Kern AI

If you’re not using custom chatbots for the documentation of your product, you are definitely missing out on one of the best use cases for #LLMs. Check out our latest video to see how it's done in under 20 minutes. Theory and code included. youtu.be/URYPsgVGDRc

Curious about how to use Entity RAG alongside vector databases to build highly accurate and trustworthy Large Language Models? Check this webinar we recorded recently with @MeetKern youtu.be/QUdIpXUxaAc?si…

In our latest blog post: "Data-Centric RAG" discover how a data-centric RAG approach is setting new standards for AI accuracy. ✅ Understand the data-centric RAG approach ✅ Learn how it reduces hallucinations ✅ Real-world example with ChatGPT eu1.hubs.ly/H08mL-T0

I will be talking about Entity RAG for LLMs in regulated environments together with @MeetKern on 23rd April. Want to use LLMs in a regulated environment? You need a trusted source of truth. We will build a chat bot live in the demo and show it working +/- RAG. 🤖 Link below. 👇

LLMs are a hot topic but they can be dangerous - especially in regulated environments if they are not based on a trusted source of truth. Join us and @MeetKern in a webinar on 23rd April where they will discuss how to make reliable, accurate LLMs based on Entity RAG. 👇

Transform #customersupport with Kern AI! 🚀 Instantly access company knowledge, boost your team's performance, and deliver superior customer experiences. Our #AI, built on a data-centric RAG, provides accurate, reliable help. Learn more here eu1.hubs.ly/H08mFVP0

Explore the future of AI with our exclusive interview featuring @johoetter, CEO & Co-founder of @MeetKern, at @Finovate Europe! 🎥 Watch now for a deep dive into the future of AI! 👇fintechfutures.com/2024/03/video-…

Today we’re releasing Code Llama, a large language model built on top of Llama 2, fine-tuned for coding & state-of-the-art for publicly available coding tools. Keeping with our open approach, Code Llama is publicly-available now for both research & commercial use. More ⬇️

just listed a couple of new open-source opportunities by companies like @prisma, @MeetKern, @meilisearch, or @SafeGlobal_ 👉 ossjobs.dev

Excellent comprehensive work that covers LLM security. For each security issue, they address examples, prevention methods, and attack scenarios. Check it out, there even is a shorter version as slides!

Congratulations to 8 of our #IntelLiftoff members who made it onto the German AI Startup Landscape 2023 list! Special applause to @Ask_ui, listed in the Top 20! Learn more: intel.ly/3Qa39hw @ebenbuild @motionscloud @MeetKern 🧠💡

this is wild — kNN using a gzip-based distance metric outperforms BERT and other neural methods for OOD sentence classification intuition: 2 texts similar if cat-ing one to the other barely increases gzip size no training, no tuning, no params — this is the entire algorithm:

For folks looking to learn NLP, the slides for Stanford's excellent cs224n courses are all available here (scroll to the bottom for the latest version): web.stanford.edu/class/cs224n/s…

LIMA : LLaMA 65B + 1000 supervised samples = {GPT4, Bard} level performance. From @MetaAI arxiv.org/abs/2305.11206

MMS: Massively Multilingual Speech. - Can do speech2text and text speech in 1100 languages. - Can recognize 4000 spoken languages. - Code and models available under the CC-BY-NC 4.0 license. - half the word error rate of Whisper. Code+Models: github.com/facebookresear… Paper: scontent-lga3-2.xx.fbcdn.net/v/t39.8562-6/3… Blog: ai.facebook.com/blog/multiling…

Promising. Everyone should hope that we can throw away tokenization in LLMs. Doing so naively creates (byte-level) sequences that are too long, so the devil is in the details. Tokenization means that LLMs are not actually fully end-to-end. There is a whole separate stage with its own training and inference, and additional libraries. It complicates the ingest of additional modalities. Tokenization also has many subtle sharp edges. Few examples: That "trailing whitespace" error you've potentially seen in Playground? If you end your (text completion API) prompt with space you are surprisingly creating a big domain gap, a likely source of many bugs: blog.scottlogic.com/2021/08/31/a-p… Tokenization is why GPTs are bad at a number of very simple spelling / character manipulation tasks, e.g.: x.com/npew/status/15… Tokenization creates attack surfaces, e.g. SolidGoldMagikarp, where some tokens are much more common during the training of tokenizer than they are during the training of the GPT, feeding unoptimized activations into processing at test time: lesswrong.com/posts/aPeJE8bS… The list goes on, TLDR everyone should hope that tokenization could be thrown away. Maybe even more importantly, we may find general-purpose strategies for multi-scale training in the process.

Event season continues! In May, you will find us at these events: @M3_Konferenz (Cologne, 9th - 11th) #ConversationalAIsummit by @teamrework (London, 16th - 17th) Looking forward to seeing you there!

Event season is about to start! In April you will find us at these events: #FintechWorld (Berlin, 18th - 19th) #PyCon (Berlin, 19th) #LogiMAT (Stuttgart, 25th) #InsureNXT (Cologne, 26th - 27th) Come and say hi, we'll be well-equipped with some stickers 😉

Even though Easter is over, we still have an 🥚-citing surprise for you. Did you know that there is a hidden game in our open-source tool 'refinery'? Find a place where you can execute Python code and see the magic happen! Try it out on demo.kern.ai :)

With Twitter open-source, the issues of its repository have gone wild. Check out my blog post on this topic here: dev.to/meetkern/twitt…