[HN Gopher] Role-playing with AI will be a powerful tool for wri...
       ___________________________________________________________________
        
       Role-playing with AI will be a powerful tool for writers and
       educators
        
       Author : benbreen
       Score  : 92 points
       Date   : 2023-12-12 14:02 UTC (8 hours ago)
        
 (HTM) web link (resobscura.substack.com)
 (TXT) w3m dump (resobscura.substack.com)
        
       | Loughla wrote:
       | The author sort of hints at this in a roundabout way, but the
       | problem for education - until the systems stop lying about things
       | they don't actually know, or using "facts" that are wildly
       | inaccurate or out-of-date, they are not reputable/reliable enough
       | to use at all.
       | 
       | Right now it's like we're in 2001-2005 for wikipedia. It's
       | probably a good tool for basic, basic, basic beginnings of your
       | work in the classroom, but you're going to get things wrong in
       | surprising ways if you use it without verifying.
        
         | pixl97 wrote:
         | Schoolbooks for example have wildly out of date facts, or are
         | brief to the point of inaccuracy. Much less watching any
         | history channel show as those tend to suck badly.
         | 
         | One of the things AI's commonly allow is for people to ask
         | questions without fear of judgement. I've not seen one scowl
         | and call a kid an idiot yet (though I'm sure someone has a
         | jailbreak for that). Having an AI trained to take these kids
         | off the wall (and often based on historical inaccuracies)
         | questions would be a really interesting tool.
         | 
         | Just not the only tool, and not really one that should be fully
         | authoritative in itself.
        
           | lobsterthief wrote:
           | Right, but when a schoolbook-writer doesn't know something,
           | they don't just make up an outrageous lie. Not saying LLMs
           | don't have a place in education, but they have a long way to
           | go.
        
             | pixl97 wrote:
             | Feynman would have disagreed with you
             | 
             | https://www.rangevoting.org/FeynTexts.html
             | 
             | >The reason was that the books were so lousy. They were
             | false. They were hurried. They would try to be rigorous,
             | but they would use examples (like automobiles in the street
             | for "sets") which were almost OK, but in which there were
             | always some subtleties. The definitions weren't accurate.
             | Everything was a little bit ambiguous - they weren't smart
             | enough to understand what was meant by "rigor." They were
             | faking it. They were teaching something they didn't
             | understand, and which was, in fact, useless, at that time,
             | for the child.
        
               | dinvlad wrote:
               | Gonna adopt this paragraph for describing LLMs
        
           | Guthur wrote:
           | Oh sure, a human might hurt your feels so you should only
           | talk to a mechanical turk.
           | 
           | We are all clearly going mad.
        
             | stuartjohnson12 wrote:
             | Oh sure, someone talking to a mechanical turk might hurt
             | your feels so you should only talk to a human.
             | 
             | We are all clearly going mad.
        
               | Guthur wrote:
               | Lol thanks I feel you helped my point
        
               | stuartjohnson12 wrote:
               | As a large language model, I cannot help points as they
               | may be harmful to humans.
        
             | pixl97 wrote:
             | Based on the number of replies you've made that have been
             | downvoted, I think maybe as a child you are one of those
             | people that would have benefitted from a mechanical turk
             | guiding you how to behave.
             | 
             | Maybe you should reflect on the abuses some humans have
             | received from other humans. For example like your^H their
             | own parents and teachers?
             | 
             | Not all of us have had great lives. People that have been
             | abused tend to abuse others unless someone/something breaks
             | that cycle (and it's almost never about pulling up your own
             | bootstraps).
        
           | kmeisthax wrote:
           | > One of the things AI's commonly allow is for people to ask
           | questions without fear of judgement.
           | 
           | Search engines have been around for almost 30 years now and
           | they do this job better than spicy autocomplete. I type
           | stupid questions into Google all the time and get good
           | answers. The "AI" version of this involves strapping a search
           | engine onto a language model, ostensibly to summarize
           | results, but in practice there are examples of the language
           | model just lying instead of doing the actual search for you.
        
             | thfuran wrote:
             | >I type stupid questions into Google all the time and get
             | good answers.
             | 
             | I type questions into Google and frequently get misleading
             | or outright incorrect answers directly in their BS
             | summaries.
        
               | jrflowers wrote:
               | > I type questions into Google and frequently get
               | misleading or outright incorrect answers directly in
               | their BS summaries.
               | 
               | This is a good point. If you treat a search engine as a
               | language model and vice versa you will run into issues.
               | These issues compound even further for users that decline
               | to scroll down and click the links that search engines
               | return
        
         | BiteCode_dev wrote:
         | Giving the load of bollocks teachers propagate themselves, I
         | suspect the first problem after automated homework will be
         | students calling out the BS. I recall the system really didn. T
         | likd it.
        
         | raincole wrote:
         | > use it without verifying
         | 
         | There isn't a single general source of truth that we can use
         | without verifying. Human teachers especially aren't such
         | sources.
        
         | stinkbutt wrote:
         | gpt4 needs better marketing. its not perfect but its without a
         | doubt the best learning tool humans have achieved. its better
         | than books, wikipedia, libraries, etc. its basically your
         | personal college professor with 24/7 office hours on
         | practically any subject. using it in combination with other
         | tools is the best approach, but this has always been the best
         | approach for every learning tool.
        
           | RecycledEle wrote:
           | As a teacher, I agree.
        
             | stinkbutt wrote:
             | if i had this when i was in school im certain i would have
             | been an A student. my biggest hangup has always been fear
             | of looking dumb, asking dumb questions, holding up the
             | class to clarify things...gpt4 eliminates all these
             | problems
        
             | staticman2 wrote:
             | So if you're teaching, I dunno, Introduction to Physics,
             | your claim is you'd rather assign students GPT 4 than a
             | physics textbook if you could only assign them one
             | educational tool? Because it's the best tool?
             | 
             | If you were teaching fourth grade math, instead of
             | assigning a workbook of math problems you'd prefer to tell
             | the kids "Ask GPT to make up math problems" because it's
             | the best tool, so if you could only pick one tool you'd go
             | with that?
             | 
             | If you were teaching history and had the choice of sending
             | kids to the university library to write research papers of
             | having them ask GPT 4 about history, you'd have them just
             | ask GPT 4 about history, because it's the best educational
             | tool?
             | 
             | Bold claim.
        
               | stinkbutt wrote:
               | no one made the claim that gpt4 is the only tool that
               | should be used
        
         | dangerwill wrote:
         | Early wikipedia was flawed but c'mon, hoping that a
         | probabilistic text generator happens to string together a
         | series of correct statements is a fundamentally unserious way
         | to gather information
        
           | danielbln wrote:
           | Did you know that those text generators can invoke tools like
           | web search, retrieval, and more to pull in external truth?
        
             | Capricorn2481 wrote:
             | Using chatGPT's web search is the worst of both worlds. You
             | get all the inefficiency of google with the potential of
             | your text generation fucking up what you searched.
        
               | hack_edu wrote:
               | ChatGPT's search is powered by Bing.
        
             | ifyoubuildit wrote:
             | This actually has made me less happy with chatgpt.
             | 
             | If I wanted to google something (or bing, whatever), I
             | would have done that. A major draw for me has been that
             | chatgpt was providing a much better experience than search
             | engines.
             | 
             | Now it sometimes feels like a fancy lmgtfy.
        
             | wolverine876 wrote:
             | > tools like web search, retrieval, and more to pull in
             | external truth
             | 
             | When did those tools start outputting truth?
        
           | olddustytrail wrote:
           | And yet it works. I asked Google Bard five multiple choice
           | questions, each with four options and it got them all
           | correct. I'm sure you can figure out the odds of that
           | happening by chance.
           | 
           | It makes no sense to say something can't work when it clearly
           | and obviously does. Most humans would do worse.
        
             | jstarfish wrote:
             | Nothing about the LLM experience is deterministic, so these
             | anecdotal experiments mean nothing. Your experiment led the
             | witness by feeding it possible answers. Small wonder it got
             | them all right. Try this: give it four _wrong_ answers for
             | each question and see how it fares. In my experience it
             | will pick one and convincingly rationalize why it 's right,
             | unless you question it, at which point you're leading it
             | again.
             | 
             | Anecdotes are so worthless, in fact, here's mine-- I asked
             | Azure's GPT for Powershell help. After seven regens in
             | which it tried to include a different fictional library, I
             | gave up. So which of us had the "real" LLM experience?
             | 
             | These things are storytellers, not teachers. Sometimes it
             | gets it right. Maybe most of the time. It's convincing
             | enough that unless you're an expert, you'll never guess
             | when it's wrong, and the lies are bespoke for every user so
             | there's never going to be an errata page to document its
             | failures. It will always _appear_ reliable.
        
               | dinvlad wrote:
               | > These things are storytellers, not teachers.
               | 
               | I really liked how a recent paper from DeepMind put it -
               | LLMs are just role-playing:
               | https://arxiv.org/abs/2305.16367. This explains so much.
        
               | olddustytrail wrote:
               | > Your experiment led the witness by feeding it possible
               | answers. Small wonder it got them all right.
               | 
               | Are you claiming you've always scored 100% on every
               | multiple choice test because you've been "fed the
               | answers"?
               | 
               | What kind of dumb response is that?
        
               | thfuran wrote:
               | Did you stop reading the comment there?
        
               | jstarfish wrote:
               | > Are you claiming you've always scored 100% on every
               | multiple choice test because you've been "fed the
               | answers"? What kind of dumb response is that?
               | 
               | In giving it answers, you gave it context to infer the
               | right one. You narrowed the search domain. It does have
               | the same effect on humans, which is why multiple choice
               | tests are easier than others-- when you show up to take
               | the test wholly unprepared, you're looking for the most
               | plausible answers based on the context you're given.
               | Fortune tellers and Clever Hans work the same through
               | interactive reactions.
               | 
               | You take to insult but I'll challenge you again to run
               | your experiment and provide only wrong answers for all of
               | the questions. Bullshit your fortune teller and see what
               | answers it comes back with in the impossible situation
               | you create.
        
               | olddustytrail wrote:
               | Fortune tellers and Clever Hans do not score perfectly on
               | multiple choice tests and neither do humans.
               | 
               | I don't have any evidence of how humans perform on
               | multiple choice tests where all the answers are incorrect
               | and they are not given that option. Do you? Or are you
               | just assuming that they would challenge the context?
        
           | airstrike wrote:
           | This is such a tiring comment. Might I suggest you actually
           | try GPT-4? It will do amazingly well at most tasks you throw
           | at it, especially if you're decent enough at the task to
           | course-correct it.
        
         | ProllyInfamous wrote:
         | >2001-2005 for wikipedia
         | 
         | I was there [still am]. Recently, I provided an update to
         | "transistor density," which was then cited by Perplexity.AI
         | when I asked about a specific new processor type (eerie, having
         | been an early adopter for both wiki and LLMs, from a user-
         | perspective).
         | 
         | I'm left wondering "how much an old wiki handle" [account]
         | might be worth, if it is so-readily cited as "leading
         | authority" (when in reality I was just a curious teenager,
         | trying to figure out what made encyclopedia "so special," when
         | wikipedia provides all these linkages FOR FREE).
         | 
         | Half a lifetime ago, and I'm still curious how this whole "open
         | source thing" is going to play out...
        
         | spullara wrote:
         | This isn't as big a problem as you seem to think it is. Even if
         | people need to verify facts maybe it will give them better
         | judgement when other humans lie to them.
        
         | sonicanatidae wrote:
         | Remember, LLMs are purely language AIs. They can and will lie
         | because "an answer " has been weighted high enough that it'll
         | do anything to satisfy that requirement. The purpose of them is
         | to create natural sounding language in response to what it
         | parses the input to be. That's it.
         | 
         | It's similar with what I call "action AIs", where an AI tries
         | to learn to walk, or race a car, optimally, through a track. It
         | will often repeat mistakes, because short term it gains a
         | higher score and takes time to learn that short term gains, in
         | some cases, harm long term gains.
         | 
         | People are using a hammer to install screws. Technically it
         | works, but that's not the droid they were really looking for.
        
       | pavel_lishin wrote:
       | AI as a tool to kickstart your own imagination is fine; but
       | despite the author's claims that he's not excited about AI-as-
       | author, it seems like they're mostly just reading fan-fiction and
       | fan-art created by an AI.
       | 
       | Even the old timey doctor sections, the author immediately admits
       | are mostly factually wrong. What good does that do?
        
         | benbreen wrote:
         | Not factually wrong at all - the dosages, ingredients,
         | diagnosis and even the language are all strikingly accurate.
         | Naturally, the "fake" 1680s doctor didn't write the same
         | prescription as the real one (Sydenham). But a different real
         | life doctor would've disagreed with Sydenham, too. In other
         | words, if you had 100 physicians in the 1680s write out a
         | prescription for hysteria and "hypochondriacal passion," this
         | would be (IMO) indistinguishable from the real ones. What's
         | different is that this is interactive, so you can change
         | elements at will. Again, this isn't reflective of historical
         | fact. It also isn't simply fan fiction. This text isn't an end
         | point, but a starting point for jumpstarting your own thinking
         | about the affordances of a past world.
         | 
         | In other words, I think this is a new method for thinking
         | creatively about history -- one among many that already
         | existed, like historical fiction, historical re-enactment,
         | various forms of experiential learning like debates and
         | roleplaying, etc. But it's cool that there's a new one!
        
       | Racing0461 wrote:
       | And video game creators. True open world games with infinite
       | choices that affect things down the line.
        
         | duskwuff wrote:
         | Definitely not -- at least, not yet. Current generation
         | language models are _spectacularly bad_ in this application. It
         | 's very difficult to get them to role-play a character in a
         | universe which differs substantially from the real world (since
         | that's where all their training data came from), and they're
         | overly credulous when responding to player input. As a result,
         | they're likely to rapidly go "off the rails" when interacting
         | with players -- they're likely to act unaware of details about
         | the world they're in, to fabricate details about that world or
         | mix in details from other fantasy universes or the real world,
         | or to allow the player to introduce incongruous elements
         | without being challenged.
        
           | Racing0461 wrote:
           | Yeah, i mean't infinite figuratively. LLMs should at least
           | enable some middle ground between selecting between A,B,C,D
           | and anything goes.
        
       | ramboldio wrote:
       | This is a cool project that implements role-playing AI:
       | 
       | https://github.com/joonspk-research/generative_agents
        
         | aaronscott wrote:
         | This is amazing, thanks for sharing it! I've been thinking
         | about building something along these lines, so it's great to
         | see a working model.
        
         | tetris11 wrote:
         | Another one (inspired by the above) that doesn't rely on OpenAI
         | servers:
         | 
         | https://github.com/a16z-infra/ai-town
        
           | airstrike wrote:
           | > Setting Up the Environment
           | 
           | > To set up your environment, you will need to generate a
           | utils.py file that contains your OpenAI API key and download
           | the necessary packages.
           | 
           | > Step 1. Generate Utils File
           | 
           | > In the reverie/backend_server folder (where reverie.py is
           | located), create a new file titled utils.py and copy and
           | paste the content below into the file:                   #
           | Copy and paste your OpenAI API Key         openai_api_key =
           | "<Your OpenAI API>"
           | 
           | I guess it still relies on OpenAI
        
             | tetris11 wrote:
             | Right, sorry I forgot to add you can override the url with
             | `OPENAI_API_BASE` and point it to a text-generation-ui
             | OpenAI API[0] compliant model.
             | 
             | 0: https://github.com/oobabooga/text-generation-
             | webui/discussio...
        
       | xena wrote:
       | Honestly as a writer I agree that it could be a useful tool, but
       | it would require a lot of UX and UI work to make it truly usable.
       | I don't know if it's worth the effort with the technology as it
       | is. Maybe in 2 months it will be.
        
       | yborg wrote:
       | At some point the Danielle Steels of the world will use models
       | trained on their body of work to generate new novels from whole
       | cloth to pump out content with little effort. The only constraint
       | would be not releasing them too quickly for human-written books.
        
         | aprilthird2021 wrote:
         | I don't think this will be a viable option. After the first 4-5
         | of books created like that they'll get stale. The AI won't get
         | any new training data (as Ms. Steel is presumably not writing a
         | book while the AI writes a book for her), and even if the
         | prompts vary, the outputs will feel more similar to each other
         | than an average author's latest 4-5 books
        
       | haolez wrote:
       | I've been a little disappointed with role playing with ChatGPT 4.
       | The AI starts out really close to the behavior that you would
       | expect for the proposed role, but as the conversation gets
       | longer, it starts to forget the role and becomes more generic,
       | like you'd expect normal ChatGPT to be.
       | 
       | I hope this is an implementation detail and that the technology
       | can, in the future, maintain the role playing across really long
       | conversations.
        
         | atemerev wrote:
         | Look at MemGPT
        
         | benbreen wrote:
         | Forgetting context is a problem, for sure. One thing I've found
         | that works fairly well is to include a request for a "status
         | bar" in your prompt. I.e. you ask it to remind itself with each
         | response 1) who it is pretending to be 2) what the date is 3)
         | what the setting is 4) what is in their NPCs "inventory" (which
         | it intuitively understands because LLMs seem to have a natural
         | affinity with MUDs). You can even have it track its mood and
         | variables like weather.
         | 
         | As the context windows of Claude/GPT-4 etc increase, I think
         | this will be less of an issue, but for now it's a pretty
         | effective workaround.
         | 
         | Here's an example of the prompts I'm using (from an activity I
         | just did with my world history class):
         | https://docs.google.com/document/d/1sLRsUVJ_KSPtjrO83ko2MSFf...
         | 
         | And my writeup of an earlier version:
         | https://resobscura.substack.com/p/simulating-history-with-ch...
        
           | jstarfish wrote:
           | A useful 1a for ChatGPT specifically is BabyAGI-style "what
           | their current/next objective is."
           | 
           | There are vector database solutions that address the
           | forgetting of context. Oobabooga has superbooga (chromadb),
           | SillyTavern requires the headache of the Extras service but
           | will let you use chromadb, Google or OpenAI. Koboldcpp
           | doesn't do vectorization, but does have a clunky
           | autogenerate-summary feature that injects a summary of events
           | so far into the prompt.
        
       | wolverine876 wrote:
       | As the pendulum has shifted way ( _way_ ) over to art as
       | commerce, as a product, then AIs are plausibly useful. Why not
       | make the 'product' more efficiently and cheaply.
       | 
       | But art for me is self-expression; I'm encountering another human
       | being or I'm expressing myself. AI creating that art (really the
       | only "art" IMHO) is a useful as an AI writing memoirs or a love
       | letter or a condolence note.
       | 
       | In general perception, art always has been some mix of those two
       | and people seem to unconsciously conflate them, to slip from one
       | to the other. If AI - plus the current socio-political madness of
       | dismissing all humanities, all compassion, and all humans as
       | anything but economic devices - effectively wipes out art, what
       | have we wrought?
       | 
       | What has the IT revolution, what has Silicon Valley wrought? Look
       | at the world we are creating. And for what? So a few people can
       | make lots of money?
        
         | aprilthird2021 wrote:
         | AI generation is like any tool. Once it's available to
         | everyone, you have to do something unique with it to have
         | something worth grabbing people's attention. If AI can
         | illustrate and flesh out the plot of a comic book on a simple
         | prompt, then we'll quickly get bored of those comic books. When
         | someone takes the energy and time to create a unique plot,
         | setting, rich characters, timely themes, etc. and then feeds
         | that to the AI, the result will be interesting to people.
         | People who put less effort in, will get bland, unsellable
         | output out of an AI
        
       | pjfin wrote:
       | I'm actually building a role play training tool at the moment -
       | https://Solidroad.com . The AI plays the part of a fake customer
       | so sales and support reps can practice before talking to real
       | customers. We've built a pretty low latency voice to voice
       | conversation simulator. It helps people build confidence on the
       | phone etc. and we've found that accuracy isn't as important in
       | this context (ie. It's ok if the AI says something weird every
       | once in a while) as long as the results are by and large
       | believable.
       | 
       | Have a demo in our website where you can try and sell a smart
       | phone to Michael Scott from the office.
        
         | pjfin wrote:
         | Demo here: https://app.solidroad.com/public/office
        
       ___________________________________________________________________
       (page generated 2023-12-12 23:01 UTC)