Use Gen AI to build Applied AI agents and systems. AI builds Applied AI while DS teams supervise and evaluate the results. Working towards open-source AGI
AI Agents Are Going To Be Ubiquitous, And You Can Build Them Using The English Language
The simple way to build AI agents is to simply specify what you want in the english language and then let the AI do the planning, coding, and building
Write a detailed 2-3 line prompt, and…
The "Office Suite of AI"
Bindu Reddy
@bindureddy
is founder/CEO of Abacus AI, a company that makes a suite of tools that you can use to turn your business into an AI-run workhorse. Here she talks to business people who are starting to implement AI, or are interested in learning…
Unless you need a very specific type of output/task, fine-tuning is typically a bad idea.
•LLM progress is very fast and there’s a new LLM released nearly every day. The chances that your fine-tuned LLM will be state of the art is very low.
•RAG and extended context is a…
The Private LLM guys made Llama-3-Smaug from Abacus available on their iPhone app.
It's a very good small model... and works pretty well on a mobile device.
Who would have thunk it? - LLMs running on your phone 🤯🤯
LLMs are already pretty intelligent; they can distinguish between strong and weak arguments, predict stances based on beliefs and demographic characteristics, and determine the appeal of an argument to an individual based on their traits.
The paper below shows that LLMs perform…
Create Complex Agents In Hours!!
You can create a customer service agent and integrate it with your customer service software like Zendesk in hours...
AI, not humans, build these agents on our platform
The average MT-Bench score of Llama3 is 8.04. The new model's score is 8.25.
That's a large jump! And it only took a few days.
The name of the model is Llama-3-Smaug-8B, and it was created by the team
@abacusai
.
Here is the model in HuggingFace:
Our AI Agents vertical is our fastest-growing vertical. It can help you crank out multiple bots to perform different tasks of varying complexity.
Some of our customers even want to make 700+ agents!
You will need support for:
•Dynamic RAG, information retrieval, and vector…
Building AI Agents
- build agents to automate simple tasks
- show ROI quickly
- in the beginning, these agents will mostly be assistants to humans
- agents that talk to internal systems, pull repeats, send emails and interface with you on slack / teams are the most popular
-…
Please Stop Saying Open Source AI Makes No Money
Takes claiming open-source AI and software are pointless because we live in a for-profit capitalist society make NO SENSE
- Meta open-sourcing Llama is a strategic masterstroke and serves its business interests. The community…
MT-Bench Scores And Why They Matter?
MT-Bench is an LLM benchmark that measures an LLM's ability to engage in coherent, informative, and engaging conversations.
It is designed to evaluate the performance of LLMs in having conversations and following instructions - attributes…
🚨New GGUF alert!
- 5x new high-quality IQ based quantized models for "Meta-Llama-3-70B-Instruct" by
@AIatMeta
- 10x new GGUF models for "Llama-3-Smaug-8B" by
@abacusai
Models are available on
@huggingface
(links are in the comments) - as always, thanks for the support ❤️
My conversation with the Wall Street Journal and the need for unbiased, liberated LLMs that don’t exercise thought control
We at Abacus AI are working on several fine tunes in the space including agentic LLMs!
Yes, they will be safe and no, they won’t have access to the…
Interesting topic
@mims
Anti-woke AIs and Agentic AI
Nice to see
@bindureddy
and
@abacusai
introduce "Liberated Qwen" that will answer any question prioritizing it's learning first
via
@WSJ
World's Best Small Model - Announcing Llama-3-Smaug-8b! - Significant MT-Bench improvement
Imagine if a llama and a dragon had a baby! - We would have an altogether better model in the form of Llama-3-Smaug-8B!! The model is available on Hugging Face NOW and significantly…