RAG Explained

  Рет қаралды 136,679

IBM Technology

IBM Technology

Күн бұрын

Пікірлер
@RyanRosario
@RyanRosario 3 ай бұрын
The most important step here is left to the end. You can only use RAG with a transparent or locally built LLM.
@joshjones6779
@joshjones6779 13 күн бұрын
Well to be fair, you can use it, but you may get bias or inaccurate results because you don't have any control over what was used to train the LLM which may cause bias or hallucinations. So it's "recommended" that you use a transparent or locally build LLM.
@amritbro
@amritbro 8 ай бұрын
Very simple and clear explanation.. cheers to IBM
@HustlerCoder
@HustlerCoder Ай бұрын
Correction 1. Vector databases use mathematical representations like vectors, not arrays. Array-like similarity oversimplifies. 2. Transparency reduces bias but cannot entirely eliminate it without robust training data. 3. Embeddings are typically static; new embeddings are created for updated data.
@thedevmachine
@thedevmachine Ай бұрын
3, I think generating new embedding is easier then updating the old embeding. How are you going to update the old one by similarity scroe? What if a word was changed in an earlier document? Also you have the scan existing embeddings and match it with the new one is more time consuming and power.
@myidanny
@myidanny 4 ай бұрын
Thought Shawn was very flirty until I realised he didn't say "love" but "Luv"
@saminhaque2233
@saminhaque2233 4 ай бұрын
Same Here XD
@aaryaxz
@aaryaxz 7 ай бұрын
Hey! If I ask a RAG-based language model, "Tell me the features of the iPhone 17," what will it tell me? Will it say it doesn't know or will it hallucinate? I understand that once the iPhone 17 is released, the database will be updated to provide the correct information. But what happens if I ask about it before its release?
@EduardoSantAnna
@EduardoSantAnna 6 ай бұрын
I can see two scenarios here: if it is indeed RAG-based, then you have provided info about the yet-to-be-released iPhone 17. So the LLM will respond based on that. If you don't have it in your additional documents/vector DBs, then I'd recommend you to have always added something along the lines of "only answer with facts you have access to" to your system prompt and to set Temperature to a low number. (Temperature is a parameters in LLMs that defines how "creative" the model can be) Great question and it highlights the importance of having experts in GenAI guiding enterprises on how to implement this in a way that suits their use cases.
@paulaichniowski968
@paulaichniowski968 8 ай бұрын
Shawn & Luv!!!! Awesome job!!!!
@rogers2934
@rogers2934 7 күн бұрын
We don’t feed embeddings directly to an LLM; instead, we use them to retrieve similarity data. For example, if a customer searches for something, we generate an embedding for their search query and use it to find similar items from a Vector Database. The top K most relevant pieces of text are then fed to the LLM to help it make an informed decision. LLMs don’t understand embeddings directly-they work with natural language text. This is a classic case of what happens when salespeople try to explain technical concepts!
@LorenzoMarkovian
@LorenzoMarkovian 6 ай бұрын
Clear and simple, thank you guys and thank you IBM
@akshitgoel1183
@akshitgoel1183 5 ай бұрын
Hi I have few Questions , Please find time to answer 0. Are we filtering the Data in the Vector DB ? (If Yes? then ) 1. How are we filtering the relevant data from our vector DB to augment to our prompt for the LLM ? 1. 1 who is doing this process , another LLM, our own code , or some-different tool? 1.2 Are we feeding the complete data as whole to the LLM? 1.2 if we are filtering the vector data using Rule Based Mechanism then what will be the use case of LLM , how is the power of LLM being drawn if we are the one who is making the decision what to feed as a relevant data to the LLM?
@satish1012
@satish1012 5 ай бұрын
Hi This is what my understanding Storing Data as Embeddings: Correctness: Storing data (documents, images, etc.) as embeddings in a Vector DB is a valid approach. Embeddings represent the data in a high-dimensional vector space, capturing its semantic meaning. Consideration: Ensure that the embedding model you use is appropriate for your data type. For images, you might use a different model (e.g., CLIP) compared to text embeddings. Searching with Embeddings: Correctness: Converting the search query into embeddings and then comparing these embeddings with those stored in your Vector DB is correct. This allows for semantic search, which is more effective than keyword-based search. Consideration: Ensure that the conversion process and similarity calculations (e.g., cosine similarity) are implemented correctly. The returned plain text should be accurately relevant to the search query. Summarization by LLM: Correctness: Sending the retrieved plain text content to an LLM for summarization is appropriate. LLMs are designed to generate summaries and provide concise explanations based on the input text. Consideration: Ensure that the LLM is correctly configured for summarization tasks. Provide clear instructions or prompts to achieve the desired summarization quality. Returning Summarized Text to User: Correctness: Receiving the summarized text from the LLM and returning it to the user is the final step in the process. This is standard practice for providing user-friendly summaries. Consideration: Validate that the summarized content meets user expectations and provides accurate, meaningful information. 1. We Store all out relevant data in Vector DB ..like documents , images etc as part of Embeddings 2. When User Searches, it will not hit LLB directly, it will convert our search into Embeddings and return the resutlt with plain text 3. Then we send the same to text for summarization to LLM 4. Then LLM returns the summarized text back to user
@zehaankhan1
@zehaankhan1 3 ай бұрын
The guy on the left wanted to laugh out loud at that sketch 😂
@angelinagokhale9309
@angelinagokhale9309 2 ай бұрын
Very Well explained! Thank you so much.
@ShivamKumar-kx6rb
@ShivamKumar-kx6rb Ай бұрын
great quick session, thanks !!
@ChanceTEK
@ChanceTEK 21 күн бұрын
Excellent! Thank you. 🔥
@commentatorxyz5514
@commentatorxyz5514 Ай бұрын
The unhappiness in their eyes and frowns tell the pain of working for Artificial Intelligence jobs
@research2you-su9om
@research2you-su9om 8 ай бұрын
Thanks for the straight forward description of RAG.
@paoloalberti8792
@paoloalberti8792 2 ай бұрын
IMO. Data Gov Management seems to be the same as correct Database data input workflow. Prompt is a another way to query the database. LLM avoid use of query language to interrogate the DB. In this way a common people can query the DB. Appears to be a good idea to fine tuning a large LLM. But is not a fine tuned train is more similar to data embedding. In RAG how the vector database interact with LLM ? The vector database grows the LLM's latent space ? Is there a possibility that the LLM parameters can overlaps vector parameters making a mix of knowledge?
@5uryaprakashPi
@5uryaprakashPi 6 ай бұрын
So with a rag approach, can I say that we can update the original vector db with our own processed data?
@emteiks
@emteiks 8 ай бұрын
There is good point of halucinations of AI and the video unfortunately does not address it. The data governance is not addressing this issue we still can have a scenario where input is valid but output generated by AI is a garbage.
@vintastic_
@vintastic_ 8 ай бұрын
What's the solution?
@frackinfamous6126
@frackinfamous6126 8 ай бұрын
@@vintastic_you have to make sure the relevant data is going to the model. Good info into the data base is only half the battle. Semantic chunking. Size of chunks..types of search. Type of vector database used. For example PG Vector is a Postgres plugin and is not near as good at retrieval (usually) as something like pinecone
@frackinfamous6126
@frackinfamous6126 8 ай бұрын
Then the prompt used can also tremendously affect the model. You have to put it in the right context and use industry specific terms when prompting. Even a genius needs context or a bit of time to think. No matter who good the model, you have to know a bit about the specific industry to obtain great results. It’s like explains a noise to you mechanic or telling them you have a miss-fire on cylinder 1.
@miraculixxs
@miraculixxs 7 ай бұрын
​@@vintastic_ human expert review for every response
@joeillingworth1141
@joeillingworth1141 29 күн бұрын
Have it use python code to extract the info once it finds it and then having it check that against what it actually tells you and providing receipts to you for how it determined what to look at (keywords etc)
@fanzhang8823
@fanzhang8823 6 ай бұрын
1. Vector DB may or may not stored relevant information related to the question. 2. LLM may already have information or more accurate information to the question. So RAG may not always be helpful for GenAi applications
@saaien_
@saaien_ 8 күн бұрын
LLMs don't have memory.....
@SocialSketch
@SocialSketch 3 ай бұрын
Is there a pane of glass in front of them, or is this some other technology?
@thedesignbro
@thedesignbro 3 ай бұрын
yes it's a glass
@LinkingL-x4v
@LinkingL-x4v 4 ай бұрын
Clean database, stable generator and clever retriever.
@ColmFearon-dc8dr
@ColmFearon-dc8dr 6 ай бұрын
Do these guys write on the whiteboard backwards or how does that work?
@JShaker
@JShaker 6 ай бұрын
Yes they learned to write backwards for this video because it's cheaper to do that than to run the algorithm to flip a video horizontally
@mackkaputo8989
@mackkaputo8989 5 ай бұрын
@@JShaker 😂😂😂
@hathanh4650
@hathanh4650 6 ай бұрын
Thank you for your sharing! Very helpful and easy to follow. Just one question, is there anyway we can test or reinforcement train the model to make sure the outputs are appropriate?
@Thomas___018s
@Thomas___018s 5 ай бұрын
Poetic journey: the essence of refund details and expected actions
@maxcontini348
@maxcontini348 4 ай бұрын
Nice explanation. Well done boys 😁
@akhil7110
@akhil7110 7 ай бұрын
Does not address how do you validate the Q1 results returned are accurate. You should have built in a process parallel to querying the LLM of actually querying the results and training the LLM to address any discrepancies, if that is possible or correct them.
@stephaniakossman2923
@stephaniakossman2923 5 ай бұрын
Very clear explanation, thanks! How do you manage to avoid using "blackbox" models?
@DiegoSarasua-jn2wh
@DiegoSarasua-jn2wh 8 ай бұрын
Thanks guys, very clear!
@CumaliTurkmenoglu-zn7hp
@CumaliTurkmenoglu-zn7hp 6 ай бұрын
Great explanation. Thank you
@vvishnuk
@vvishnuk 8 ай бұрын
Neat and detail explanation.
@gatsby66
@gatsby66 8 ай бұрын
Nobody's ever been fired for buying RAGs from IBM.
@miraculixxs
@miraculixxs 7 ай бұрын
... yet
@dak2009
@dak2009 4 ай бұрын
Love the 1980s meme.
@complexity5545
@complexity5545 Ай бұрын
I suggest prioritizing a terminal rather than a drawing board.
@AgentBangla
@AgentBangla 8 ай бұрын
Excellent video, love it
@bangyor5949
@bangyor5949 6 ай бұрын
Do LLMs store our sensitive data when using RAG?
@satish1012
@satish1012 5 ай бұрын
No , Vector DB stores all the sensitive and confidential data Once obtained send that data to LLM to get it summarized because Vector DB would have given only the connecting data for the string
@yinlan2672
@yinlan2672 5 ай бұрын
How to do this video? screen as the board.
@bayesian7404
@bayesian7404 7 ай бұрын
Good job. I still need to learn more about data accuracy in a LLM.
@Roy-h2q
@Roy-h2q 8 ай бұрын
Interesting , thanks both
@mictow
@mictow 5 ай бұрын
Did you need to learn to write backwards for this videos?? or is there a product that help you with this nice board?
@nguyentran7068
@nguyentran7068 4 ай бұрын
You record the video then mirror the image. They simple write on the clear board
@CraigMatadeen
@CraigMatadeen 3 ай бұрын
kzbin.info/www/bejne/i6DdkKKpe9mqbJo
@DeanJohn7
@DeanJohn7 Ай бұрын
Bias within LLM's is a topic that needs more like shed on.
@mzimmerman1988
@mzimmerman1988 8 ай бұрын
nice work.
@yojackleon
@yojackleon 2 ай бұрын
Meh, they leave the biggest question unanswered, how are enterprises expected to govern the data that was used to train the LLM ?
@THEaiGAI
@THEaiGAI 8 ай бұрын
Awesome video
@satishrachaiah
@satishrachaiah Ай бұрын
They are just talking the high level concept
@nachoeigu
@nachoeigu 8 ай бұрын
Cool explanation
@siddid7620
@siddid7620 5 ай бұрын
7:10 sounds like support for open source.
@PretendCoding
@PretendCoding Ай бұрын
You'd think with this being IBM the mic would be better
@PaulMoreira
@PaulMoreira 3 ай бұрын
This was an excellent explanation! Thank you.
@ricko13
@ricko13 4 ай бұрын
good explanation
@brianvandermey4223
@brianvandermey4223 2 ай бұрын
Great backwards writing skills!
@Laura__sa8q
@Laura__sa8q 5 ай бұрын
The demanding world of refund specifics and anticipated actions
@George-j3George_372u
@George-j3George_372u 5 ай бұрын
Ironically, because transfers between banks and cards always go so smoothly, don't they? But seriously, it's all good.
@HazelChikara
@HazelChikara 3 ай бұрын
My left arm started tingling. Quite hard to concentrate now. 😅
@ammularajeshsagar7953
@ammularajeshsagar7953 8 ай бұрын
Informative
@topcommenter
@topcommenter 7 ай бұрын
🤔I think Luv saw the connection the entire time
@Ash-bc8vw
@Ash-bc8vw 4 ай бұрын
Exactly love.
@datoalavista581
@datoalavista581 8 ай бұрын
Thank for sharing !!
@hassanabida
@hassanabida 8 ай бұрын
Gotcha.
@BlueBearOne
@BlueBearOne 8 ай бұрын
And then a wide spread global epidemic crisis is brought to light wherein our gold standard "books" (peer reviewed journals) are rife with bad and corrupt data due to mismatched incentivization and misalignment of directives; and we then realize...how much good data through science do we really have? Shame we polluted the books we are supposed to be able to trust now that we have this magnificent technology here. 😭
@godlymajins
@godlymajins 8 ай бұрын
Dude, keep on topic. This isn’t the place for your grievances
@gcg8187
@gcg8187 2 ай бұрын
i know right? it's too bad we not have unbiased data to make the most of this technology :(
@BachBaldwin
@BachBaldwin 3 ай бұрын
8615 Balistreri Forge
@jonniuss
@jonniuss 8 ай бұрын
ok, gotcha 👌
@MelissaLee-k8c
@MelissaLee-k8c 3 ай бұрын
Alverta Field
@RosaSam-n3k
@RosaSam-n3k 3 ай бұрын
1411 Carson River
@VioletaEasterwood
@VioletaEasterwood 3 ай бұрын
946 Cremin Ranch
@vrohan07
@vrohan07 7 ай бұрын
kabhi haans bhi liya karo.. (Smile a bit bros)
@dheerajkumar-uk6ec
@dheerajkumar-uk6ec 6 ай бұрын
exactly love
@sk3ffingtonai
@sk3ffingtonai 8 ай бұрын
👏👏
@DonnaGonzalez-r9e
@DonnaGonzalez-r9e 3 ай бұрын
Dibbert Throughway
@TenishaBentrem-f8t
@TenishaBentrem-f8t 4 ай бұрын
Etha Road
@maliciousinferno
@maliciousinferno 7 ай бұрын
How in the world is this dude writing inverted for us too read straight lol
@inriinriinriinriinri
@inriinriinriinriinri 7 ай бұрын
they mirror the video) you can notice that most of the people writing on a glass board are left handed(in reality 90% of planet’s population is right handed), that’s also because they mirrored the video
@ObscuredByCIouds
@ObscuredByCIouds 7 ай бұрын
It's a skill only left-handed people have
@EasyCodingStudio
@EasyCodingStudio Ай бұрын
good
@markmotarker
@markmotarker 7 ай бұрын
lol. must be annoying to talk to "Luv". "Hi, Luv", "Exactly, Luv"
@KennyGennaria-d1n
@KennyGennaria-d1n 3 ай бұрын
Sipes Forks
@LutherHalifax-e4m
@LutherHalifax-e4m 3 ай бұрын
Malika Spur
@EllenJared-x3y
@EllenJared-x3y 3 ай бұрын
Dayna Course
@HowardBlair-m5d
@HowardBlair-m5d 4 ай бұрын
Tom Harbor
@SidStrong-l9t
@SidStrong-l9t 6 ай бұрын
Terrible Analogy! When a journalist is ants to do research, he goes to a library and asks the librarian?? As opposed to doing a google search ? This scenario is from last century before Luv was born ?
@WilliamsBarco-z3x
@WilliamsBarco-z3x 3 ай бұрын
Pacocha Estates
@vishwanathchintala2000
@vishwanathchintala2000 2 ай бұрын
Single take artists
@Lemonsstored
@Lemonsstored 8 ай бұрын
Boring
@snakebyteOne
@snakebyteOne 2 күн бұрын
CRINGE!!!
@FirmsJoktsate
@FirmsJoktsate 3 ай бұрын
77091 Schowalter Passage
RAG vs. Fine Tuning
8:57
IBM Technology
Рет қаралды 109 М.
What is RAG? (Retrieval Augmented Generation)
11:37
Don Woodlock
Рет қаралды 187 М.
The Best Band 😅 #toshleh #viralshort
00:11
Toshleh
Рет қаралды 22 МЛН
REAL or FAKE? #beatbox #tiktok
01:03
BeatboxJCOP
Рет қаралды 18 МЛН
We Attempted The Impossible 😱
00:54
Topper Guild
Рет қаралды 56 МЛН
What is a Vector Database?
8:12
IBM Technology
Рет қаралды 101 М.
What are AI Agents?
12:29
IBM Technology
Рет қаралды 1 МЛН
How to set up RAG - Retrieval Augmented Generation (demo)
19:52
Don Woodlock
Рет қаралды 44 М.
Transformers (how LLMs work) explained visually | DL5
27:14
3Blue1Brown
Рет қаралды 4,2 МЛН
What is Retrieval-Augmented Generation (RAG)?
6:36
IBM Technology
Рет қаралды 862 М.
I put ChatGPT on a Robot and let it explore the world
15:24
Nikodem Bartnik
Рет қаралды 1,2 МЛН
AI, Machine Learning, Deep Learning and Generative AI Explained
10:01
IBM Technology
Рет қаралды 850 М.
What is Agentic RAG?
5:42
IBM Technology
Рет қаралды 67 М.
What are Generative AI models?
8:47
IBM Technology
Рет қаралды 1 МЛН
OpenAI Embeddings and Vector Databases Crash Course
18:41
Adrian Twarog
Рет қаралды 524 М.
The Best Band 😅 #toshleh #viralshort
00:11
Toshleh
Рет қаралды 22 МЛН