Comments (4)
It was a fascinating talk! The way each one was explained was clear-cut and simple, but it shows how powerful we can use AI models (hopefully open-source) to tackle otherwise challenging parsing scenarios.
Shout-out to
Task.async
, makes it look super easy to shoot a process to the AI model and not having to wait for it - dwyl/elixir-http-request-tutorial#2 (comment).
GenServer
The GenServer
approach piqued my interest, mainly because I've had the chance to see how LLMs are prompted in these applications (like through Langchain
) and how they are parsed. I've seen Rafa using Azure AI's platform to deploy a simple chatbot, like in the image below:
Adding a system message
and specifying how the bot should act is prompt engineering (even though considering this "engineering" is debatable by itself but whatever) at its core and it's funny how crucial it is to get the models to properly understand user queries and how to properly get an acceptable output (which is why things like https://github.com/guidance-ai/guidance exist).
The reason why I'm saying this is because it's interesting seeing him employ these techniques to yield correct results from the agent, which is what makes these demos successful.
And the way he's parsing the output from the model is exactly why frameworks like guidance
exist.
When adding AI to an application, the hardest part seems to be how to get a good prompt going. I've had my run-ins with https://github.com/CompVis/stable-diffusion and 99% of the work is prompting, knowing which keywords go hand-in-hand, and adding or reducing weights to some keywords.
But when using LLMs to get these simple responses and answers as shown in the video (and actually using an LLM to check the content of the answer and not having to consider all the edge scenarios when parsing user input) is THIS simple, even more so with platforms like Azure AI making it even easier for us, the hardest part is being creative with it.
Was actually curious about the
~X
sigil he implemented, was looking forward to knowing more about it :/
Generative Agents
Got SUPER excited with the third approach of self-sovereign and aware agents and how EASY it seems to make it in Elixir!
The Shinstagram
example was uber-cool. Seeing each agent interact with each other whilst having their own personalities, all in real-time, looked pretty much like real life! I would LOVE to see the code for the demo, I absolutely adored this part of the presentation.
It's so interesting to see how this approach can effect other fields. For example, games! Characters in Red Dead Redemption 2 (apparently) had their own routines hard-coded. But imagine how much more dynamic (thus boosting replayability) the world would be if each one was an LLM agent, with their own set of stories/memories, and seeing them interact with each other and the player! I know for sure this is already being implemented somehow, just can't wait to see it!
I'm already starring https://github.com/replicate/replicate-elixir!
And seeing https://replicate.com/collections/image-to-text, it's curious to see that BLIP is available (which I previously mentioned in dwyl/image-classifier#1). So this could very much be viable in imgup
(although it's paid, so it might be easier to scale it down to a local model for now).
from learn-elixir.
"Magic Box" ...
In a nutshell, this is the problem with AI. 😕
from learn-elixir.
Recommend watching. Thinking of building something similar.
from learn-elixir.
@LuchoTurtle keen to hear your thoughts/feedback. 💭
from learn-elixir.
Related Issues (20)
- HTML function component
- How to speed up `Elixir` compile times
- Bumblebee: GPT2, Stable Diffusion, and more in Elixir!
- Sheldon: The Erlang Spell Checker
- Create a Livebook sub repo? HOT 10
- About comprehensions in Elixir
- livebook on genservers HOT 2
- Simplifying Systems with Elixir • Sasa Juric • YOW! 2020 HOT 1
- livebook (in progres) on Finite State Machines HOT 1
- Block Based Editor in Phoenix LiveView
- React to LiveView for Performance
- Phoenix LiveView 0.19 + `todo_trek`
- Machine Learning with Elixir book HOT 2
- Elixir for cynical curmudgeons HOT 1
- Unpacking Elixir: Concurrency
- Secure Coding and Deployment Hardening Guidelines
- [Video] ElixirConf 2023 - José Valim - foundations of the `Elixir` type system HOT 1
- Env variables: how to? HOT 16
- How to install `Erlang` and `Elixir` on `macOS` with `M1` Chip HOT 2
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from learn-elixir.