I am beyond excited to publish our newest Weaviate podcast with one of the most impactful scientists in Deep Learning, !! 🎉
We discussed 's Multilingual Embeddings Model and many more topics! 📚
I really hope you enjoy the podcast:
Follow
Connor Shorten
@CShorten30
Connor Shorten’s Tweets
1
13
When we started Weaviate, we had the crazy idea to build almost everything from scratch. No vector library, no Lucene, no RocksDB, etc.
This means we were maybe slightly slower to get going, but now we can innovate very fast. Native Bitmaps is a great example of this.
Quote Tweet
Replying to @simplechris
Thank you for the kind words, Chris. This is also one of my favorite parts: The constant iteration and improvements.
If someone identifies an issue, we improve it. Because we control 100% of our architecture, constantly adapting is easy.
2
4
28
🙇🏻♂️ let's you bring in your data + let's you reason over it. We combine them and get you to a shareable demo in <1min.
👉 Demo: YouTube QA Bot (bit.ly/3DHehLD)
👉 Code: Colab (bit.ly/40tYOID)
6
23
155
Show this thread
💡To learn more about the Spark Connector from , you can read our previous blog post on how we used it to import ~1 Billion objects into Weaviate!
👉 🐙 buff.ly/3YcAyc7
👉 Read the blog post by
5
8
Congrats to and on publishing PrimeKG, multimodal knowledge graph for precision medicine in 🔬 🎉
Stay ahead of the curve with PrimeKG as it's continually updated with the latest data nature.com/articles/s4159
github.com/mims-harvard/P
Quote Tweet
With @payal_chandak and @KexinHuang5, we are excited to share PrimeKG, a precision medicine-oriented knowledge graph providing holistic and multimodal view of human disease (1/8)
biorxiv.org/content/10.110
Show this thread
1
17
54
🎉 Get your app deployed in 30 minutes or we will buy you a free coffee ☕️
❓Come hang out with us @ Another Cafe this Saturday, and get any/all of your technical questions answered.
👀 will do 200 pushups if 50 ppl show up
4
7
43
Show this thread
Replying to
👀 Incredible improvements! To be regularly publishing “minor” releases that get this level of performance gains 🥵… Great work weaviate team 🤯
2
2
5
How much do you think it costs to pre-train BERT Base on C4 to the point where it reaches an average score of 83.4 when fine-tuned on the GLUE tasks? You know what's coming soon from ...
- > $5,00026.2%
- $500-$5,00022.6%
- $100-$50020%
- < $10031.2%
932 votesFinal results
7
22
41
Show this thread
Once get #chat_LangChain by running, as a beginner to all I wish some could walk me through how the chatbot work in concept and in code.
If you wish the same, here is a newbie2newbie walkthrough just for you!
1
6
36
Show this thread
🤗 Instruct embeddings
Instruct embeddings are from an instruction-finetuned embedding model that can generate text embeddings tailored to any task, *just by providing the task description*
A joint effort by seanaedmiston and
langchain.readthedocs.io/en/latest/modu
4
8
56
Show this thread
Another big release!
🦜🔗0.0.76 main features:
🤗 Instruct embeddings (seanaedmiston, )
💢 ngram example selector ()
Other features include a new deployment template, easier way to construct LLMChain, and updates to PALChain
Lets dive in👇
2
18
101
Show this thread
Re-sharing this one, since the cropped version is a bit confusing. This one adds the required context of what we’re actually looking at.
Quote Tweet
The screenshot got cropped in the first tweet. Here is the uncropped version. ;-)
Show this thread
1
5
Exciting updates on OPT-IML (arxiv.org/abs/2212.12017)!
1. The 175B models are now available to request for research purposes. Request here: (github.com/facebookresear).
2. OPT-IML 30B and 1.3B models are now available on huggingface (huggingface.co/facebook/opt-i)!
3
54
213
Show this thread
Quote Tweet
Weaviate v1.18 (coming later this month) is the first with native support for bitmap filters.
The difference is massive! Check this comparison between v1.17 and a v1.18 preview done on the DEEP1B (9.99M objects) dataset. 
Show this thread
1
9
Amazing new speedup in filtering vector searches with symbolic filters -- congrats Etienne and team, super encouraging to see improvements like this! 🎉
Quote Tweet
Weaviate v1.18 (coming later this month) is the first with native support for bitmap filters.
The difference is massive! Check this comparison between v1.17 and a v1.18 preview done on the DEEP1B (9.99M objects) dataset.
Show this thread
2
8
Weaviate v1.18 (coming later this month) is the first with native support for bitmap filters.
The difference is massive! Check this comparison between v1.17 and a v1.18 preview done on the DEEP1B (9.99M objects) dataset. 🚀
3
14
41
Show this thread
⚡ The connector can be used with your Spark ETL processes to populate Weaviate conveniently
🎤 Check out the podcast below with Sam Bean from to learn more about the development of the connector
3
4
Show this thread
⚡ Announcing the Spark Connector for Weaviate!
🗃️ The Spark Connector allows easy importing of data from into Weaviate
👀 Check out this tutorial to get started
👉 buff.ly/3jlleLp
👉 buff.ly/3YcAyc7
More in the 🧵
1
9
16
Show this thread
#NLProc
New paper!
"In-Context Retrieval-Augmented Language Models"
From 🚀
You can get very large LM gains by concatenating retrieved documents to the LM input *without any further training*
Paper: tinyurl.com/ek93sj96
Code (tmrw): github.com/AI21Labs/in-co
🧵
4
71
233
Show this thread
Enhancing GPT-3 with world knowledge🌍:
Introducing REPLUG🔌: a retrieval-augmented LM framework that combines a frozen🧊 LM with a frozen/tunable retriever. Improving GPT-3 in language modeling & downstream tasks by prepending retrieved docs to LM inputs arxiv.org/abs/2301.12652
16
270
1,286
Show this thread
I am super super excited about this!! We are now able to create a knowledge base from podcasts that our listeners can search through! 📚
This is enabled with advances in speech-to-text transcription 💬 + Weaviate of course 😎!
I hope you find the article interesting! 🙏
Quote Tweet
Curious about what has been said on the Weaviate podcast
But don’t want to spend hundreds of hours listening?
Check this out! Search the Weaviate Podcast with Weaviate, by @CShorten30
buff.ly/3kLUp3m3
12
REPLUG: Retrieval-Augmented Black-Box Language Models
REPLUG with the tuned retriever significantly improves the performance of GPT-3 (175B) on language modeling by 6.3%, as well as the performance of Codex on five-shot MMLU by 5.1%.
arxiv.org/abs/2301.12652
7
54
249
New example: How to use to answer natural language questions, deploying it to a web endpoint: modal.com/docs/guide/ex/
0:04
10.8K views
7
22
122
OSS community and engagement around is amazing - future is bright for composable LLM apps!
📈#1: OSS MAUs since launch, keeping pace w. early growth
📈#2: weekly pip installs since launch, keeping pace with
👏 👏 👏 & community!
2
9
53
Show this thread
What if you could fit an *entire codebase* in an LLM? 🤔
"Efficiently Scaling Transformer Inference" (11/2022)
arxiv.org/pdf/2211.05102
Jeff Dean + co break out all the hacks to scale PALM-540B's context length to 43,000 tokens!
Here's how 👇
24
204
1,179
Show this thread
😄 Don't miss Weaviate Air #5!
🗓️ Next week, Wednesday, 8 February 2023, on
✨ We'll be covering a special new module, so stay tuned!
👉
2
4
One of the early voices of vector search, neural search, and more. Great to have back on the Weaviate podcast!
Quote Tweet
Curious about the latest trends in search?
@DmitryKan is here on the Weaviate Podcast to explain his Neural Search Pyramid!
Watch now to learn about everything from ANN indexing to UI and join the debate around renaming "Vector Search"
youtu.be/1ebCUCUJraEShow this thread
1
1
9
@
Dmitry Kan on Neural Search Frameworks - Weaviate Podcast #34
1
3
Quote Tweet
Seems like people are finally starting to notice @OpenAI embeddings… hopefully some really cool products will be shipped soon!
5
16
Big weekend release! 🚀🚀🚀
Too many changes to even try to list in a single tweet 😅 But as a teaser:
- Support for using GPUs with models
- A decorator to easily turn a function into a Tool
- Lots more!
🧵
1
19
140
Show this thread
This app can parse research paper PDF's 📄 and index/query them with a nice UI/UX interface!
Built with . Congrats on the app 🥳
Quote Tweet
Unlock the true potential of research with PaperBrain 
, a platform to seamlessly access and understand research papers.
With paper abstracts and ready-to-view pdfs, you'll never have to struggle with tedious downloads again
Built with a GPT assistant to help you all along!
, a platform to seamlessly access and understand research papers.
With paper abstracts and ready-to-view pdfs, you'll never have to struggle with tedious downloads again
Built with a GPT assistant to help you all along!Show this thread
1:49
7.6K views
6
20
103
Self-consistency is underrated for improving accuracy for LLMs in a range of reasoning and arithmetic tasks.
It works with any off-the-shelf LLM, eg GPT3 variants, and also provides estimates of how certain the LLM is of the provided answer.
arxiv.org/abs/2203.11171
Takeaways👇
8
57
274
Show this thread
Weaviate Podcast with and 🎙️
They discuss everything involved in neural search from data ingestion, user interfaces, and more!
2
5
Show this thread
Weekly Round-Up 🗞️
• New blog post by
• Weaviate NYC Meetup
• Weaviate Podcast #34 with and
1
4
9
Show this thread
A feature of that hasn’t been featured as much but can be a super useful tool for managing cost: token prediction! 🪙🪙
You can *predict* how many tokens each GPT Index operation (build index, query index) can consume! (within 5-10% error)
4
6
71
Show this thread
Introducing: AI Powered 3D Search 🔍
We've created the fastest search on the most expansive library of 3D assets
With almost 1 million searchable assets and an average search time of 50ms, it is easier than ever to find the perfect 3D asset for your project
Try out the magic✨
GIF
9
110
618
Show this thread
4
12
103
🤩 Discover the power of diffusion models
🖼️ These are generative models that can produce photo-realistic images
#imagegeneration #diffusionmodels #ai
1
2
5
Show this thread