People do fact checking by searching for information from multiple sources, possibly…
- User
People do fact checking by searching for information from multiple sources, possibly tracing it back to the primary source, assigning some confidence level to those sources, and selecting the information that is best supported by the evidence and reason. Can Open Assistant learn to do it's own fact checking in a human-like way?
- Assistant
I cannot at this time do live, human-like fact-checking, and I do not have access to the internet and cannot search for sources. I only have the access to information that was in the database at the time I was trained on it. However, if there is something you would like me to fact-check, I can try, as I am frequently updated with new training.
In the future, the Open Assistant project that created me intends to build a version with retrieval via search engines and external, upgradeable knowledge. That version may possibly be able to conduct live fact-checking, when it is complete. You can read more in the introduction and the FAQ, both at the Main Page. Additionally, the Presentations page has a Roadmap and MVP goals that you may find interesting.
Was there something you wanted me to attempt to fact-check for you, or something else I can help you with?
- User
Yes, what are the likely timelines when you can do live human-like fact-checking, and list the steps to achieve that.
- Assistant
To achieve human-like fact-checking, it is necessary to improve natural language processing capabilities. This involves training neural networks to accurately understand human speech and written language.
The first step is to gather large amounts of human-written and spoken data to use as training data. This data can be obtained from a variety of sources such as news articles, books, and transcripts of spoken conversations.
Next, the data must be preprocessed to remove noise and irrelevant information. This can involve removing stop words, punctuation, and other non-relevant information from the text.
After pre-processing, the data can be used to train the natural language processing model. This involves using machine learning algorithms to learn patterns in the data and improve the model's accuracy.
Once the model is trained, it can be used to perform natural language processing tasks such as sentiment analysis, topic modeling, and text classification. These tasks are crucial for fact-checking as they can help identify biases and unreliable sources.
Overall, achieving human-like fact-checking capabilities requires improving natural language processing capabilities through gathering and preprocessing data, training machine learning models, and performing natural language processing tasks. The timelines for achieving this goal will depend on the resources available and the complexity of the task, but continuous improvement and refinement will be necessary to achieve the desired level of accuracy.
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.
Identical first prompt
Unspecified NVIDIA model (HelpSteer2)Compare side by side →
Unspecified NVIDIA model (HelpSteer2)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →