Metaphor is a search engine that understands language – in the form of prompts – so you can type what you’re looking for in all the expressive and creative ways you can think of.
Search today looks largely like search 20 years ago. Search engines do a reasonable job at returning results that match the literal content of your query but often fail to understand your real goal. The effect is that even for simple questions, it’s often quite easy to get lost in a sea of irrelevant results.
The model that powers Metaphor is trained using a form of self-supervised learning, the same paradigm behind models like Stable Diffusion and GPT-3. Stable Diffusion tries to generate images based on their captions, GPT-3 tries to predict the next word based on the previous ones, and the model behind Metaphor tries to predict the next link on a webpage based on all the words that come before it. Because of this way it was trained, searching with our current model works best using prompts that resemble how someone on the internet might refer to a link.
There are no reviews yet. Be the first one to write one.