I spent a couple of hours today and yesterday and created AINIRO Oracle, the Cut The Crap Search Engine. My motivation? I'm tired of wading through tons of garbage when I search for things.
For instance, trying to figure out what year the book 1984 was published requires reading through entire sections of the Spanish Civil War, George Orwell's birth, 11 of his 47 marriages, how they ended, and who his wifes had as their siblings - For then to skim through his claim to fame, etc. Finding simple facts like these typically takes 30 minutes. With our search engine, finding the answer takes 30 seconds.
Basically, 5 orders of magnitudes improvement compared to WikiPedia.
For crying out loud, I was looking for a simple number, the number was 1949!
Of course, the very first question in the system had to be what is 42 for obvious reasons.
This is NOT a Search Engine
Warning, this is not a search engine, it's a "question/answer machine". We've added a little bit of finesse to make that clear too. You have to finish your question with a ?. If you don't, the system will refuse to answer you. This is to separate us from other search engines where you typically phrase questions in cryptic quasi English, such as "site:stackoverflow.com cohesion encapsulation polymorphism".
This is a natural language answer machine and not a search engine!
Argument killer machine
Another one of its features I suspect will be highly useful, is as a "source provider" every time some schmuck at Facebook asks you for a source. We've even explicitly added sources at the bottom of all articles to make this clear, allowing you to phrase your question to our oracle machine, copy the URL, and paste it into some stupid Facebook debate where some schmuck asks you for a source, for then to simply move on with your life ...
Internally it's using DuckDuckGo and ChatGPT, and you will get a link to DuckDuckGo's search results, and the article will be cached. If you ask a question somebody else asked previously, or your question has a "similarity" of 90% or more, and the answer previously provided is less than 5 days old - The machine will return the previous answer. This is to reduce the token count, and conserve database space.
Psst, feel free to share whatever answers it comes up with anywhere you wish. "Re-using" answers in debates and such, was a big part of our motivation to create this.
Top comments (5)
Tried this out with two questions "How many indictments does Donald Trump Have?" and "What is the voice to parliament in Australia?". Both excellent answers, good detailed summaries, ran quickly and liked having the reference links avaialble as well as being able to go through to the duck duck go search. Great tool!
Thx Liam, we're trying to embed images inline into its result. We've got it on our chatbots, so it shouldn't be too difficult :)
Hi, I love the work you are doing, and am very thankful for your efforts. However, I am curious how your software differs from Phind, thanks!
Thank you for nice words, but I have no idea what Phind is, and I don't spend time analysing what "competitors" are doing, I do my stuff, period. If people like it, that's a bonus :)
I really like how it gives you references as well. Not only does it save you time, it will let you double check the facts before you use the info if needed.