I do not agree -- to me, "lying" implies intent.
These models are just trained to complete a sentence in the most plausible way, based on the millions of texts that humans have generated and blasted into the web. This does not even require consistency.
If we want consistency, we need to enforce it -- but don't blame models ...
-----Ursprüngliche Nachricht-----
Von: Code for Libraries <[log in to unmask]> Im Auftrag von CODE4LIB automatic digest system
Gesendet: Montag, 13. Mai 2024 05:00
An: [log in to unmask]
Betreff: CODE4LIB Digest - 11 May 2024 to 12 May 2024 (#2024-99)
There are 2 messages totaling 293 lines in this issue.
Topics of the day:
1. rag - retrieval-augmented generation (2)
----------------------------------------------------------------------
Date: Sun, 12 May 2024 08:41:33 -0700
From: Karen Coyle <[log in to unmask]>
Subject: Re: rag - retrieval-augmented generation
I wish we would just call it "lying". Hallucinations can be delightful, or scary, but what these do is simply lying. Hallucinations are sensory, not data driven. I think the term was used by the LLM people to make it seem less like a huge mistake and more like a twinkle in the eye.
They are LYING.
kc
On 5/8/24 10:43 AM, Lee, Seong Heon wrote:
> Hi,
>
> I agree. Hallucination seems big deal in adopting LLMs for research. I don’t see a perfect answer for this issue yet although AI engineers work hard to resolve it for future technology. I know that they use ‘temperature’ to control the degree of AI’s creativity, so that they determine how much AI responses are grounded on the user-provided documents.
>
> However with this limitation, LLMs are widely accepted as a research assistant. This is legit even though they do not guarantee 100% fact check. In my opinion, it is like that faculty get help from research assistants. Faculty is still responsibility to verify all contents of their own writing. But employing research assistants will certainly boost their work, especially in the beginning stage.
>
> Seong Heon Lee
>
> From: Code for Libraries <[log in to unmask]> on behalf of Lena
> G. Bohman <[log in to unmask]>
> Date: Wednesday, May 8, 2024 at 10:10 AM
> To: [log in to unmask] <[log in to unmask]>
> Subject: Re: [CODE4LIB] rag - retrieval-augmented generation External
> Message
>
>
> Hi all,
> I think this thread is highlighting that the main issue with using LLMs in library work is hallucinations. My impression is that at this point no one really knows how to correct that flaw, and since our work requires a high level of accuracy/truth, it really is a fatal flaw in our field.
>
> I am constantly telling researchers that they cannot use LLM for research where they cannot independently fact check the results. This makes them far less attractive to my researchers, since they really want LLMs to be able to do things they can't already do themselves...
>
> Lena
>
> Lena Bohman
> Senior Data Management and Research Impact Librarian Long Island
> Jewish - Forest Hills Liaison Donald and Barbara Zucker School of
> Medicine at Hofstra/Northwell
> [cid:cbe21533-1efd-4b3a-9506-1ed4e834a004]
> ________________________________
> From: Code for Libraries <[log in to unmask]> on behalf of
> Parthasarathi Mukhopadhyay <[log in to unmask]>
> Sent: Wednesday, May 8, 2024 12:57 PM
> To: [log in to unmask] <[log in to unmask]>
> Subject: Re: [CODE4LIB] rag - retrieval-augmented generation
>
> EXTERNAL MESSAGE
>
> Dear Eric
>
> Thanks for bringing the RAG pipeline to the attention of the
> community. I actually came to know about it from your earlier post on
> RAG dated March 1, 2024, and was trying to play with a RAG pipeline by
> using all open source tools like LlamaIndex-based PrivateGPT, Qdarnt
> as a vector database, and open source LLMs like
> mistral-7b-instruct-v0.2.Q4_K_M.gguf (as quantized GGUF formatted
> models are more friendly for a CPU-based system like my laptop), Orca, etc.
>
> Today I tried with the journal articles you referred to in your
> earlier post and using in your current system (around 135 articles,
> mainly from CRL and ITAL) to upload, index, and retrieve them in my
> local RAG pipeline. And then came a very thought-provoking post from
> Simon critically studying this new RAG system, which actually came
> into existence to reduce two big issues of LLM, like hallucinations
> and out-of-date non-contextual responses. It seems hallucination is an
> inherent feature of LLM, even when contextualized through a RAG pipeline.
>
> However, one interesting point to be mentioned here is the effect of
> prompt engineering on a RAG pipeline. When I ask the same questions as
> Simon did on the same set of documents in a similar kind of pipeline
> with prompt engineering, the result shows some differences (see
> additional system prompt in the snapshot):
>
> [image: image.png]
>
> Regards
>
> Parthasarathi
>
> Parthasarathi Mukhopadhyay
>
> Professor, Department of Library and Information Science,
>
> University of Kalyani, Kalyani - 741 235 (WB), India
>
>
> On Wed, May 8, 2024 at 10:07 PM Eric Lease Morgan <
> [log in to unmask]> wrote:
>
>> On May 8, 2024, at 11:20 AM, Simon Hunt <[log in to unmask]> wrote:
>>
>>> I thought you might be interested in a few tests I tried out- they
>>> reveal some interesting hallucinations and misalignment of
>>> expectations. Of course, I don't know the content of the 136
>>> articles you used, so this might also demonstrate how the chatbot
>>> attempts to answer questions that fall outside of scope.
>>>
>>> My input:
>>>
>>>> Please recommend three recent articles that discuss how to catalog
>> musical
>>>> scores.
>>> It confidently gave me three articles that don't exist (that is,
>>> based on searching my own library catalog and Google Scholar), from
>>> three authors that don't exist (as far as I could tell), then
>>> provided four references that have nothing to do with cataloging musical scores.
>>>
>>> In a new session, I tried a more controversial topic:
>>>
>>>> List the ways that current classification systems reflect a culture
>>>> of white supremacy
>>> The answer suggests that it self-censored due to the sensitive topic
>>> (I assume there are guardrails behind the scenes). The titles and
>> publication
>>> dates of the references, while real, suggest to me that they aren't
>> likely
>>> to contain much information on the topic of white supremacy in
>>> classification systems (though again, without knowing the sources
>>> you
>> used,
>>> they might represent the closest matches).Finally, as a follow-up in
>>> the
>> same session, I asked
>>>> What are the most recent articles on the topic of classification
>>>> and
>> white
>>>> supremacy?
>>> Like the first answer, the reply is decent, but if the articles
>> referenced
>>> below it actually discuss what the answer claims, the titles sure
>>> don't suggest it. The bot also loves the article *Cataloging Theory
>>> in Search
>> of
>>> Graph Theory and Other Ivory Towers* -- it also referenced that in a
>>> colleague's question about subject headings.
>>>
>>> In short, it seems like the effect RAG is having is to provide real
>>> articles as references, but it isn't clear how/if those articles
>>> have any content that lines up with the chatbot's output.
>>>
>>> --
>>> Simon Hunt
>>> Director, Automation, Indexing & Metadata
>>
>> Simon, thank you for the feedback, and my short reply is, "Yes!"
>>
>> There are many characteristics going into the process of indexing
>> ("vectorizing") a collection and then providing a generative-AI
>> inteface against the index. Some of them include:
>>
>> * creating a collection - What set of content is to be queried? In
>> this case, I created a collection of 136 articles on cataloging.
>>
>> * curating the collection - This mean providing some context, and I
>> provided authors, titltes, dates, and file names. Curating the
>> collection really helps when it comes to addressing questions and
>> supporting information literacy issues.
>>
>> * indexing - This is the process of vectorizing each document and
>> caching the result. This process can be accomplished through the use
>> of a model or through the use of a tradtional database. The process is not trivial.
>>
>> * prompt engineering - On the surface, these chatbots seem to take
>> anything as input, but under the hood the inputs are reformulated to
>> create "prompts". Different models use different prompts. Many of the
>> mis-steps outlined above could be avoided by better prompt engineering on my part.
>>
>> * generation - My demonstrations use a model called Llama2 to
>> formulate the response. Other models are better at generating
>> structured data like JSON, CSV, etc. Other models are better at
>> outputing software -- Python scripts. I believe the results of my
>> demonsdtration would be better if I were to use ChatGPT, but I'm
>> unwilling to spend the money; I like open source software and making
>> sure everything is computed locally, not remotely.
>>
>> Alignment? RAG works like this:
>>
>> 1. vectorize ("index") content
>> 2. get query and vectorize it too
>> 3. identify content having a similar vector as the query
>> 4. give the generating model (ex: Llama2) both the query
>> as well as the similar content to create the response,
>> and the reponse works similarly to autosuggest on your
>> telephone, but only on steroids
>>
>> Simon, many of the things you outline can be improved, and my hopes
>> is that they will be. "Software is never done, and if it were, then
>> it would be called 'hardware'." Again, thank you.
>>
>> P.S. This morning I created a different chatbot, and this time it is
>> rooted in the works of Jane Austen:
>> https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fe70
>> 53a831a40f92a86.gradio.live%2F&data=05%7C02%7Cselee%40CHAPMAN.EDU%7C9
>> ed78d2f5823491ecf4d08dc6f81c825%7C809929af2d2545bf9837089eb9cfbd01%7C
>> 0%7C0%7C638507850416777211%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMD
>> AiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C0%7C%7C%7C&sdata=S
>> QXn7bohv0yY6xE2koE%2Fkjb94lvgzovTHAD%2F%2FcXNbZQ%3D&reserved=0<https:
>> //nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fe7053a831
>> a40f92a86.gradio.live%2F&data=05%7C02%7Cselee%40CHAPMAN.EDU%7C9ed78d2
>> f5823491ecf4d08dc6f81c825%7C809929af2d2545bf9837089eb9cfbd01%7C0%7C0%
>> 7C638507850416788308%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQ
>> IjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C0%7C%7C%7C&sdata=QAjaKfQ
>> 46Oeqf5NwVhen7H3lPm2vc1XskIObmJfbNUw%3D&reserved=0><https://e7053a831
>> a40f92a86.gradio.live/>
>>
>> --
>> Eric Morgan
>> University of Notre Dame
>>
> **** CAUTION: This email originated from outside of Hofstra
> University. Do not click links or open attachments unless you
> recognize the sender and know the content is safe. ****
>
> NOTE: This email originated from outside Chapman’s network. Do not click links or open attachments unless you recognize the sender and know content is safe.
--
Karen Coyle
[log in to unmask]
http://kcoyle.net
------------------------------
Date: Sun, 12 May 2024 09:56:11 -0700
From: Karen Coyle <[log in to unmask]>
Subject: Re: rag - retrieval-augmented generation
I played around a bit with the library cataloging interface. The results remind me of a paper written by a high school student who didn't do any homework; the cites have one or two of the key words in them, but aren't relevant to the actual question. It also reminds me of this Saturday Night Live sketch in which contestants have to answer like that kind of
student:
https://www.youtube.com/watch?v=e0HGEZXTy8Y
When I asked about the concept of Work in cataloging, I got an answer about working in libraries.
I know that this is fascinating technology, and it may eventually result in useful answers, but I think the most interesting study today would be in HOW it gets things wrong. A lot of that will have to do with the fact that language is amazingly imprecise (and our brains seem to work around that).
I'm not trying to rain on anyone's parade, and experimenting with this is valuable, but as librarians I think we really need to harshly evaluate its relationship to facts.
kc
On 5/10/24 10:34 AM, Eric Lease Morgan wrote:
> Parthasarathi Mukhopadhyay <[log in to unmask]> wrote:
>
>> ...However, one interesting point to be mentioned here is the effect
>> of prompt engineering on a RAG pipeline. When I ask the same
>> questions as Simon did on the same set of documents in a similar kind
>> of pipeline with prompt engineering, the result shows some
>> differences (see additional system prompt in the snapshot):
>>
>> --
>> Parthasarathi Mukhopadhyay
>
> Yes, when it comes to generative-AI, prompt engineering is a real thing. Prompts are akin to commands given to a large-language model, and different large-language models have different prompts. The prompt I have been using in my proof-of-concept applications have this form:
>
> Context information is below.
> ---------------------
> {context_str}
> ---------------------
> Given the context information and not prior knowledge, answer the query
> Write the answer in the style of {speaker} and intended for {audience}.
> Query: {query_str}
> Answer:
>
> Where the placeholders (the things in curly braces) are replaced with values from the interface. For example, {context_str} is the content of documents pointing in the same vectored direction as the query. The {speaker} placeholder might be "a second grader", "a librarian", "a professor emeritus", etc. The same thing is true for {audience}. The value of {query_str} is whatever the user (I hate that word) entered in the interface. The prompt is where one inserts things like the results of the previous interaction, what to do if there is very little context, etc. Prompt engineering is a catch-as-catch-can. Once a prompt is completed, it given as input to the large-languagae model for processing -- text generation.
>
> Over the past week, I have created five different generative-AI chatbot interfaces, each with their own different strengths and weaknesses:
>
> * climate change - https://5c0af9ffadb4b3d2ba.gradio.live
> * library cataloging - https://6a147d360a3fc1d7df.gradio.live
> * Jane Austen - https://e7053a831a40f92a86.gradio.live
> * children's literature - https://a10e1d2687be735f40.gradio.live
> * What's Eric Reading - https://e462cd2ac6d1e35d1c.gradio.live
>
> These interfaces use a thing called Gradio (https://www.gradio.app/) for I/O, and they are supposed to last 72 hours, but all of them still seem to be active. Go figure.
>
> Finally, today I saw an announcment for a AI4LAM Zoom meeting on the topic of RAG where three different investigations will be presented:
>
> * Kristi Mukk and Matteo Cargnelutti (Harvard Library Innovation Lab), Warc-GPT
> * Daniel Hutchinson (Belmont Abbey College), the Nicolay project
> * Antoine de Sacy and Adam Faci (HumaNum Lab), the Isidore project
>
> See the Google Doc for details: https://bit.ly/3QEzQ6f
>
> --
> Eric Morgan
> University of Notre Dame
--
Karen Coyle
[log in to unmask]
http://kcoyle.net
------------------------------
End of CODE4LIB Digest - 11 May 2024 to 12 May 2024 (#2024-99)
**************************************************************
|
