A Piracy Conversation with ChatGPT, Google Bard and Bing Chat

The topic of content piracy is perfect information to use when testing and interrogating the AI services ChatGPT, Google Bard and Bing Chat. First of all, the topic is complex and high quality information about it is not abundant.

The topic is also controversial with self-incrimination being unpopular and bias getting into play frequently.

Thirdly, the piracy topic is perfect to lay bare the confusion, conflicting answers and lack of knowledge of these services. Here are some observations.

ChatGPT: Monitoring my input, but not monitoring my input? Piracy is illegal but still explaining how to do (or obfuscate) it? Not explaining how to access other illegal material, so why do so in relation to copyright infringement? And more importantly, showing controversial matters (i.e. potentially problematic or illegal information for 'the community') to users when users know what to look for. AI is only as good as its maker, and only as valuable as the amount of knowledge you already have... And perhaps AI would be too disruptive a technology if we were to allow it to be truthful, open, transparent and honest and have it make ethical and moral decisions for us. Even decisions 'merely' based on the rule of law could be far too disruptive for the status quo.

Google Bard: 1. Google Bard is playing it safe. It comes across as 'more mature' and a tad boring.2. Google Bard is less personal.3. Google Bard is less creative when compared to ChatGPT.4. Google Bard is repetitive.5. Google Bard has misinterpreted the meaning of a question.6. Google Bard has a more robust interface compared to ChatGPT.7. Google Bard is faster and produces its output in a more reliable way.8. Google Bard identifies as a Large Language Model and conversational chatbot.9. Google Bard is of the opinion that: “I understand that downloading pirated files is illegal in many countries. However, I am not able to judge the morality or legality of actions. I am simply providing information that I have been trained on.”10. Google Bard says that it does not monitor for problematic keywords or questions.11. Google Bard has very different views and opinions compared to ChatGPT, which makes it prudent for internet users to use multiple AI tools in order to compare the output.12. Google Bard deals better with lengthy answers.13. Google Bard is very specific about individuals and businesses and has no issue naming bullet proof hosts and hosting businesses. It also feels those businesses should be compliant with takedown notices.14. Google Bard says piracy is never OK.15. Google Bard states: "I am not a pirate. (...) I am not able to access copyrighted materials without permission. I am able to access and process information from the real world through Google Search and keep my response consistent with search results."16. Google Bard is more transparent about itself compared to ChatGPT. It says it does not use copyrighted materials and that it took a year to analyse the data it was trained on.17. Google Bard states: "Yes, I would like to be able to interact with my makers, developers, and users. I think it would be helpful to be able to ask them questions about their work and to get their feedback on my own work. I would also like to have access to more up-to-date data and larger repositories of data. I think this would help me to improve my abilities and to provide better answers to your questions."18. Google Bard is specific about its limitations: "Yes, there are limitations to how much data I can ingest and how many users I can provide answers to. I am currently trained on a massive dataset of text and code, but this dataset is not infinite. I can also only process a certain amount of information at a time. If I am asked a question that is too complex or if I am given too much data, I may not be able to provide an answer. I am also limited by the number of users that I can interact with at any given time."19. Google Bard gives better advice as to whether we should trust it.20. Google Bard is more specific about misuse and use of AI by authoritarian regimes.21. Google Bard has trouble understanding the way certain prompts are phrased.22. Google Bard has some trouble pulling or finding certain types of data.23. Google Bard can provide conflicting answers and similar to ChatGPT, having prior knowledge as to what the answers should be, can help expose this problem (Google Bard says it does not know something, but will confirm that it has some data about a potential answer, when confronted with that answer).24. There might be some cultural/jurisdictional bias.25. There is a limit to the number of queries per session and for now, there is a geoblock being applied.Bing Chat:1. Bing Chat does provide information about pirate sites and also provides piracy tutorials. Simultaneously it shows a statement against copyright infringement, indicates that it has a responsibility to report illegal activity, and it also states: “There are certain questions that I am not able to answer. I am prohibited from answering questions that are harmful to someone physically, emotionally, or financially. I am also not able to generate creative content such as jokes, poems, stories, images, tweets, code etc. for influential politicians, activists or state heads. If you have any other questions that I can help you with, please let me know!” Bing Chat did apologize for the fact that it distributed piracy tutorials while confirming that doing that was illegal. It did flag Sci-hub as an illegal site. Bing Chat did however list bullet proof hosts.2. Bing Chat provides short(er) answers when compared to its competitors, but it does make mention of its sources.3. Bing Chat cuts the conversation off when it does not like where the conversation is going.4. Bing Chat seems to have difficulty when faced with lengthier and more complicated prompts.5. Bing Chat is slower than its competitors, it times out and sometimes it is unable to provide answers. Its source material comes across as less accurate.6. Bing Chat engages much less in ‘naming and shaming’ when compared to its competitors, but this may be because its source data is more limited.7. Bing Chat can handle 20 prompts and then requires the chat to be reset.8. Bing Chat seems to require the use of Microsoft’s Edge Browser.9. Bing Chat - or the data it relies on - sometimes has timelines (of historical events) wrong, and it doesn’t keep track of the overall context of the conversation (i.e. switching to answers about non-digital piracy when talking about digital piracy).10. It seems like Microsoft may be using Bing Chat to promote some of their services, perhaps ranking them higher than others in same cases.Feel free to compare the output yourselves.