Metaphor
A new AI search engine in the form of prompts. It is based on Generative AI, using the same ideas behind DALL-E and Stable Diffusion. Watch this great video by HitPaw that explains this new innovation in search.
Search today
Search today resembles search from 20 years ago in many ways. Search engines do a decent job of matching literal results to your query, but frequently struggle to comprehend your true objective. As a result, it's frequently very simple to get lost in a sea of irrelevant results, even for simple questions.
Search engine of the future
Metaphor is a link-predicting search engine, much like how GPT-3 predicts the next word. In order to use it, you must first compose a "prompt," which is a phrase that sounds like it could be followed by a link. It's a search engine powered by generative AI, which means you can express your search in any way you like through prompts and the language you choose.
Prompt writing
Metaphor search is interacted with through the creation of "prompts," or fragments of text that may or may not precede a link. How your word search prompts can greatly affect the quality of results returned. Metaphor was taught to perform best with search queries that mimic common ways that users might refer to links.
To better understand this, picture that your ideal link has already been discovered by someone else online. What words could they use to explain that connection to others? Imagine you're writing a blog post, a tweet, or a comment on your favourite online forum; this can be a useful mental image to keep in mind when coming up with prompts.
How Metaphor is powered and used
Metaphor's underlying model is self-supervisedly trained, following the same principles as popular models such as Stable Diffusion and GPT-3. Models like Stable Diffusion's attempt to generate images based on their captions, GPT-3's attempt to predict the next word based on the previous ones, and Metaphor's attempt to predict the next link on a webpage based on all the words that came before it are all examples of this kind of technique.
Metaphor's current model is best suited for searching with prompts that are similar to how someone might refer to a link on the internet due to the way it was trained. A good search prompt might read something like this:
"My two favourite blogs are TechCrunch and"
On a webpage, you’d expect this text to be followed by a link to another blog, so it’s a good prompt!
The existing models of Metaphor are only the beginning. In the future, search engines will function more like extensions of your mind, with the ability to engage in complex interactions spanning different sensory channels and levels of reasoning to assemble fresh perspectives from the sum total of all knowledge.
Examples and best practises
Example 1
Suppose you were interested in motorcycle riding and wanted to find the personal pages of interesting people to talk to about the hobby. What might an ideal prompt on Metaphor look like?
You might think something like “Motorcycle riding personal page” or maybe something like “People interested in motorcycles” would be reasonable prompts.
If you try searching for either of these however, you’ll notice that the results aren’t very dialed in. Why are the results so poor? The keyword-like nature of the prompts is the clue. When describing links they’re about to share, people don’t tend to talk in keyword phrases. In other words, the reason for the poor results is that neither of those prompts look like how someone on the internet might actually describe a link they found.
A more optimized prompt might look something like:
"This person would be great to talk to about motorcycles (personal site here:"
This prompt has a couple of things going for it. Primarily, it truly resembles the way someone on the internet might actually talk about a really useful link they just found. We can imagine that this prompt would be just at home in the comments section of a particularly helpful subreddit or in the middle of someone’s personal blog page. These properties make it a good link.
Another great thing about this prompt is that near the end of it we specify exactly what kind of website we’d like for Metaphor to return; “personal site”. If, for example, we were searching for something where a blog post or arxiv link would be more appropriate we could specify exactly what we want by adding “blog post here” “arxiv paper:” to the end of the prompt.
Example 2
Let’s take a look at another example. Suppose you were interested in finding the best math blogs on the internet.
Maybe you start off with something like “What are the best math blogs on the internet?”
As you might have guessed this isn’t a particularly great prompt because it’s not a very natural way to precede a link. How might we improve it?
Well, we can imagine that someone searching for great math blogs might already have a great math blog in mind that they wish to find content similar to. We could try a query like
“Besides Terry Tao’s blog, this is my favorite math blog :”
Much better prompt, and much higher quality results as well.
Tips and Tricks
When possible, try to avoid keyword searches
The results tend to be less robust and useful, because they look very different from the type of ways people might talk about links on the internet. Instead of typing “Jeopardy archive” try something like “Here is the Jeopardy archive”
Rephrase questions to look more like answers.
Questions tend to make poor prompts because people sharing links on the internet don’t tend to precede them with questions. If you find yourself typing in a prompt in the form of a question, ask yourself if phrasing it as though you already found an answer would make for a more natural prompt?
For example, instead of searching for “What’s the best way to get started with cooking?”, try rephrase the question to resemble an answer;
“This is the best tutorial on how to get started with cooking:”
Use modifiers to control the type of results returned:
If you’re looking for a particular kind or style of result (say a goodreads link, or a funny post) try specifying exactly that in the prompt.
Pay attention to the punctuation at the end of prompts.
You’ll notice in many of the prior examples that the sentences ended in colons “:”. As you’ve probably guessed by now the purpose of this is to mimic the way someone sharing a link on the internet might end a sentence.
Conclusion
Link sharing is a fundamental part of the internet, and as such it’s important to be able to find links that are high quality and useful. The Metaphor search engine is designed to help with this by allowing users to specify exactly the kind of link they’re looking for through the use of natural language prompts. With a little practice, you should be able to find the perfect link for whatever you’re looking for in no time!
Please visit and explore Metaphor here - https://metaphor.systems/ and let me know what you think and discover!
留言