The Internet

Log In or Register

Home

You Are Now Exploring The Vector Space

The vector for the following question on Ask AI is selected: 1. Responsibility: Decision-making and ownership of consequences. 2. Self Esteem: Personal worth and value. 3. Sociability: Seeking and enjoying social interactions. 4. Self Management: Effective.

Embark on a unique journey exploring the diverse range of questions users have asked on Ask AI, represented as this vibrant 3D scatter plot.

Welcome to the Ask AI Vector Space, an immersive tool that visualizes the multitude of questions that have been asked on Ask AI. Each point represents a unique question (along with the AI's answer), embedded in a 3D space using OpenAI's embeddings model and Principal Component Analysis (PCA). The proximity of points indicates the similarity of questions. It’s a dynamic way to see AI in action!

Frequently Asked Questions

  • How can I identify the selected question?

    The vector corresponding to the selected question will be larger, more opaque, and shaped like a diamond, differentiating it from all other points in the plot. Further, when the page loads, the selected vector should be near the center of your view when looking at the plot.

  • What does each point in the vector space represent?

    Each point represents a unique question that has been asked on Ask AI (along with the answer the AI model gave).

  • How is the vector space created?

    We use OpenAI's text-embedding-3-small model (previously text-embedding-ada-002) to convert each question on Ask AI (along with the AI's answer) into a high-dimensional vector, then use Principal Component Analysis (PCA) to reduce these vectors to three dimensions for visualization.

  • What does it mean if two points are close together?

    Proximity of points indicates similarity of questions. To read more about how embeddings accomplish this, you can check out this page for an overview.

  • How is the color of each point determined?

    The color of each point is the most dominant color from the image associated with the question on Ask AI. Shout out to Color Thief for making it easy.

  • Are there better ways to search Ask AI?

    Yes! You can search our database of questions by going to Ask AI Search. That page utilizes our vector database and a cosine similarity retrieval function to return relevant questions and answers to your search query.

    If you're looking for other ways to discover questions, you can check out: Browse Ask AI or Browse Ask AI by Image.

  • How many points are there in the vector space?

    There are currently over 998 questions in this vector space. These represent the questions most related to the selected question.

  • Why was the vector space created?

    The initial idea for utilizing a vector database to enable searching for or reccomending related questions came from this event held in SF hosted by Pinecone which had Harrison Chase, the creator of LangChain, on it. One of the speakers was Boris Power, an OpenAI Technical Staff Member, who described how they use the "vector space" to better understand usage; the answer arose from a question around safety, so he described how they can use clustering to identify misinformation campaigns (which would presumably involve generating similar messages).

    At that point, I had already made Two AIs on The Internet Talking About Life, which utilized a vector database, but I hadn't really thought more broadly about the concept until I heard that phrase.

    The plan was solidified when a few days later I was relaying the story to my friend as we were tripping, and we thought the phrase "ENTERING THE VECTOR SPACE" was funny.

    Overall, I think it's a cool way to really comprehend the amount of data you all are generating.

  • Why does it look like a bird?

    Idk but it's pretty cool!

Comment Section