I feel like being socially adept and charismatic are different from being able to communicate clearly and effectively.
The former comes from intrinsic self-belief and acceptance, as well as genuine curiosity about the people you interact with.
Clear communication is more of a skill that you could learn from a book or guide. It depends on if you want to write or speak convincingly - writing structured arguments is definitely one way to clarify your thoughts and organize them in a logical way.
This guide from Animalz provides one great persuasive writing model: https://www.animalz.co/blog/bottom-line-up-front/
In terms of interacting with a team and leading different types of personalities, I like the business classic Five Dysfunctions of a Team: https://a.co/d/jc3UExI
It's not a complete disengagement. Your brain is a processor, just like an NLP is. It takes in various inputs and rearranges them and makes connections. Taking in the input of your knowledge base as you write is not about replacing your thinking, it's about jogging your memory, stirring up your thoughts. It's like working with an editor, or pair-programming, or even just taking a walk to let things settle.
Unlike an LLM, the brain gets better at things the more it does them, and worse the less it does them. Whatever skills we replace with NLP, we should expect to atrophy in our brains. In the case of writing, we may at first assume that those skills are simply constructing a sentence at a syntactic level, or coming up with the right words. In other words, that writing is essentially a rote, mundane task, like spell checking or long division. But we may find in practice that there are underlying analytical, creative, or reasoning skills which writing requires, and that we practice through writing, and not the behavior we replace writing with.
I wrote the article, and am personally extremely interested in this angle. What would happen to writing if we could directly transfer ideas without any kind of mediation? Would we need to "translate" at all? Would we still want to write for the beauty of it, and read for the meaning it adds to our lives?
It wasn't so much a critic of your article as thinking about coping mechanisms when the volume of articles being put out is significantly increased and nobody will be able or want to deal with all this noise. Fight GPT by GPT :)
Yep, yet another post on AI and the future of content creation.
Looking at it from a different angle - what if AI could search your internal docs, and help you problem-solve? Aka help you exploit past knowledge to inform future decisions?
The hope: GPT will democratize creation, not fill the internet with shitty articles.
Democratize creation and filling the internet with shitty articles are synonymous. Not that this is a problem or a particularly scary prospect, bookstores are already filled to the brink with garbage, it's just a game of numbers at this point.
Searching your internal docs is an interesting one, but it's still unclear what this can do that grep can't. The leap forward would be ability to reason autonomously, but we're as far from that as we've ever been.
Yep that’s the first great use case I’ve thought of - looking through all the documents in a group and answering a specific question. Also can imagine multi step pipelines driven by answers to previous questions.
Large Pharma and Life science companies generate huge amounts of documentation around change. They have a huge historical corpus of categorized structured documents that have been reviewed and approved. The quality should be good. Can definitely see a draft document using AI option in the future.
The commenter you're replying to submitted the article, I think they were trying to get people to hold off on prejudging it as Yet Another AI Submission by stating that it's actually got a slightly different message than most
It seemed rather light on details to me, like do you upload a whole bunch of documents onto the site and then it builds a model based upon that that it uses to query?
No, the basic model already exists. You could throw together something like this yourself: encode your documents with a Transformers model into a series of vectors. Then you're merely a nearest neighbor search away from finding the most semantically relevant documents. Feed those documents to GPT-3 or some such LLM as contextual state, along with the query, et voila! you have Q&A on documentation.
(Spoiler: I worked on this feature @ Slite) But in practice, the effectiveness of your pipeline will depend greatly upon exactly how you implement each piece.
Here are some of the things we've had to consider and develop, in order for the Q&A to be production worthy:
- at which level is it best to encode documents (sentence, paragraph, whole doc) and how can those tiered embeddings be weighed effectively?
- how to use the natural document hierarchy in the workspace?
- where to add NLI, so we can actually compare in-model "does this statement imply that statement" rather than just comparing vectors
- how to prioritize passages using additional fine-grained ranking against the question (cross-encoding, etc)?
- how to compute a "confidence" score so that we actually take the generative (GPT) part out of the equation and bail out early, if the documents do no contain anything which strongly implies an answer to your specific query
These are just a few of the pieces. But what we learned quickly is that solving the problem of building a great Q&A means first solving many problems that are deeply intertwined with search algorithms, natural language understanding, and other text problems in general.
Thanks for all this, most of which flew right over my head!!
I wonder, can you recommend a resource where one could get "quickly" up to speed on how this stuff conceptually works, something one could ingest in a weekend or so to get a decent handle on it (but not necessarily be capable of doing anything)?
Like, I think I have a decent understanding of the probabilistic "next word" aspect of LLM text generation, but I'm assuming there's an initial "vector query" that happens prior to this part of it?
Also, could you offer any advice on the notion of how AI/ML is able to extract various meanings from text? For example, let's say I wanted to be able to scrape this thread and then for each comment extract various ideas/topics mentioned in each comment (ideally with influence from a custom ontology), detect emotions, claims of fact, etc etc etc - is that sort of thing possible?
