Breaking things down so you don't.
May 11, 2023 9:29 AM   Subscribe

goblin.tools is a collection of small, simple, single-task tools, mostly designed to help neurodivergent people with tasks they find overwhelming or difficult.

Example: Suppose you are someone who has difficulty with executive function and the task of painting an office wall has been something you've procrastinated from putting into action.
1. Write the following in the "Add new item..." field: “How do I paint the office wall?”
2. Adjust spiciness level based on the level of granularity and detail you prefer.
3. Click the magic wand button. Wait.
4. Boom. You have a detailed breakdown/list of your requested subject.
That list can be exported to iCal or as a Todoist CSV file. That CSV file can also be imported into Asana.
posted by Fizz (59 comments total) 110 users marked this as a favorite
 
This is really cool. A three-pepper "Clean kitchen" request generated 9 steps:

Remove all items from counters and table

Put away all dishes and utensils

Wipe down counters and table with cleaning solution

Sweep/vacuum the floor

Mop the floor with cleaning solution

Clean any dirty dishes or utensils in sink

Wipe down appliances and stove top

Take out the trash and recycling

Put away any miscellaneous items that don't belong in the kitchen.
posted by Iris Gambol at 9:38 AM on May 11, 2023 [1 favorite]


Looks like it's using ChatGPT on the backend. Neat! From the about page
Most tools will use AI technologies in the back-end to achieve their goals. Currently this includes OpenAI's models. As the tools and backend improve, the intent is to move to an open source alternative.

The AI models used are general purpose models, and so the accuracy of their output can vary. Nothing returned by any of the tools should be taken as a statement of truth, only guesswork. Please use your own knowledge and experience to judge whether the result you get is valid.
posted by Nelson at 9:41 AM on May 11, 2023


"Be happy" @ four peppers; 14 steps:

Think positive thoughts

Focus on things to be grateful for

Spend time with loved ones

Engage in activities that bring joy

Practice self-care and relaxation techniques

Express kindness and compassion towards yourself and others

Allow yourself to feel and process negative emotions, but don't let them consume you

Set goals and work towards achieving them

Celebrate small successes and achievements

Surround yourself with positivity and uplifting people

Maintain a healthy and balanced lifestyle

Find purpose and meaning in your life

Embrace your uniqueness and individuality

Choose happiness as a mindset and way of life.

--
+Add Subtask feature allows the user to add their own step(s) to a generated list.
posted by Iris Gambol at 9:52 AM on May 11, 2023 [4 favorites]


I tried "sell fishing tackle" which is an actual real life problem I'm dealing with right now - cataloguing and offloading the dozens of rods and sea of reels and accessories my father-in-law left behind when he passed last month. It was... not good at figuring this task out, starting me with research on my target market and fishing spots, and moving on to coming up with a brand name, website, and logo pretty quickly. At no point did it suggest that I might want to take stock of the fishing tackle that I hope to sell.

I then tried "learn to ollie" (which I guess is kind of a task I'm grappling with in real life, and it looked super promising. Genuinely impressed that it started with locate a skateboard and put on safety gear. Stand on the board. Learn to balance. Learn to back for pivot. (These are all great steps for familiarising yourself with being on a board, and you should absolutely learn to do them before even starting to learn to actually ollie, and most explanations entirely omit it.) It fell apart a bit once it got to the actual ollie though - jump on the board, then do that while popping the tail. Now continue to practice ollies and try variations!

A popped tail jump is not an ollie. You also need to level the board of by dragging your front foot along the board. It omits this crucial step entirely.

It's a neat idea, let down somewhat by ChatGPT.
posted by Dysk at 10:04 AM on May 11, 2023 [6 favorites]


Dysk, just curious, but was this at the highest level of spiciness? It probably wouldn't change the flow too much but I suspect as these systems iterate down the line, we'll see them improve but yeah, not ideal for all solutions but a good starting point.
posted by Fizz at 10:13 AM on May 11, 2023


It was broadly the same at 3 and 5 peppers, yeah. Five had more granular steps relating to website, branded promo items, logo, etc, which 3 rolled up into one or two steps. Didn't try less than that.

The formalizer is fun! Terrible at its stated purpose, but brilliant in the same way that translating phrases back and forth between lots of languages in the early days of machine translation was.
posted by Dysk at 10:15 AM on May 11, 2023 [1 favorite]


Using neurodivergent people as test subjects for an AI oracle seems weird. Or do we trust ChatGPT now?
posted by gwint at 10:18 AM on May 11, 2023 [11 favorites]


Using neurodivergent people as test subjects for an AI oracle seems weird. Or do we trust ChatGPT now?

This comment comes across a bit uncharitable.

First, what is actually meant by the term "using"?

If by "using" you actually mean "exploiting," and I don't know if you are, then we could take a look at who programmed this and who, if anyone, is or may benefit from any data gained. Did you read or observe something that documents the programmer's intentions for how this will be used? If so, kindly feel free to share.

As a neurodivergent person who already approaches AI with a healthy dash of skepticism: it seems neither fair nor accurate to make a blanket statement that this tool is "using" neurodivergent people as test subjects. I don't speak for the neurodivergent community but I wouldn't be surprised if a number of other neurodivergent-identifying folks agreed with me. Neurodivergent people have agency, too. If and when they don't have agency to make their own decisions, which certainly does happen, that's a separate conversation—certainly not the conversation we're having here.

Respectfully, your comment feels more like a disingenuous drive-by trolling to stoke animosity, than it does an attempt to meaningfully engage the very necessary conversations around ethical applications of AI.

I'm happy to reassess my response to your comment if you provide more information—not assumptions, but actual documented information—to demonstrate what, specifically, is exploitative about this particular tool.

This is a tool that anyone can use. Yes, it's most helpful to people who are neurodivergent, but people who are anxious or generally task-averse may find it helpful as well. Is the tool (or AI in general) perfect? No. And of course we should continue to discuss the ethics of AI's use and implementation. But, I don't think it's at all helpful for us to frame the conversation here in this specific way.
posted by nightrecordings at 10:45 AM on May 11, 2023 [29 favorites]


From the About page:
goblin.tools is written and maintained by Bram De Buyser, a freelance software engineer. You can find his contact information at skyhook.be. He's on twitter, reddit, and sometimes mastodon. He is also building an ad-free, tracking-free social network at drabble.social. It is now in public preview and accepting registrations.
The reddit post also has this bit about privacy:
How about privacy?

Your todo list and spiciness setting is not stored on any server, it sits in your browser's local storage, it never leaves your computer or phone. When you put in a new task, or ask to split a task with the wand button, it will send just the task's text to the backend. The backend doesn't store any of it, not even in logs, but it is passed on to OpenAI's API for now. They might store it for eventual further training. Once i switch to a self-hosted model your input won't be stored anywhere at all. I also use Google Analytics and MS Clarity to help me understand how many people are using it, what's going wrong, etc. This doesn't allow me to identify a particular user, and it's configured to not send me your text in- or output.”
I think its worth sharing in regards to some of the concerns raised up above.
posted by Fizz at 10:51 AM on May 11, 2023 [14 favorites]


Respectfully, your comment feels more like a disingenuous drive-by trolling to stoke animosity, than it does an attempt to meaningfully engage the very necessary conversations around ethical applications of AI.

I don't mean to interrupt, but if we're talking about tone, we should really ask goblintools' Judge function for a ruling:

"Based on the text, it seems likely that the author is expressing skepticism and possibly even concern about the ethics of using neurodivergent people as test subjects for AI. They may feel that it is wrong to use individuals who may have unique needs and experiences as guinea pigs for technology, and they may question the validity and reliability of an AI oracle created in this way. The author also seems to be questioning the trustworthiness of ChatGPT, the system that is presumably using these test subjects. Overall, the text reads as skeptical and potentially critical in tone."
posted by mittens at 10:58 AM on May 11, 2023 [3 favorites]


this is really fun.

i asked it how long it would take to learn the guitar parts and bass part of just like heaven by the cure.

it said 3-6 weeks! i am feeling like a virtuoso over here
posted by capnsue at 11:01 AM on May 11, 2023 [3 favorites]


Maybe it's just the way my brain is structured, but when I enter in a "holy shit, maybe I'm not up to this" task like "paint a wall" it's making me more anxious.

One of the steps is clean the wall. Clean the wall? With what? Kitchen soap? Lysol wipes? Special grade degreaser? Do I have to let it dry? For how long? Every step is that vague! And why is it telling me I need to procure painting funds first? Does it think I'm just going to walk into the store and take the supplies? I'm not about to attempt to shoplift paint cans! Stop it, you little Yoda-looking nag! I'm putting YOU in the can in two seconds etc etc.

In contrast, I can pull up a video on YouTube where the host both reassures me that anyone can paint a wall and shows the kit necessary in the first three minutes or so.

There must be a better use case for this thing...
posted by kingdead at 11:05 AM on May 11, 2023 [13 favorites]


kingdead, for Magic ToDo you can click the wand to have it break a task down to subtasks. Here's a little drill down I just did (each list item comes from me expanding the previous one into a subtask list; I only show one subtask.)
  • Paint a wall →
  • Prepare the wall by cleaning, patching holes, and sanding →
  • Gather cleaning supplies →
  • Identify necessary cleaning supplies →
  • Research and identify specific cleaning solutions for each surface →
  • Match suitable cleaning solutions with each surface →
  • Research and gather information on suitable cleaning solutions for each surface
The AI has let us down here but I think the user interface for iterating makes sense. I tried being more specific ("paint a plaster wall with acrylic paint") but then it didn't prompt me to clean it at all.

I dunno, I really like the design of these tools. They're basically just simple wrappers for ChatGPT conversations, presumably with some prompt engineering built in. But they're useful and clean and simple. It's easy to imagine this kind of thing getting significantly better, quickly.
posted by Nelson at 11:36 AM on May 11, 2023 [3 favorites]


Hey, it's okay to just not use a thing that other people might find useful.
posted by Etrigan at 12:00 PM on May 11, 2023 [31 favorites]


My upcoming task is painting my deck. I have generally been happy with YouTube DIY videos, but I thought I would try this.

The overall list of steps was reasonable. However, when breaking down "Choose a suitable primer", the first
step was "Determine the DNA sequence you want to amplify"; and continued with DNA related directions. I assume a "primer" is used in DNA amplification.

Drilling down into later steps didn't go so wrong, although at some point it "forgot" that I was painting a deck, and gave generic instructions about painting an "object".
posted by Vegiemon at 12:19 PM on May 11, 2023 [7 favorites]


It's kind of fun to keep clicking the magic wand tool, just to see how many possible steps something can get broken down into. I started with "Buy a hot pepper" and gave up after it generated about 47 steps.
posted by spilon at 12:25 PM on May 11, 2023 [3 favorites]


Formalize: You better check yourself before you wreck yourself

Output1: It would be advisable to introspect and reevaluate your actions before they lead to negative consequences.

Output2: It is advised that you conduct a self-evaluation prior to engaging in any activity or behavior that could potentially lead to negative consequences.

Output2: It would be advisable for you to examine your behavior before causing any potential harm to yourself or others.


Interestingly the output is not repeatable.
posted by euphorb at 12:30 PM on May 11, 2023


Yeah, apologies for coming off snarky (I obviously did not run my comment through the Formalizer->Less Emotional filter) I'm just a little shocked that all those worries about ChatGPT and the like having inherent biases, hallucinating, etc. and the privacy concerns about OpenAI and other big tech companies retaining user data aren't even part of the conversation anymore? It's just "here's a useful tool"? I mean, I get it, there's 1,000 startups out there trying to use LLMs to build all sorts of products. Google just today announced LLM integration into pretty much all of their software. I use similar tools all the time! But we're hooking up all of these tools to big black boxes. That's cause for concern, even for well meaning projects like this one.

Or maybe I just woke up on the wrong side of the bed this morning. I just plugged in "Get in a better mood" to Magic ToDo and the advice seems... pretty good.
posted by gwint at 12:34 PM on May 11, 2023 [2 favorites]


Ok, I typed "Bake an apple pie from scratch" and I'm going to keep clicking until I reach "Invent the universe"
posted by gwint at 12:38 PM on May 11, 2023 [21 favorites]


Me: "Count to one million"

Estimator: 2-3 weeks

Me: "Count to one million at a rate of one number per minute"

Estimator: 16667-20000 minutes

Me: "Crawl to Jupiter"

Estimator: "It is impossible for a human to crawl to Jupiter."

Me: "Crawl to Jupiter, North Carolina"

Estimator: "Impossible"
posted by RonButNotStupid at 1:07 PM on May 11, 2023 [2 favorites]


Estimator: "And don't call me North Carolina."
posted by box at 1:31 PM on May 11, 2023 [32 favorites]


Some of the advice is only so helpful. I asked it how to "look beautiful" because I was afraid of what it would tell me to do and yup it's not helpful:

Wash your face
Apply moisturizer
Apply foundation
Apply concealer under eyes
Apply powder to set foundation and concealer
Fill in eyebrows
Apply eyeshadow
Apply eyeliner
Apply mascara
Apply blush
Apply highlighter
Apply lipstick or lip gloss


I don't wear makeup or use moisturizer. So I guess I don't look beautiful because I have the gall to show my bare-ass face to the world.

Boys and girls of all ages: if you want to look beautiful you should obviously wear a LOT of makeup. Ugggh.
posted by RobinofFrocksley at 1:33 PM on May 11, 2023 [7 favorites]


I note that "become the world champion crawfish eater" did not include "cook crawfish" in its list of subtasks.
posted by credulous at 1:40 PM on May 11, 2023 [1 favorite]


take over the world

Determine a clear goal and vision for the takeover
Assess the current state of global politics and military power
Identify potential allies and enemies
Plan and execute propaganda campaigns to gain support
Strategize and carry out covert operations to weaken opponents
Build a powerful and loyal army
Develop advanced technology and weaponry
Coordinate and execute surprise attacks on key targets
Establish and enforce a new global political order
Implement policies and laws to maintain control
Suppress any dissent or rebellion
Maintain a strong military presence to prevent opposition
Expand the empire through colonization or annexation
Cultivate close relationships with other nations and world leaders
Constantly monitor and adjust tactics to maintain control.
posted by Ickster at 1:41 PM on May 11, 2023 [3 favorites]


Flip a coin to heads 100 times in a row without flipping tails even once.:
Flip a coin
If it lands on tails, flip again
If it lands on heads, continue
Repeat until it lands on heads 100 times in a row

It's not wrong. Apparently could take as little as 31 years.
posted by HeroZero at 2:00 PM on May 11, 2023 [3 favorites]


I appreciate the the sentiment behind this and it's fun to play with, but ChatGPT's instructional ability is just lightyears ahead in detail and neurodivergent or not, you can't sort out tasks if you're not getting all the steps. Neurodivergent people deserve correct information!

If there was some way to merge the checklist functionality here with the ChatGPT set, I'd love it. Sadly, ol Chatty can create checkboxes, but they can't be ticked off or copied to a document.
posted by kingdead at 2:15 PM on May 11, 2023


My input: What the fuck, man?

Make my text: More formal

Formalizer output: Excuse me, may I inquire as to the reasoning behind your actions or statement?
posted by ZenMasterThis at 2:25 PM on May 11, 2023 [18 favorites]


wow this is rad and i'm going to show it to my ND daughter. what a neat project.
posted by Sebmojo at 3:02 PM on May 11, 2023 [3 favorites]


Input: I can't believe it's not butter.
Formalizer output: I find it difficult to comprehend that the substance before me is not genuine butter.
posted by FissionChips at 4:34 PM on May 11, 2023 [5 favorites]


Input: metafilter

Open a web browser
Type in "metafilter.com" in the address bar
Press enter or click on the search button
Wait for the website to load
Scroll down and explore the homepage
Click on any category or topic that interests you
Read articles or posts related to the chosen category
Leave comments or participate in discussions if desired
Use the search bar to look for specific content
Sign up for a Metafilter account if interested in posting content
Follow Metafilter social media accounts for updates and news
Bookmark the website for future visits.
posted by Hume at 4:46 PM on May 11, 2023 [1 favorite]


Pepsi is the preferred option of a fresh cohort of consumers.
The flavor is exquisite while the calorie content is reduced.
The concept of perpetual longevity is commonly associated with diamonds.
Rest assured that you will be well taken care of when you choose Allstate as your insurance provider.
The United States economy is significantly fueled by Dunkin', a popular coffee chain.
The morning repast of individuals who are recognized for their exceptional skills or prowess.
The milk chocolate confectionery which has a propensity to dissolve in the mouth rather than the hand.
Perhaps her innate qualities are responsible for it. Alternatively, it may be due to the usage of Maybelline products.
The act of kissing is often associated with the purchase of jewelry from the renowned retailer Kay Jewelers, as their marketing campaign emphasizes that the beginning of any romantic relationship can be traced back to a gift from their establishment.
There exist certain intangible entities that remain beyond the purview of monetary transactions. However, for all other materialistic requirements, the usage of MasterCard is deemed to be effective.

May I inquire as to who released the canines?
posted by HeroZero at 5:39 PM on May 11, 2023 [2 favorites]


Hmm...

"How long to complete" the graph traversal thing I've pounded my head on for years (very sparingly, admittedly) but far more then stated 16-32 hours.

Thanks for uh. The vote of confidence.
posted by symbioid at 5:55 PM on May 11, 2023


I have reached my limit with the presence of these venomous reptiles aboard this aircraft.
posted by credulous at 5:57 PM on May 11, 2023 [2 favorites]


this post tricked me into using AI and i find that repugnant since i have hertofore avoided it vigorously and vehemently. i feel kind of sick. i can never get this back.

please consider when posting AI tools in the future that you explicitly mention that you are linking to some AI-backed frontend, as some people find such things extremely offensive to interact with
posted by glonous keming at 6:32 PM on May 11, 2023 [5 favorites]


I'm just a little shocked that all those worries about ChatGPT and the like having inherent biases, hallucinating, etc. and the privacy concerns about OpenAI and other big tech companies retaining user data aren't even part of the conversation anymore?

AI is a mixed bag and we all have super complicated feelings about it.

It's like if knives were just invented, some people are saying "clearly this is dangerous, we will cut ourselves, we all agree to ban knives right?"

And other people out there are wiggling all sorts of different shaped knives around different types of materials and items just to see if they can find a cool use for the tool.

And with the rate of speed things are going, personally I feel that within ten years, these tools will have really changed how we do both good and bad things.
posted by rebent at 6:44 PM on May 11, 2023 [4 favorites]


Formalize: You better check yourself before you wreck yourself

Jacketed buckshot ammunition is indeed disadvantageous to an individual’s bodily fidelity.
posted by slogger at 7:31 PM on May 11, 2023


Is it just me, or is the emphasis on framing these tools for a neurodivergent audience really obviously "because neurodivergent people often have difficulty with executive function, and lists like this can help with decision-making and translating motivation into action" rather than an attempt to exploit or otherwise stereotype neurodivergent people. This sort of thing is really popular in communities of neurodivergent people, so of course it would come up in the "genre" of the toolset.

Bear in mind that people often have different specific difficulties with aspects of executive function. I generally struggle with starting movements, transitioning between tasks, and getting overwhelmed by complex projects without a clear starting point. On the other hand, my spouse is much more easily overwhelmed by system complexity than I am and struggles to sort through too many choices at once.
posted by sciatrix at 8:01 PM on May 11, 2023 [12 favorites]


I can use the Formalizer to be more sociable. That can actually be a big help.
posted by NotLost at 8:12 PM on May 11, 2023 [1 favorite]


Wow, this is very cool!
posted by aurynn at 10:35 PM on May 11, 2023 [1 favorite]


The Judge finds Admiral Kirk's Star Trek III monologue to Christopher Lloyd after blowing up the Enterprise and an entire crew of Klingon troops "unapologetic and callous."
posted by johngoren at 12:26 AM on May 12, 2023 [1 favorite]


Input: My client is an asshole.

Make my text more: Formal

Output: Unfortunately, it is not appropriate to use such harsh language to refer to a client. I suggest refraining from making such derogatory statements and to instead approach the situation with professionalism and tact.
posted by idlethink at 3:21 AM on May 12, 2023


Hey, guys! How about we make the most of our weekly free threads by trying out this cool new idea? We could use the time to experiment with passing comments through the fomalizer filter and see how it works out! What do you all think? :)

Let's pour our hearts into making a change! Every week, we have a chance to try something new and innovative, so why not use that opportunity to test out a brand new way of filter-automating comments? The formalizer filter could be the solution we've been searching for to finally streamline the commenting process and bring out the best in each and every conversation. Let's join together and channel our passion into making this idea a reality!
posted by rebent at 5:16 AM on May 12, 2023 [1 favorite]


Rebent, I see what you did there.
posted by AlSweigart at 5:52 AM on May 12, 2023


The magic to do seems like a thing I could find super helpful, actually. If there was an easy way to export the lists to habitica I would be all over it.
posted by Adridne at 6:04 AM on May 12, 2023


Okay, this is an interesting use of this stuff. Potentially useful, though, so I'm going to send this to some of my neurodivergent friends.
posted by rmd1023 at 6:15 AM on May 12, 2023 [1 favorite]


How to overcome sentimental attachments to objects when decluttering?

Acknowledge and accept that it can be difficult to let go of sentimental items
Set a goal or intention for decluttering, such as creating more space or reducing stress
Start with less emotionally charged items, like clothing or kitchen items
Sort sentimental items into categories, such as photographs or mementos
Ask yourself why each item is important and if it truly brings you joy or serves a purpose in your life
Take a break if needed and revisit sentimental items later with a fresh perspective
Consider taking photos or creating a memory box of sentimental items to keep the memories without the physical clutter
Donate or gift items to someone who will appreciate and use them
Remember that letting go of physical possessions does not diminish the memories or love associated with them.
posted by carmicha at 6:50 AM on May 12, 2023


The to-do list magifier felt, well, magic... until I drilled down a little bit.

I tried "do homework." The first level of complexity is not too far off. Where it loses me is the expansion of individual subtasks. It turns out that there is no bottom. If I try to magify "gather materials," for instance, it will expand that into a list of things that look an awful lot like the main list. You can repeat this in a vaguely fractal way. It's gathering materials all the way down.

This reflects other experiences I've seen with this type of AI. It reminds me of ticky-tacky houses built quickly in a sudden economic boom. Impressive facades, less impressive bones.
posted by eirias at 8:28 AM on May 12, 2023 [6 favorites]


Is there a prize for totally breaking it on my first go?

I had to read a few comments to sort out how to use the tool- pretty neato once I scaled back and reformatted.
posted by zenon at 11:35 AM on May 12, 2023


Indulging in the creamy, luscious taste of paste is like experiencing a divine sensation that envelops your senses and leaves you feeling utterly satisfied. The rich, velvety texture of this delectable delight caresses your taste buds, and the burst of savory flavors tantalizes your palate in a way that is simply irresistible. There's just something unequivocally magical about the way paste seduces you with its enchanting taste, leaving you hankering for more and more. It's not just a culinary experience; it's an emotional journey that takes you on a culinary escape, and before you know it, you're lost in a world of pure ecstasy. Paste doesn't just taste good; it's a culinary revelation that speaks to your soul and satisfies your every craving.
Well alright then.
posted by flabdablet at 12:55 PM on May 12, 2023 [1 favorite]


Curious as to how it would do with some vaguer language, I plugged Faulkner's famous "Memory believes before knowing remembers. Believes longer than recollects, longer than knowing even wonders" into the Formalizer and told it to waffle it and it came up with this:

Hey there! Did you know that our memories often believe things before we even consciously remember them? It's true! And sometimes those beliefs stick around even longer than our conscious recollections. Crazy, right? It's almost like our memories have a life of their own. 😊

Which is actually... not a terrible interpretation of the passage. Huh.
posted by lorddimwit at 4:49 PM on May 12, 2023 [2 favorites]


I stayed up way too long last night to feed it bits and pieces of old letters and drafts for unpublished stories to make the "Judge"-tool say nice things about my writing, how it's thoughtful, and poignant, nuanced, emotionally layered, tense, suspensful, dramatic, dynamic, engaging, "a rich, intriguing portrayal of a complex character", how a text has "moments of humor, romance, tension, and intrigue, creating a complex emotional landscape for the reader to navigate".

Then I had Chat-GPT write an emotionally charged scene between a couple, reflecting frustration, disappointment, regret, but also moments of intimacy and vulnerability. Chat-GPT produced its usual garbage, just the dullest dogshit drivel, and lo, and behold, goblin tools too perceives it as emotionally heavy, identifying an emotional tone of introspection and vulnerability.

There's no value in being considerd "poignant" by something that has to ability to recognize something as trite. Which should have been, of course, obvious from the start.

Alas, I am a bit closer now to understanding how incels can fall in love with their chatbots. These things are really getting better and better at mirroring your intentions back to you, and I guess, that's enough to make one feel understood.

It has no discernment, not taste, no actual judgement - but a fairly good score when it comes to catching on to me trying to make a joke. Much better success rate than 90% of the men I've dated so far!
posted by sohalt at 4:34 AM on May 13, 2023 [3 favorites]


The text comes across as a mix of frustration, disappointment, and sarcasm. The author seems to be exhausted from staying up late to feed their writing to a tool for validation. They express disappointment with the results produced by Chat-GPT but also seem to find humor in the situation. The author also expresses a certain level of vulnerability, acknowledging their own desire for validation through these tools and even making a reference to incels falling in love with chatbots. Overall, the tone is somewhat cynical, with the author recognizing the limitations of these tools while also acknowledging their usefulness in providing a sense of validation.

Aww. You get me, goblin tools.
posted by sohalt at 4:45 AM on May 13, 2023 [1 favorite]


Chat-GPT produced its usual garbage, just the dullest dogshit drivel

You should try it next using Bard instead; the goblin judge may self-destruct from the awfulness.

(I know this isn't a thread about Bard! But it's terrible! I've spent the past few days running story premises through Bing, ChatGPT and Bard. Bing has the most fun responses, and actually managed to add two 'new' ideas (caveat lector) and a whole lot of bad ones, but at least made for fun reading, the downsides being the 2k character/20 comment limits, and its bad habit of erasing its responses whenever it thinks it's being offensive. ChatGPT, not as great, but at least it doesn't flash answers in front of you before taking them away! Bard...Bard essentially reworded my sentences like a nervous cheater in a freshman comp class. I was honestly shocked by its lack of trying. It's running on a C64 in some dusty back room of a Google datacenter, the power cord becoming uncomfortably hot.)
posted by mittens at 4:51 AM on May 13, 2023 [1 favorite]


Many sensationalizing headlines are around how kids will "cheat" at school (not sure who they are cheating...) by getting ai to write their papers.

But, is it also cheating if the teachers use sentiment analysis to grade homework?

With how overworked teachers are, who could blame them! But, is it an equivalent disruption of the learning process?
posted by rebent at 6:14 AM on May 13, 2023 [1 favorite]


But, is it also cheating if the teachers use sentiment analysis to grade homework?

I still don't see how that would work; as I said, this thing has no actual judgement. It can name emotions the text intends to express, but can't tell you whether they are conveyed authentically, plausibly, in a compelling manner. It can say that a text seems analytical, but it can't tell you whether the analysis is facile, misguided, baseless, profound, shallow, well- or ill-informed. It can only say something about the apparent intention of the text, often just parroting back the wording of the assignment, not whether the text actually accomplishes its aims.

I guess it can produce some verbiage you can add to a letter-grade, which won't sound completely out of place, but also won't actually justify anything, and I guess for some teachers, and for some students, that will do. But I guess those always gave grades based on vibes and added verbiage as needed even in the before times. And if you have a student who'll demand a more specific feedback to learn how to improve on the next occasion, and comes to see you in your office hours about it, this won't help you at all.
posted by sohalt at 6:46 AM on May 13, 2023 [1 favorite]


If I were still teaching composition, then I would definitely be tempted to use Chat-GPT, not to get out of the real work of reading papers and giving feedback on them... but to reduce the time I spent transforming my first blunt reaction ("Your evidence doesn't support your conclusion") into the version with the compliment sandwich and the encouragement and the kindness and the "here's what you should do" instead of "this is bad."

Sometimes you're grading papers, and you hit a bad one, and you have to close the laptop and take a walk to the kitchen and have a snack, and you're thinking, "Okay, well, what is the NICE way to say what this student needs to hear?" and you're also thinking "Heck, it doesn't matter, they're not going to read anything I write anyway; if they were in the habit of listening to feedback they wouldn't be writing essays this bad," so you get depressed, and you have another snack, and finally you figure out an acceptably nice way to say what you want to say, but you've wasted 20 minutes on a paper that should've taken 10 minutes, and you feel bad about how much time you've spent writing words that no one will read, and how much time you've spent trying to be nice to students who will not care about that effort. And eventually you might well come around to thinking, "Well, if they don't want me to automate my way out of that a little bit, maybe they should be paying me more than $21,000/yr."
posted by Jeanne at 8:08 AM on May 13, 2023 [2 favorites]


Many sensationalizing headlines are around how kids will "cheat" at school (not sure who they are cheating...) by getting ai to write their papers.

It's not sensational - it's actually happening right now in schools. My husband, a HS teacher, has been dealing with it in every single assignment requiring writing since January. They are cheating themselves, obviously, but teachers are villains for flagging things like this and also villains for noticing but not caring enough to flag it, and also villains if they aren't in tune enough with the writing style of each student to notice when the tone dramatically shifts etc etc etc.
posted by kimberussell at 10:09 AM on May 13, 2023 [3 favorites]


1. Relax and sit down at a table
2. Place a plate of beans in front of you
3. Take a deep breath and observe the plate of beans
4. Begin analyzing the beans on the plate - their color, texture, size, shape, etc.
5. Consider where the beans came from, how they were grown, how they were prepared
6. Think about the cultural significance of beans in different countries and cuisines
7. Contemplate the health benefits of beans and how they contribute to a balanced diet
8. Start pondering the philosophical question of existence and wonder if the beans have any meaning in the grand scheme of things
9. Continue overthinking until satisfied, or until your beans get cold.
posted by dg at 8:45 PM on May 14, 2023 [1 favorite]


someone in a neurodivergent cleaning group I follow typoed "clear area near my bed" as "clear area bear my bed" into goblin tools & after telling them to clear out the space it instructed them to locate the bear & move the bear somewhere else

it didn't specify where exactly, just, somewhere else

anyway "locate the bear" really got me
posted by taquito sunrise at 1:17 PM on May 17, 2023 [1 favorite]


« Older Merging the Negro Leagues into MLB stats has not...   |   Of course, you don’t have much time to make your... Newer »


This thread has been archived and is closed to new comments