imo there’s a ton of use cases i’d use an #llm for if they were cheaper. mostly text classification. top of mind: mastodon filter/mute (filter by meaning, not substring) is probably too expensive to run every message in my timeline through an llm, although maybe an embedding model might be cheap enough
The Ask the SQL DB App 🦜🔗 is a cool Streamlit application made by
Harrison Chase and it is based on LangChain and LLM. This app translates the user questions into SQL queries 👇🏼
“The assumption that cognitive or academic tests designed for humans serve as accurate measures of #LLM capability stems from a tendency to anthropomorphize models and align their evaluation with human standards,” says Shapira. “This assumption is misguided.”
New #LLM#vulnerability attack just dropped, and it’s hauntingly beautiful because it uses the same techniques used to fine tune LLMs to fine tune the attack prompt. The technique used for this attack should continue to be useful because it can be redone on any future models. The attack can also be performed on one model and applied to another related model.
Put your most realistic #LLM prompt in a safe place, and come back in 10 years. This is how real its gonna look.
You're not experiencing a revolution on par with the printing press. You're experiencing the sublime effect of being dominated by the latest iteration of the Spectacle, just as the stupefying effects of its former appearance begins to wane.
Looks like Amazon and Anthropic are following the Microsoft + OpenAI playbook of "I cant buy you, because of regulators, but you will have to rent from me"
Not a subsidiary, but a subsidiaroid
> Amazon will invest up to $4 billion in Anthropic
> As part of the investment, Amazon will take a minority stake in Anthropic.
> AWS will become Anthropic’s primary cloud provider for mission critical workloads
Anthropic released today their new LLM model - Cloude 2. According to the release notes, this model includes improvements in coding, math, and reasoning with respect to previous versions of their model.
The model was tested on the Bar exam and GRE (reading and writing) exam and scored 76.5% and 90th percentile, respectively.
Liu et al show that an #LLM fails to effectively use relevant information in the middle of its input context, and a longer context (e.g., gpt-3.5-turbo-16k) helps little. The implications are significant when using LLMs to perform tasks based on custom information (like retrieved documents) added to the prompt. It also matches my experience, that a context with concepts A and B will work well for a task involving B, but a context with A, B and C will stop working. https://arxiv.org/abs/2307.03172
Ohne Angabe von Trainingsdaten, Kalibrierungs- und Trainingsprotokoll sind "Zitate" von #LLM/#Sprachmodellen 100% beliebig. Jeder kann behaupten, was er will.
Das macht es ja so "interessant", wenn jeder und jede aktuell in Folien presst, was #ChatGPT zu weiß Gott welch abseitigem Thema konfabuliert hat.
Ihr könnt auch sagen: "Ich selbst im Übrigen stimme mir jederzeit zu" - größer ist die Wertigkeit von ChatGPT-Snippets nun wirklich nicht.
FreeCodeCamp released today another data science course focusing on LLM Engineering. This two hours course focuses on how to embed an LLM model on your own project using tools such as OpenAI, Langchain, Agents, Chroma, etc.
Automatic Generation of Visualizations and Infographics with LLMs 🚀🚀🚀
Microsoft released this cool application for creating data visualization with a language model - LIDA. The tool enables the creation of data visualization with the use of prompts. Demo available here 👇
