
AI Export- 28+ years of Dev Exp - Founder @KodeBase & @KodeAgency
United States
Joined May 2009
- Tweets14.5K
- Following530
- Followers856
- Likes120
Will (Kode) Burgess retweeted
Best resources and tools for Designers π₯π₯π₯
- Supahero.io β Hero sections
- Navbar.gallery β Navbars
- Bentogrids.com β Bento grids
- H1gallery.com β Headlines
- CTA.gallery β CTA sections
- Footer.design β Footers
- 404s.design β 404 pages
- Icoon.co β Icon and design assets
- Mocku.co β Mockup and AI Design tool
- Framer.com β Website Builder
Will (Kode) Burgess retweeted
Traditional RAG vs. HyDE, visually explained!
RAG is great, but it has a major problem:
Questions are not semantically similar to their answers.
Consider an example where you want to find context similar to "What is ML?"
It is likely that "What is AI?" will appear more similar than "Machine learning is fun."
This semantic dissimilarity leads to several irrelevant contexts being retrieved during the retrieval step.
HyDE solves this issue.
The visual below illustrates how this approach differs from RAG.
Here's how it works:
- Use an LLM to generate a hypothetical answer (H) for the query (Q). This answer doesn't have to be entirely correct.
- Embed the answer using a contriever model to obtain E. Bi-encoders trained with contrastive learning are generally used for this purpose.
- Use the embedding E to query the vector database and retrieve relevant context (C).
- Pass the hypothetical answer H, the retrieved context C, and the query Q to the LLM to produce a final answer.
Done!
Now, of course, the hypothetical generated will likely contain hallucinated details.
But this does not severely affect the performance due to the contriever.
More specifically, this model is trained using contrastive learning, and it also functions as a near-lossless compressor whose task is to filter out the hallucinated details of the fake document.
This produces a vector embedding that is expected to be more similar to the embeddings of actual documents than the question is to the real documents.
Several studies have shown that HyDE improves the retrieval performance compared to the traditional embedding model.
But this comes at the cost of increased latency and more LLM usage.
I have shared a hands-on guide to HyDE by Milvus in the replies.
π Over to you: Have you used HyDE in production yet?
____
Find me β @_avichawla
Every day, I share tutorials and insights on DS, ML, LLMs, and RAGs.
Will (Kode) Burgess retweeted
Give massive value first. Sales will follow. #ContentStrategy #ContactMarketing #InboundMarketing #ValueFirst
Will (Kode) Burgess retweeted
The first 3 months of SEO are EVERYTHING. π‘ Test, analyze, and build data-driven strategies for lasting success.
Read more π weareboldr.com/blog/why-the-β¦
#SEO #DigitalMarketing #MarketingTips
Will (Kode) Burgess retweeted
π Discover the Power of AI in SEO Research! ππ Revolutionizing keyword analysis, content optimization, and competitor analysis. π€ Stay ahead in the SEO game with AI insights! #SEOtech loom.ly/qkg8xFw
Will (Kode) Burgess retweeted
Unlock new possibilities with AI in digital marketing! π Embrace the power of artificial intelligence and take your marketing strategies to new heights. #AIinDigitalMarketing #MarketingStrategies #AItechnology loom.ly/PuuKpaU
Will (Kode) Burgess retweeted
Discover the magic of AI-driven digital marketing! π Unlock the potential of AI to boost your online presence and engage your audience. #AIforMarketing #DigitalMarketing #AIpower loom.ly/PuuKpaU
Will (Kode) Burgess retweeted
Check out our most recent #digitalmarketing Guide for #localbusinesses
#smallbusinessmarketing #hvacmarketing #localmarketing #leadgen
ow.ly/6QFR50Kba4g
Will (Kode) Burgess retweeted
Learn how you can use Google My Business to boost SEO, drive traffic, and engage with customers.
#digitalmarketing #SEO #GMB #homeservices local-agency.us/google-my-buβ¦
Will (Kode) Burgess retweeted
The Definitive Guide to Google Local Services Ads for Electricians
#localserviceads #googleads #ppc #homeservices local-agency.us/the-definitiβ¦
Will (Kode) Burgess retweeted
26 Key SEO Statistics You Need to Know for 2022 buff.ly/3ez3pT8
Will (Kode) Burgess retweeted
Do you feel that your content marketing lacks personal touch? This article brings great insight on how to fix this!
Too many content marketers use PR-centered metrics to prove the value of their content. Here are four steps to change that. contentmarketinginstitute.coβ¦
Will (Kode) Burgess retweeted
Weβre delivering new innovations that address the specific challenges and opportunities of the hybrid work world. Announcing our newest features:
βοΈ Spaces in #GoogleChat
βοΈ Companion mode in #GoogleMeet
βοΈ Client-side encryption
And more β goo.gle/3zjl0I1
Will (Kode) Burgess retweeted
Have you also noticed that people RT less than they did before?
I see plenty of comments on Tweets, but less RTs.
Will (Kode) Burgess retweeted
Have you heard the news? A Velocity #fordbronco just broke world record at @barrett_jackson auto auction.
https://teapawt.net/t.co/U0P6DeDRrT #ford #Bronco #bronco2021 #classiccar #classiccars #automotiveinnovations
Will (Kode) Burgess retweeted
I believe I've finally worked out what a huge part of 'crawl anomaly' is in GSC. It's parameters filling the wrong fields on data driven sites or data not matching other known data sources such as Wikipedia tables or Ordnance Survey. Lesson is 'make sure your data is accurate'
Will (Kode) Burgess retweeted
After yesterday's Lighthouse changes going live you probably want to know how to optimise for the new metrics which are
Largest contentful paint
Total Blocking Time / First Input Delay
Cumulative Layout Shift
Here is a thread with some quick pointers.
#sitespeed #webperf #ux
Will (Kode) Burgess retweeted
Twelve WordPress Plugins That Boost Your Content Marketing Efforts - - tinyurl.com/uwrhk4s ##plugins
























