value of a specific response <em>y </em>to a rough price estimate for an AI chat is illustrated.</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_19604647 |
| institution | Zenodo |
| language | eng |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | What does information cost? – On the price of an AI Chabot's work Weinmann, Siegfried Information AI Chatbot Energy Costs <p class="AbstractLabel">Abstract</p> <p>This paper examines the information provided by an AI Chatbot. The starting point is Shannon’s technical concept of information, which treats the message <em>y</em> ∈ as a formal selection from a set of possible responses to a prompt <em>x</em> ∈ . The relevant statistical quantities are the conditional self-information <em>i</em>(<em>y|x</em>) and, on average, the conditional entropy <em>H</em>(<em>Y|X </em>= <em>x</em>), where <em>X </em>and <em>Y </em>denote the random variables of the prompt and the response. Landauer’s principle links information theory with thermodynamics; through the concept of entropy, it forms a bridge between information and energy in the form of a thermodynamic lower bound on heat for logically irreversible processes. This thermodynamic minimum energy is vanishingly small compared to the actual production energy <em>E</em><sub>prod</sub>(<em>x</em>,<em>y</em>) of the response <em>y </em>on the provider’s side, which in turn represents only one component of the production costs <em>K</em><sub>prod</sub>(<em>x</em>,<em>y</em>) alongside computing time, storage, infrastructure and proportionate system costs. Using the example <em>x </em>= ‘Name a prime number less than 10’, the transition from the theoretical information value of a specific response <em>y </em>to a rough price estimate for an AI chat is illustrated.</p> |
| title | What does information cost? – On the price of an AI Chabot's work |
| topic | Information AI Chatbot Energy Costs |
| url | https://doi.org/10.5281/zenodo.19604647 |