Yay, I too got my 7-day suspension badge from Stack Overflow from adding an #LLM#AI disclaimer back after it was first reverted to my four (4) answers!
A study that confirms what I’ve been suspecting for a while: fine-tuning a #LLM with new knowledge increases its tendency to hallucinate.
If the new knowledge wasn’t provided in the original training set, then the model has to shift its weights from their previous optimal state to a new state that has to accommodate both the previous and new knowledge - and it may not necessarily be optimal.
Without a new validation round against the whole previous cross-validation and test sets, that’s just likely to increase the chances for the model to go off the tangent.
LLMs looting the internet will lead to a significant increase in insularity, barriers to entry, suspicion, and siloing by its users.
Expect to see an increase in invite-only forums and communities which vet everyone who applies for access to make sure they aren’t a scraper in a human suit. An increase in experts refusing to help newbies, for fear of their help being copied, mulched, and resold by massive corporations. A decrease in the “social” part of the net.
I just issued a data deletion request to #StackOverflow to erase all of the associations between my name and the questions, answers and comments I have on the platform.
One of the key ways in which #RAG works to supplement #LLMs is based on proven associations. Higher ranked Stack Overflow members' answers will carry more weight in any #LLM that is produced.
By asking for my name to be disassociated from the textual data, it removes a semantic relationship that is helpful for determining which tokens of text to use in an #LLM.
If you sell out your user base without consultation, expect a backlash.
What happens when China builds an #LLM? DeepSeek just released v2 of its model, which is open source.
I tried it on deepseek.com. Ask it about Tiananmen square, and the chatbot self-censors its answer while it is generating (that presumably is limited to their deployment). On variations not caught by the filter, it refuses -- and replies in Chinese:
"The content of your question is not in line with the core values of socialism, nor is it in line with China's laws, regulations and policies."
@mcc So developers will stop sharing information on #StackOverflow and future #Copilot and friends will be forever stuck in the past, answering questions about historically relevant frameworks and languages. #LLM#StuckOverflow
So, how can we get a proper answer? Ten years ago, when I wrote “The Milwaukee Soup App”, I used the Kimono (which is long dead) to scrape the soup of the day. You could also write a fiddly script to scrape the value manually. It turns out that there is another option, though. You could use Scrapegraph-ai. ScrapeGraphAI is a web scraping Python library that uses LLM and direct graph logic to create scraping pipelines for websites, documents, and XML files. Just say which information you want to extract and the library will do it for you.
Let’s take a look at an example. The project has an official demo where you need to provide an OpenAI API key, select a model, provide a link to scrape, and write a prompt.
As you can see, it reliably gives you the flavor of the day (in a nice JSON object). It will go even further, though because if you point it at the monthly calendar, you can ask it for the flavor of the day and soup of the day for the remainder of the month and it can do that as well.
I am running Python 3.12 on my Mac but when you run pip install scrapegraphai to install the dependencies, it throws an error. The project lists the prerequisite of Python 3.8+, so I downloaded 3.9 and installed the library into a new virtual environment.
Let’s see what the code looks like.
You will notice that just like in yesterday’s How to build a RAG system post, we are using both a main model and an embedding model.
At this point, if you want to harvest flavors of the day for each location, you can do so pretty simply. You just need to loop through each of Culver’s location websites.
Have a question, comment, etc? Please feel free to drop a comment, below.
Just got an e-mail from #storyblok informing me that one of their sub-processors (whatever that means) is now #OpenAI.
I immediately asked them if the content I created will be used in any way to train their #LLM and if so, I want no part in it. Requested a back-up of my data and to delete everything.
I used it for the first iteration of my blog but switched to #markdown a while ago.
Stack Overflow, a popular forum for programmers and software developers, announced a partnership with OpenAI earlier this week, selling the site’s data, including users’ forum posts, to train ChatGPT.
Now unhappy users are finding themselves banned for editing their popular posts in protest, and even finding those posts changed back by admin – “a reminder that anything you post on any of these platforms can and will be used for profit,” concluded one. Futurism has more.
I just found out one actually useful use case for #LLM (a.k.a. “AI”).
There is thousands of tonnes of documents from #german#nazi#regime detailing basically everything what happened, who was involved and who participated what and said what.
Feed ALL of that to an LLM and ask it “how can we avoid this happening ever again”. I have a pretty good guess my self but some people aren’t that convinced. Perhaps this is The Key. #history#sciense#neverAgain
Back in January, we started looking at AI and how to run a large language model (LLM) locally (instead of just using something like ChatGPT or Gemini). A tool like Ollama is great for building a system that uses AI without dependence on OpenAI. Today, we will look at creating a Retrieval-augmented generation (RAG) application, using Python, LangChain, Chroma DB, and Ollama. Retrieval-augmented generation is the process of optimizing the output of a large language model, so it references an authoritative knowledge base outside of its training data sources before generating a response. If you have a source of truth that isn’t in the training data, it is a good way to get the model to know about it. Let’s get started!
Your RAG will need a model (like llama3 or mistral), an embedding model (like mxbai-embed-large), and a vector database. The vector database contains relevant documentation to help the model answer specific questions better. For this demo, our vector database is going to be Chroma DB. You will need to “chunk” the text you are feeding into the database. Let’s start there.
Chunking
There are many ways of choosing the right chunk size and overlap but for this demo, I am just going to use a chunk size of 7500 characters and an overlap of 100 characters. I am also going to use LangChain‘s CharacterTextSplitter to do the chunking. It means that the last 100 characters in the value will be duplicated in the next database record.
The Vector Database
A vector database is a type of database designed to store, manage, and manipulate vector embeddings. Vector embeddings are representations of data (such as text, images, or sounds) in a high-dimensional space, where each data item is represented as a dense vector of real numbers. When you query a vector database, your query is transformed into a vector of real numbers. The database then uses this vector to perform similarity searches.
You can think of it as being like a two-dimensional chart with points on it. One of those points is your query. The rest are your database records. What are the points that are closest to the query point?
Our main model for this demo is going to be phi3. It is a 3.8B parameters model that was trained by Microsoft.
LangChain
You will notice that today’s demo is heavily using LangChain. LangChain is an open-source framework designed for developing applications that use LLMs. It provides tools and structures that enhance the customization, accuracy, and relevance of the outputs produced by these models. Developers can leverage LangChain to create new prompt chains or modify existing ones. LangChain pretty much has APIs for everything that we need to do in this app.
The Actual App
Before we start, you are going to want to pip install tiktoken langchain langchain-community langchain-core. You are also going to want to ollama pull phi3 and ollama pull nomic-embed-text. This is going to be a CLI app. You can run it from the terminal like python3 app.py "<Question Here>".
You also need a sources.txt file containing the URLs of things that you want to have in your vector database.
The May 7th event is too recent to be in the model’s training data. This makes sure that the model knows about it. You could also feed the model company policy documents, the rules to a board game, or your diary and it will magically know that information. Since you are running the model in Ollama, there is no risk of that information getting out, too. It is pretty awesome.
Have any questions, comments, etc? Feel free to drop a comment, below.
(1/2) Prompt Fuzzer - a new open-source project for LLM security 👇🏼
Prompt Fuzzer is a new open-source project that provides a set of functions for assessing the security of GenAI applications. This CLI-based tool enables you to run and test your system prompts to identify security vulnerabilities against potential dynamic LLM-based attacks.
"Microsoft and OpenAI are reportedly working on a massive datacenter to house an AI-focused supercomputer featuring millions of GPUs. The Information reports that the project could cost "in excess of $115 billion" and that the supercomputer, currently dubbed "Stargate" inside OpenAI, would be U.S.-based."
So much could be done with that amount of money... such a huge waste.