williamgunn, Given how many websites try to carve out a little ad revenue by basically reposting everything large publishers publish, I'm not sure this meaningfully restricts what the models can ingest, but it does meaningfully impact attribution. Do people using these models care about the quality of the sources ingested by the models? Mostly no, but this does make it harder to legitimize business use.
https://www.theguardian.com/technology/2023/aug/25/new-york-times-cnn-and-abc-block-openais-gptbot-web-crawler-from-scraping-content
#llm #llms #ai #artificialintelligence #chatgpt #openai #publishing
kellogh, 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
ErikJonker, (edited ) It's good to remember the immense datasets LLMs are trained on, a bit beyond human comprehension. For example just one (!) of those datasets is known as "Books3" , it contains the text of 183.000 books. The Atlantic wrote about it, behind a paywall sadly,
https://www.theatlantic.com/technology/archive/2023/09/books3-database-generative-ai-training-copyright-infringement/675363/
but they also made it possible to search in this dataset.
https://full-stack-search-prod.vercel.app/
#LLM #Books3 #Atlantic #GenerativeAI #AI #Books
(via JulietEMcKenna@wandering.shop )
williamgunn, Nice work characterizing a LLM-powered botnet on Twitter. LLM-generated content can't be detected as such, so detection relies on behavioral cues, like connectedness of the social graph among accounts linking to scam websites. Not going to be easy to do this at scale. https://arxiv.org/abs/2307.16336
#llm #llms #ai #artificialintelligence #chatgpt #gpt #infosec #research
br00t4c, LLM Summary of My Book Beyond Fear
https://www.schneier.com/blog/archives/2023/09/llm-summary-of-my-book-beyond-fear.html
brennansv, “Voice assistants need a shake-up.” - Jennifer Pattison Tuohy from The Verge
Alexa is going to be supported by an LLM soon. That will definitely shake things up. I’ve been waiting for this for years.
https://www.theverge.com/2023/9/20/23880764/amazon-ai-alexa-generative-llm-smart-home #Alexa #LLM
ramikrispin, 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 👇🏼https://sql-langchain.streamlit.app
Code available here ➡️: https://github.com/hwchase17/sql-qa
#llm #DataScience #dataengineering #sql #streamlit #nlp #PromptEngineering
_dm, “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.”
ErikJonker, It's fun but as any #LLM or #GenerativeAI it can't be trusted and is only good in answering superficial/general questions, ask US presidential candidates anything https://www.chat2024.com/
#Biden #Trump #Elections #AI
drahardja, 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.
Paper here: https://llm-attacks.org/zou2023universal.pdf
rml, 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.
A view from the cockpit landing in NYC in Flight Simulator 5.0
Jigsaw_You, Dutch
Barredo, 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
ErikJonker, Nice presentation on LLMs (slides and video) Making Large Language Models work for you by Simon Willison, @simon , posted on 27th August 2023
https://simonwillison.net/2023/Aug/27/wordcamp-llms/
#LLM #AI #GenerativeAI
ramikrispin, 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.
More info is available on the release notes below 👇
https://www.anthropic.com/index/claude-2Paper: https://www-files.anthropic.com/production/images/Model-Card-Claude-2.pdf
#llm #nlp #chatgpt #deeplearning #MachineLearning
PieterPeach, “The Manipulation Problem: Conversational AI as a Threat to Epistemic Agency” #LLM
dragfyre, gets an email from stackoverflow characterizing the output of their new value-add LLM feature as "trustworthy" and "accurate"
uhhhh that's a nope, chief
#LLM #AI #SALAMI #StochasticParrots #PleaseDearGodStopTheHypeCycle
philiphubbard, 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
ftranschel, German 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.
ramikrispin, Happy Friday!
New LLM Engineering Course 🚀👇🏼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.
Resources 📚
Colab notebook: https://colab.research.google.com/drive/1gi2yDvvhUwLT7c8ZEXz6Yja3cG5P2owP?usp=sharing
Code: https://github.com/pythonontheplane123/LLM_course_part_1
Video: https://www.youtube.com/watch?v=xZDB1naRUlk#llm #deeplearning #datascience #promptengineering #NLP #MachineLearning #python
ianRobinson, This Gartner blog on LLM Hallucinations (let’s skip over this euphemism for being wrong!) suggests:
“fine tune public models by training them on private data and restrict responses to those that are grounded with private verified data.”
Any organisation that puts private data into a public LLM needs an intervention 🤦🏻♂️
https://blogs.gartner.com/avivah-litan/2023/06/30/how-to-limit-genai-and-llm-hallucinations/
ricmac, In the evolving #LLM app stack, a British company called Humanloop has — perhaps accidentally — defined a new category of product: an LLM "playground". I spoke to Humanloop CEO Raza Habib, to find out how #AiEngineers are using its platform. https://thenewstack.io/a-playground-for-llm-apps-how-ai-engineers-use-humanloop/
Jigsaw_You, Dutch Interesting research…
“Study participants had a harder time recognizing disinformation if it was written by a #LLM than if it was written by a real person”
#AI #Twitter #Musk #chatgpt #openai #generativeai
https://www.theverge.com/2023/6/28/23775311/gpt-3-ai-language-models-twitter-disinformation-study
peterdrake, Here's the recording of the talk on LLMs and computer science education I mentioned earlier:
https://acm-org.zoom.us/webinar/register/WN_BxKKfwgrSrK8jnl5v-En5g#/registration
ramikrispin, 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 👇
➡ Installation:
pip install lidaResources 📚
Documenation: https://microsoft.github.io/lida/
Code: https://github.com/microsoft/lida