Merge pull request #1 from Kubus42/wise2024

Wise2024

_freeze/about/projects/execute-results/html.json

@@ -1,10 +1,10 @@
 {
-  "hash": "fd1df698d85be1c0799bf4ed90908b68",
+  "hash": "f51f1e846444ec9873220f31880e7605",
   "result": {
     "engine": "jupyter",
-    "markdown": "---\ntitle: Projects\nformat:\n  html:\n    code-fold: true\n---\n\nIn the final part of the seminar we are going to tackle our very own projects involving a language model. \nAt best, you find your ideas and work on them, maybe you even have a work-related application mind. \nThe following list can serve as inspiration. \n\n\n### Project ideas\n\n1. **Question-Answering Chatbot**: Build a chatbot that can answer questions posed by users on a specific topic provided in form of documents. Users input their questions, the chatbot retrieves relevant information from a pre-defined set of documents, and uses the information to answer the question.\n2. **Document tagging / classification:** Use GPT and its tools (e.g., function calls) and/or embeddings to classify documents or assign tags to them. Example: Sort bug reports or complaints into categories depending on the problem.\n3. **Clustering of text-based entities:** Create a small tool that can cluster text-based entities based on embeddings, for example, groups of texts or keywords. Example: Structure a folder of text files based on their content.\n4. **Text-based RPG Game**: Develop a text-based role-playing game where players interact with characters and navigate through a story generated by GPT. Players make choices that influence the direction of the narrative.\n5. **Sentiment Analysis Tool**: Build an app that analyzes the sentiment of text inputs (e.g., social media posts, customer reviews) using GPT. Users can input text, and the app provides insights into the overall sentiment expressed in the text.\n6. **Text Summarization Tool**: Create an application that summarizes long blocks of text into shorter, concise summaries. Users can input articles, essays, or documents, and the tool generates a summarized version.\n7. **Language Translation Tool**: Build a simple translation app that utilizes GPT to translate text between different languages. Users can input text in one language, and the app outputs the translated text in the desired language. Has to include some nice tweaks.\n8.  **Personalized Recipe Generator**: Develop an app that generates personalized recipes based on user preferences and dietary restrictions. Users input their preferred ingredients and dietary needs, and the app generates custom recipes using GPT.\n9.  **Lyrics Generator**: Create a lyrics generation tool that generates lyrics based on user input such as themes, music style, emotions, or keywords. Users can explore different poetic styles and themes generated by GPT.\n\n\n### Project setup\nTODO: Describe the idea of Dash and the app in Jupyterlab.\n\n",
+    "markdown": "---\ntitle: \"Projects\"\nformat:\n  html:\n    code-fold: true\njupyter: python3\n---\n\n\n\n\nIn the final part of the seminar we are going to tackle our very own projects involving a language model. \nAt best, you find your ideas and work on them, maybe you even have a work-related application mind. \nThe following list can serve as inspiration. \n\n\n### Project ideas\n\n1. **Question-Answering Chatbot**: Build a chatbot that can answer questions posed by users on a specific topic provided in form of documents. Users input their questions, the chatbot retrieves relevant information from a pre-defined set of documents, and uses the information to answer the question.\n2. **Document tagging / classification:** Use GPT and its tools (e.g., function calls) and/or embeddings to classify documents or assign tags to them. Example: Sort bug reports or complaints into categories depending on the problem.\n3. **Clustering of text-based entities:** Create a small tool that can cluster text-based entities based on embeddings, for example, groups of texts or keywords. Example: Structure a folder of text files based on their content.\n4. **Text-based RPG Game**: Develop a text-based role-playing game where players interact with characters and navigate through a story generated by GPT. Players make choices that influence the direction of the narrative.\n5. **Sentiment Analysis Tool**: Build an app that analyzes the sentiment of text inputs (e.g., social media posts, customer reviews) using GPT. Users can input text, and the app provides insights into the overall sentiment expressed in the text.\n6. **Text Summarization Tool**: Create an application that summarizes long blocks of text into shorter, concise summaries. Users can input articles, essays, or documents, and the tool generates a summarized version.\n7. **Language Translation Tool**: Build a simple translation app that utilizes GPT to translate text between different languages. Users can input text in one language, and the app outputs the translated text in the desired language. Has to include some nice tweaks.\n8.  **Personalized Recipe Generator**: Develop an app that generates personalized recipes based on user preferences and dietary restrictions. Users input their preferred ingredients and dietary needs, and the app generates custom recipes using GPT.\n9.  **Lyrics Generator**: Create a lyrics generation tool that generates lyrics based on user input such as themes, music style, emotions, or keywords. Users can explore different poetic styles and themes generated by GPT.\n\n\n### Project setup\nSee slides.\n\n",
     "supporting": [
-      "projects_files"
+      "projects_files/figure-html"
     ],
     "filters": [],
     "includes": {}

_freeze/ethics/bias/execute-results/html.json

@@ -2,7 +2,7 @@
   "hash": "dbc2899275a9ac9022ac9f0ade590abd",
   "result": {
     "engine": "jupyter",
-    "markdown": "---\ntitle: Bias\nformat:\n  html:\n    code-fold: true\n---\n\nIt is crucial to also explore the concept of bias lurking within language models. \nWhile these models have revolutionized various fields and are arguably one of the most impactful new tools of the last few years, they aren't immune to inheriting and perpetuating biases present in the data they are trained on.\nSo what is Bias in Language Models?\n\nBias in language models refers to the skewed or unfair representation of certain groups, perspectives, or ideologies within the generated text. \nThese biases can stem from societal stereotypes, historical prejudices, or systemic inequalities embedded **in the training data**. \nIn particular for models trained on enormous corpora stemming from the internet, it is a nearly impossible task to examine all of the training data for dangerous our otherwise harmful content. \nAnd even simply the choice of the training data can create an inherent bias in the models. \nAs an example, consider training a model only on German data, which will inevitably introduce German opinions etc. into the model.\nWhen left unchecked, biased language models can reinforce existing prejudices, amplify underrepresented narratives, and marginalize certain communities.\n\n\n#### Types of Bias in Language Models\n\nThere are plenty of different types of bias that can occur in language models, here are just a few.\n\n- **Gender bias:** Language models may exhibit gender bias by associating specific roles, traits, or occupations with a particular gender. For example, phrases like \"brilliant scientist\" might more frequently generate male pronouns, while \"caring nurse\" might generate female pronouns, perpetuating stereotypes about gender roles.\n\n- **Ethnic and racial bias:** Language models may reflect ethnic or racial biases present in the training data, leading to stereotypical or discriminatory language towards certain racial or ethnic groups. For instance, associating negative traits with specific racial groups or making assumptions based on names or cultural references.\n\n- **Socioeconomic bias:** Language models might exhibit biases related to socioeconomic status, such as portraying certain occupations or lifestyles as superior or inferior. This can contribute to the reinforcement of class stereotypes and disparities.\n\n- **Cultural bias:** Language models may demonstrate cultural biases by favoring certain cultural norms, values, or references over others, potentially marginalizing or erasing the perspectives of minority cultures or communities.\n\n- **Confirmation bias:** Language models can inadvertently reinforce existing beliefs or viewpoints by prioritizing information that aligns with preconceived notions and ignoring contradictory evidence, leading to the perpetuation of misinformation or echo chambers.\n\n\n#### Implications of bias in language models\nThe presence of bias in language models has plenty of implications, in particular when societies start using language models frequently. \n\n- **Reinforcement of stereotypes:** Biased language models can perpetuate harmful stereotypes, further entrenching societal prejudices and hindering efforts towards inclusivity and diversity.\n\n- **Discriminatory outcomes:** Biased language models may lead to discriminatory outcomes in various applications, including hiring processes, automated decision-making systems, and content moderation algorithms, potentially amplifying existing inequalities.\n\n- **Underrepresentation and marginalization:** Language models may marginalize or underrepresent certain groups or perspectives, leading to the erasure of minority voices and experiences from the discourse.\n\n- **Impact on society:** Biased language models can have far-reaching consequences on society, shaping public opinion, reinforcing power dynamics, and influencing policy decisions, ultimately exacerbating social divisions and injustices.\n\n\n\n#### Addressing bias in language models\n\nSo, what can we (or the creators of language models) do?\n\n- **Diverse and representative data:** Ensuring that language models are trained on diverse and representative datasets spanning various demographics, cultures, and perspectives can help mitigate biases by providing a more balanced and inclusive training corpus.\n\n- **Bias detection and mitigation techniques:** Implementing bias detection and mitigation techniques, such as debiasing algorithms, adversarial training, and fairness-aware learning frameworks, can help identify and address biases in language models during the development phase.\n\n- **Ethical considerations and transparency:** Incorporating ethical considerations and promoting transparency in the development and deployment of language models can foster accountability and empower users to critically assess the potential biases and limitations of these models.\n\n- **Continuous monitoring and evaluation:** Regularly monitoring and evaluating language models for biases in real-world applications can help identify and rectify unintended consequences, ensuring that these models align with ethical standards and promote fairness and inclusivity.\n\n",
+    "markdown": "---\ntitle: \"Bias\"\nformat:\n  html:\n    code-fold: true\njupyter: python3\n---\n\n\n\n\n\nIt is crucial to also explore the concept of bias lurking within language models. \nWhile these models have revolutionized various fields and are arguably one of the most impactful new tools of the last few years, they aren't immune to inheriting and perpetuating biases present in the data they are trained on.\nSo what is Bias in Language Models?\n\nBias in language models refers to the skewed or unfair representation of certain groups, perspectives, or ideologies within the generated text. \nThese biases can stem from societal stereotypes, historical prejudices, or systemic inequalities embedded **in the training data**. \nIn particular for models trained on enormous corpora stemming from the internet, it is a nearly impossible task to examine all of the training data for dangerous our otherwise harmful content. \nAnd even simply the choice of the training data can create an inherent bias in the models. \nAs an example, consider training a model only on German data, which will inevitably introduce German opinions etc. into the model.\nWhen left unchecked, biased language models can reinforce existing prejudices, amplify underrepresented narratives, and marginalize certain communities.\n\n\n#### Types of Bias in Language Models\n\nThere are plenty of different types of bias that can occur in language models, here are just a few.\n\n- **Gender bias:** Language models may exhibit gender bias by associating specific roles, traits, or occupations with a particular gender. For example, phrases like \"brilliant scientist\" might more frequently generate male pronouns, while \"caring nurse\" might generate female pronouns, perpetuating stereotypes about gender roles.\n\n- **Ethnic and racial bias:** Language models may reflect ethnic or racial biases present in the training data, leading to stereotypical or discriminatory language towards certain racial or ethnic groups. For instance, associating negative traits with specific racial groups or making assumptions based on names or cultural references.\n\n- **Socioeconomic bias:** Language models might exhibit biases related to socioeconomic status, such as portraying certain occupations or lifestyles as superior or inferior. This can contribute to the reinforcement of class stereotypes and disparities.\n\n- **Cultural bias:** Language models may demonstrate cultural biases by favoring certain cultural norms, values, or references over others, potentially marginalizing or erasing the perspectives of minority cultures or communities.\n\n- **Confirmation bias:** Language models can inadvertently reinforce existing beliefs or viewpoints by prioritizing information that aligns with preconceived notions and ignoring contradictory evidence, leading to the perpetuation of misinformation or echo chambers.\n\n\n#### Implications of bias in language models\nThe presence of bias in language models has plenty of implications, in particular when societies start using language models frequently. \n\n- **Reinforcement of stereotypes:** Biased language models can perpetuate harmful stereotypes, further entrenching societal prejudices and hindering efforts towards inclusivity and diversity.\n\n- **Discriminatory outcomes:** Biased language models may lead to discriminatory outcomes in various applications, including hiring processes, automated decision-making systems, and content moderation algorithms, potentially amplifying existing inequalities.\n\n- **Underrepresentation and marginalization:** Language models may marginalize or underrepresent certain groups or perspectives, leading to the erasure of minority voices and experiences from the discourse.\n\n- **Impact on society:** Biased language models can have far-reaching consequences on society, shaping public opinion, reinforcing power dynamics, and influencing policy decisions, ultimately exacerbating social divisions and injustices.\n\n\n\n#### Addressing bias in language models\n\nSo, what can we (or the creators of language models) do?\n\n- **Diverse and representative data:** Ensuring that language models are trained on diverse and representative datasets spanning various demographics, cultures, and perspectives can help mitigate biases by providing a more balanced and inclusive training corpus.\n\n- **Bias detection and mitigation techniques:** Implementing bias detection and mitigation techniques, such as debiasing algorithms, adversarial training, and fairness-aware learning frameworks, can help identify and address biases in language models during the development phase.\n\n- **Ethical considerations and transparency:** Incorporating ethical considerations and promoting transparency in the development and deployment of language models can foster accountability and empower users to critically assess the potential biases and limitations of these models.\n\n- **Continuous monitoring and evaluation:** Regularly monitoring and evaluating language models for biases in real-world applications can help identify and rectify unintended consequences, ensuring that these models align with ethical standards and promote fairness and inclusivity.\n\n",
     "supporting": [
       "bias_files"
     ],