Metaphor – LLMs Search Engine

Key Points: LLMs Search Engine, 1000 free requests per month.

Metaphor Systems AI-Powered Search Engine

Metaphor Systems is an innovative search engine built from the ground up using artificial intelligence (AI). The primary aim of Metaphor Systems is to deliver search results comparable to what an exceptionally knowledgeable friend might provide over the course of a week, but in an instant. Unlike traditional search engines, Metaphor offers a distinct approach to search by utilizing generative AI techniques to connect large language models (LLMs) with the vast expanse of internet knowledge. This technique enables Metaphor to present users with information beyond the capabilities of conventional search engines.

Metaphor Systems was established in 2021 and is headquartered in New York. The platform primarily caters to data scientists, data engineers, and AI practitioners, providing them with a search and discovery tool to enhance their understanding of various domains. The company's innovative approach to search is rooted in representation learning, which facilitates the generation of novel insights from the internet's vast collection of information.

This AI-powered search engine, Metaphor Systems, has garnered attention for its distinctive approach to online search. It has been praised for its ability to uncover new and valuable content on the internet that might be overlooked by other search engines. The engine's capabilities are founded on generative AI technology, the same kind used in various other AI applications.

Metaphor: A search engine based on large language models

Allows users to search using full sentences and natural language, and can also search for content in a way that mimics the way people share and talk about links on the Internet.

Metaphor can also be used in conjunction with LLMs, allowing LLMs to connect to the Internet and query content outside their knowledge base.

After testing, it only supports English, and the effect is not bad, similar to the feeling of chatting with ChatGPT, but the output result is to output a lot of content at once, and it feels very convenient!

working principle:

Metaphor's AI models are trained to make link predictions. This means that given some text cue, it tries to predict the link that is most likely to follow that cue.

This method simulates the way people talk about links on the Internet as a measure of link content and quality.

For example, a user could use the prompt: "Besides Terry Tao's blog, this is my favorite math blog:". Instead of searching for "best math blogs".

Use the hint: "This person would be a good candidate for motorcycles (personal website here:" simulates the natural way people speak when sharing a link.

The main functions

1. Search using a specific descriptor or "vibe":

•Users can use specific words or expressions to describe the feel or vibe of the content they want to find.
• For example, if a user wants to find some relaxing, romantic music, they can search using descriptors such as "relaxing jazz" or "romantic dinner music."
• This approach allows users to express their intent more precisely and find results that better match their emotions and vibes.

2. Search only for the entity types they want:

• Users can specify specific types or categories of search results they want.
• For example, if users are only interested in academic articles, they can specify to search only academic articles, not blog posts, news articles, or other types of content.
• This helps users find the specific type of content they want more quickly and filters out irrelevant results.

3. Find content that traditional search engines fail to show:

• Metaphor finds content that might not be well represented by traditional search engines due to inappropriate keywords or other reasons.

4. Search by the link itself:

•Users can provide a link to find other links that are most similar to it.
• This helps users find more resources related to a particular topic or content.

5. Integration with large language models (LLMs):

• Metaphor can be used in conjunction with LLMs, allowing LLMs to query the outside world and consume content on the Internet.
• This helps to overcome the limitations of LLMs such as generating misinformation, outdated knowledge, etc.

6. Metaphor API:

• Metaphor provides an API that allows developers to integrate its search functionality into their own applications or services.
• The API also provides clean, parsed HTML content without the need for web scraping.
• For individual developers, Metaphor API provides 1000 free requests per month.

In summary

Metaphor Systems stands out as a unique AI-driven search engine that employs generative AI techniques to deliver comprehensive and insightful search results. Its mission is to provide users with information comparable to what an expert friend might share, but instantly accessible through its innovative AI-powered platform.