In addition to technical skills, engage in AI communities, keep up with industry news, and prep for interviews with expert advice to master LLMs and understand AI's cultural and ethical aspects. My guide for learning to work with LLMs:
LLMS
-
Getting Started with LLMs: Essential Learning Path for Beginners
By
–
Beginners wanting to work with large language models (LLMs) should build a solid base in programming and AI using guides, then explore LLMs with Python skills through a selection of YouTube videos, podcasts, and articles tailored to various learning preferences.
-
Verbalized confidence from RLHF doesn’t reflect pre-train uncertainty
By
–
I think it’s wrong to assume that verbalized confidence from RLHF reflects pre-train uncertainty at all. E.g. the answer to “What is your gender?” (“None; I’m an AI.”) is both a priori unlikely and high-confidence compared to distribution of pre-train completions.
-
GPT-5 vs Bard Ultra: Race for AI Market Dominance
By
–
Unless GPT-5 hits the market before Bard Ultra is available early next year… (More stock is probably a good idea either way).
-
Groq LPU Inference Engine Demo at AI Summit NY Booth
By
–
Don't be afraid to use your elbows to push your way into booth #230 at #AISummit NY to try our demo of the #LPU ™ Inference Engine.
-
Google Debuts Gemini AI Model to Rival OpenAI GPT-4
By
–
Big AI news today from Google, which just debuted its new Gemini model to compete with others like OpenAI’s GPT-4. https://
blog.google/technology/ai/
google-gemini-ai/
… -
Gemini Technical Report: TPU Training, Performance vs GPT Models
By
–
Summary of Gemini's 60-page technical report. 1. Written in Jax and trained using TPUs. The architecture, while not explained in details, seems similar to Flamigo's. 2. Gemini Pro's performance is similar to GPT-3.5 and Gemini Ultra is reported to be better than GPT-4. Nano-1
-
ChatGPT Episode Naming: Evaluating AI Naming Capabilities
By
–
ChatGPT has named a lot of my episodes. It's actually not the best at it. LOL
-
Google’s AI Model Competes with GPT-4, Cuts Infrastructure Costs
By
–
Seems to solidly compete with GPT-4 on benchmarks.
Google has existing customers and surfaces to start the feedback loop, without worrying about adoption.
And Google will use TPUs for inference, so doesn't have to pay NVIDIA their 70% margins (like @OpenAI and @Microsoft has to -
Multi-modal RAG template unlocks Q&A on slide decks
By
–
⭐️ Multi-modal RAG template ⭐️
— LangChain (@LangChain) 6 décembre 2023
Slide decks are a rich sources of information, but their visual elements are inaccessible to most RAG apps.
Multi-modal LLMs, like GPT-4V, can unlock RAG on slides, enabling Q+A assistants over visual content.
We're releasing a new template to… pic.twitter.com/tLLPjwv8ZcMulti-modal RAG template Slide decks are a rich sources of information, but their visual elements are inaccessible to most RAG apps. Multi-modal LLMs, like GPT-4V, can unlock RAG on slides, enabling Q+A assistants over visual content. We're releasing a new template to