Why are vector databases so FAST?

[@farrael004]

Underfitted? More like Underrated.

[@MrTulufan]

The actual discussion about vector database starts at 14:45. Before that, it is a just a review of embeddings and RAG framework

[@ilyesbejia6566]

very usefull comment thanks

[@JHBG1971]

How do you only have 40k followers? Amazing content. Been looking for this for over a year. Thank you!

[@ah89971]

You are great. People who following you are the ones who care about understanding the root concepts which is rare to find nowadays because everyone copying and pasting without understanding

[@deroace]

so true

[@tee_iam78]

He is absolutely right. Unless you take course in vector database, it is not easy to find material on 'how vector database works at low level'. Thank you for your content.

[@PuerinTheHunter]

Hey Santiago, keep going with your choice of shirts!

[@underfitted]

That's the plan!

[@dorins3787]

Thanks!

[@underfitted]

Thanks

[@vinj98]

This video, from its content to your performance, is fantastic.

[@underfitted]

Thanks!

[@LiebsterFeind]

Wonderful video. Any chance of a video comparing HNSW vs Faiss vs Annoy?

[@toddroloff93]

Thanks for the lesson. Always good to understand how things are getting done in the background. Great Explanation!!

[@emrahe468]

@22:22 this really helps on understanding the efficiency of the vector search algorithms. and the drawing reminds me the SVM borders/boundaries. by the way, great shirt! :)

[@underfitted]

Love the shirt 👕

[@oseteg]

Thanks a lot, Santiago! You are one of two authors I follow in KZbin and mainly in LinkedIn. The content is just a gold. My question is about that serverless thing. You provide the cloud and region but don't provide your aws credentials. Does it mean that it is free? As far as I understood, the cloud provider in this case is used to store the data. What is we don't delete the database at the end? Will have the bills for storing the db?

[@justindressler5992]

Thanks you for explaining this, I had the intuition that this is how the indexing worked via clustering but you helped crystallise my thoughts on this. One thing I think might have been missed is the trigonometric functions used like cosine take into account the direction of the vector towards the next cluster. So the cosine function uses the vectors like a compass. When grouping the vectors your quantizing or approximately all related vectors to the centroid. So obviously reducing accuracy because your not pointing to the exact point in the cluster but to the centre. How are the results selected is there an attempt to research the selected related records using the original vector or is it simply random selection.

[@mahmoudelaskare4982]

great video as usual , love the energy 👏🏻❤

[@DarkRaviForDeath]

top tier content as always

[@ernestuz]

In many ways, when you calculate the embeddings, and you reduce a fragment of data to a single vector, you are calculating a kind of hash.

[@carterthaxton]

Yes, like a hash, it’s a compression and normalization of the data into a short common form. But better than a hash, because it’s comparable in multiple dimensions.

[@nachoeigu]

Thank you very much for this amazing content!! It is so educative :)

[@hasnainahmed7605]

Ahan!! How lucky these 48k, subscribers are... :) BTW, you look nice in this shirt!

[@KumR]

Nice ...Can u do one on Graph Database too?

[@michaelduffy5309]

Beautifully done.

[@uwegenosdude]

Thanks for the cool video to make me better understand this topic. If I do not want to put my data into a cloud, what other vector db could you recommend? ChromaDB?

[@vedant_stone]

There's weaviate which is also open source I believe

[@vedant_stone]

There's convex as well

[@riemannderakhshan1037]

Have nothing to tell, than You are fantastic!

[@alextiger548]

fantastic stuff. thank you so much

[@collinvelarde7473]

This was awesome. Thanks big guy.,

[@nope9310]

"ok so I'm going to execute this" "BOOM it's just that fast!" really?... really?.... You add a cut between those two sentences? I'm hoping this was unintentional. (thankfully the next search didn't have a cut) Great video otherwise. I'd love to see you dive into the actual indexing though so we can actually see how it works. This was quite high level.

[@underfitted]

Sorry, the cut was unintentional. My goal is to show to to build things, not how fast the tech is because that won’t be relevant in your own hardware.

[@kpm25]

Thanks a lot, subscribed

[@lokeshsharma4177]

Awesome as always. I live in Florida as well what are my chances to meet you in person AND how did you automate your responses to all comments you get as ♥ !!!!!! Please write something as well 🙂

[@underfitted]

No automation. The KZbin Studio app on my phone gives me the option to ❤️ replies. 😃

[@delvoneu]

Love the shirt, where did you buy it?

[@underfitted]

Can’t remember. Probably Dillard’s

[@johnmarshall4_]

Thank you for this

[@domineia]

Amazing content

[@rally_furymoments5294]

This guy is creating amazing content and subscriber is 40k??

[@underfitted]

Step by step

[@DataPains]

Awesome!

[@Vivek2062]

I didn't know Adam Sandlers is a VectorDB nerd!

[@deroace]

Wonderfull video Im trying to make an AI personality with vector databases lets hope I will get in my head an idea how to make it useing the information form the video 😅

[@shahjahanmirza1616]

Im sad that you dont have any paid course. I'd buy any of your AI course.

[@Drackomass]

I like the shirt.

[@sorin202]

You've correctly pointed out that you don't understand how OpenAI works. You're also questioning whether it's definitely powered by a quantum chip."

[@AtomicPixels]

Vector dbs have ML indexing built in ha

[@jtmuzix]

linear algebra. Orientation vs magnitude.

[@damonguzman]

You didn’t explain the answer.

[@raunaquepatra3966]

you can compress the video into 10 mins video, would be a lot better

[@underfitted]

Yup. Still learning how to do that.

[@dorins3787]

Not for a 5 years old to understand. The content is very good because it takes you from zero and grows the technical level. It is one of tge best i have found.

[@ricardomahfoud]

I disagree. To make the video 10 minutes, alot of the information will have to be either redacted or simplified. I like having longer videos that I can watch at 1.5/2x to get the best of both worlds. What makes this channelvaluable to me is the fact that it is not just a 10 minute surface explanation, but an in-depth technical explanation.

[@AgustinCaniglia1992]

Hi

[@johnini]

The shirt is okay, and the content overall is good. However, the video could have been shorter, 15 minutes. It felt too redundant, with not-so-useful examples. There was no need to include an example of Voronoi diagrams for cities. Maybe I am not the target viewer of your content. For now I will follow :)

[@underfitted]

Be honest: The shirt is awesome!

[@johnini]

​@@underfitted I love your carisma! The shirt is awesome! Still following and looking your new video!

[@blindConjecture]

This was WAY too much background context. You have to think who your audience is here. If we're interested in knowing the inner working of fast vector database lookups it's because we already know the basics like "what is a vector" and "how do you load a csv file in Python". I gave up watching after 15min because the video still hadn't even begun explaining anything about vector databases.

[@underfitted]

Thanks for the feedback!

[@nope9310]

There are other videos that explain that.

[@cherniaktamir612]

why are you angry?

[@teaman7v]

He's foreign, not angry. Common mistake.

[@underfitted]

I’m actually a very happy person.

[@cosmicaug]

@@teaman7v, you wouldn't like him when he's angry.

[@Hlfe0]

I like his style, he is passionate about what he shares 😁♥️

[@subusrable]

is he? isn't it just his way of passionately explaining?

[@timwake5830]

Two minutes no info. Done w you

[@mrsesh7364]

@jamesbriggs has great videos