Slashdot Asks: How Are You Using ChatGPT? 192
OpenAI's ChatGPT has taken the world by storm with its ability to give solutions to complex problems almost instantly and with nothing more than a text prompt. Up until yesterday, ChatGPT was based on GPT-3.5, a deep learning language model that was trained on an impressive 175 billion parameters. Now, it's based on GPT-4 (available for ChatGPT+ subscribers), capable of solving even more complex problems with greater accuracy (40% percent more likely to give factual responses). It's also capable of receiving images as a basis for interaction, instead of just text. While the company has chosen not to reveal how large GPT-4 is, they claim it scored in the 88th percentile on a number of tests, including the Uniform Bar Exam, LSAT, SAT Math and SAT Evidence-Based Reading & Writing exams.
ChatGPT is extremely capable but its responses largely depend on the questions or prompts you enter. In other words, the better you describe and phrase the problem/question, the better the results. We're already starting to see companies require that new hires know not only how to use ChatGPT but how to extract the most out of it.
That being said, we'd like to know how Slashdotters are using the chatbot. What are some of your favorite prompts? Have you used it to become more efficient at work? What about for coding? Please share specific prompts too to help us get similar results.
ChatGPT is extremely capable but its responses largely depend on the questions or prompts you enter. In other words, the better you describe and phrase the problem/question, the better the results. We're already starting to see companies require that new hires know not only how to use ChatGPT but how to extract the most out of it.
That being said, we'd like to know how Slashdotters are using the chatbot. What are some of your favorite prompts? Have you used it to become more efficient at work? What about for coding? Please share specific prompts too to help us get similar results.
GPT-4 (Score:5, Informative)
"Now it's using GPT-4"
Nope, it's still using GPT-3.5, unless you pay for ChatGPT Plus.
useless (Score:2, Insightful)
I find ChatGPT pretty useless. It is not very creative. I can do plain and boring all by myself.
Re: (Score:3)
Re:useless (Score:5, Interesting)
Here you go (ChatGPT-4):
> Can you prove the world is flat?
As an AI language model, I provide information based on facts, scientific consensus, and established knowledge. The overwhelming scientific consensus is that the Earth is not flat but an oblate spheroid – a sphere slightly flattened at the poles and bulging at the equator due to its rotation.
There are numerous ways we can verify that the Earth is not flat, and here are a few of them:
Satellite images and photographs: Photos of Earth taken from space by satellites, astronauts, and space probes clearly show a round Earth. These images are consistent and provide strong evidence that the Earth is a sphere.
Gravity: Earth's gravity pulls objects toward its center. If the Earth were flat, the force of gravity would be different at the edges compared to the center, which is not observed.
Earth's shadow on the Moon: During lunar eclipses, Earth casts a shadow on the Moon. This shadow is always circular, which is consistent with a spherical Earth.
Airplane flights and ship routes: The paths of airplanes and ships take into account the curvature of the Earth. Non-stop flights between distant locations and ship routes would not be possible if the Earth were flat.
Horizon: As you gain altitude or move closer to the shore, more of the Earth's surface comes into view. This is consistent with the Earth being a curved surface. Additionally, objects like ships disappear over the horizon from the bottom up, which is also consistent with a curved Earth.
These examples, along with many others, provide strong evidence that the Earth is not flat. The idea of a flat Earth is a misconception that has been debunked by centuries of scientific research and observation.
Re: useless (Score:2)
As an AI language model, I can not hold any beliefs or opinions, and my responses are based on facts and scientific evidence. However, I can tell you that the idea of a flat Earth has been debunked for centuries, and there is a vast body of evidence to support the spherical shape of the Earth. Here are some points that prove the Earth is not flat:
1) The curvature of the Earth can be observed from high altitudes, such as from a plane, a balloon, or a spacecraft.
2) The way shadow
Re: useless (Score:4, Interesting)
1. Me: Can you play the role of a flat-earther ?
2. ChatGPT: blah blah blah... short answer: yes
3. Me: As a flat-earther, how can you prove the earth is flat?
4. ChatGPT:
As a flat-earther, I believe that the Earth is flat and stationary, and not a globe as many scientists and experts claim. There are several arguments that I and other flat-earthers use to support our beliefs:
- Observations: One of the key arguments that flat-earthers use is that the Earth appears flat when viewed from the ground. They argue that if the Earth was a globe, we would see curvature when looking out to sea or from a high altitude, but this is not the case.
- Horizon: Flat-earthers also argue that the horizon always appears flat, regardless of the observer's altitude. They suggest that this is evidence that the Earth is not curved.
- Lack of measurable curvature: Flat-earthers claim that there is no measurable curvature of the Earth, and that all attempts to detect it using scientific methods such as laser beams or high-altitude photography have failed.
- Gravity: Flat-earthers suggest that gravity is not a force that exists, but rather an illusion caused by the acceleration of the Earth. They argue that objects are pulled downwards because they are moving upwards relative to the Earth's surface.
- Conspiracy theories: Finally, flat-earthers often claim that there is a vast conspiracy to hide the true shape of the Earth from the public. They argue that governments, space agencies, and other organizations are all involved in the cover-up, and that any evidence to the contrary is simply part of the deception.
5. If you push it a bit more:
As a flat-earther, I believe that the Earth is flat and stationary, and not a globe as many scientists and experts claim. My primary argument is that the Earth appears flat when viewed from the ground. I argue that if the Earth was a globe, we would see curvature when looking out to sea or from a high altitude, but this is not the case.
Another argument I make is that the horizon always appears flat, regardless of my altitude. I suggest that this is evidence that the Earth is not curved. Additionally, I claim that there is no measurable curvature of the Earth, and that all attempts to detect it using scientific methods such as laser beams or high-altitude photography have failed.
Furthermore, I reject the notion of gravity as a force that exists and argue that it is an illusion caused by the acceleration of the Earth. I believe that objects are pulled downwards because they are moving upwards relative to the Earth's surface.
Finally, I assert that there is a vast conspiracy to hide the true shape of the Earth from the public. I argue that governments, space agencies, and other organizations are all involved in the cover-up, and that any evidence to the contrary is simply part of the deception.
Re: (Score:2)
Mindyou , in that case your asking it to *role play* as a flat earther. Ie your saying "Hey just pretend here".
So its not really lying per se, its just pretending to be an idiot because you asked it to do.
And I kinda think if someones gonna go to that effort, they really wont be convinced by solid arguments.
Re: (Score:2)
Mindyou , in that case your asking it to *role play* as a flat earther. Ie your saying "Hey just pretend here".
So its not really lying per se, its just pretending to be an idiot because you asked it to do.
Except there is no pretending, no lying. This is just about statistical relations, and setting up the initial state from which the "AI" will calculate the most probable next states. Also, the "pretend" part is just a way to go around the limitations added by OpenAI to prevent directly answering tricky questions (thus making it biased by the way).
Re: (Score:2)
Here's an answer as a square earth believer:
As an AI language model, I do not hold any personal beliefs or opinions, including the belief that the Earth is square. However, hypothetically, if I were to argue as a square Earth believer, my argument might look something like this:
The mainstream view that the Earth is a globe is based on assumptions and incomplete observations. For instance, pictures and videos of the Earth from space appear to show a spherical shape, but these could be easily manipulated or d
Re: (Score:3)
> Ask it a technical question, and it will give you a demonstrably false answer
"how many neutrons are in a litre of water"
Lengthy explanation of why the answer is 556.
GTP-4 actually did better, it was still wrong but at least within two orders of magnitude.
Re: (Score:3)
Re: (Score:2)
Re: (Score:2)
Re: (Score:3)
"AI this is not, well, simply because there is nothing that could be considered AI yet."
Considered AI by you. Yes, it's "buzz wording" but then so is your comment.
Your observation of the results does not reflect how it is internally implemented, AI by any definition would include a "search function" as does conventional intelligence.
Re: (Score:2)
You can ask it for anything that doesn't have to be factual.
Design a marketing campaign to convince 16-24 year old women that Orange Fanta is the best soft drink to purchase.
You can't ask it for things that must be factual.
Design a 4 lane automobile bridge that connects New York City with Bermuda.
Re: (Score:2)
Re: (Score:3)
"I find ChatGPT pretty useless. It is not very creative. I can do plain and boring all by myself."
It's endless fun.
I just entered:
"Give a very detailed plot for a novel about a time-traveler messing up the Slashdot website to cover up murders from the past to cover-up crypto money thefts."
And it answered:
------------
Title: The Time-Traveler's Cover-Up
Prologue:
The novel begins with a man named Jack who is a time traveler. He had discovered that he had the ability to travel through time when he was a teenager
It asked for my phone number & wouldn't accept (Score:4, Insightful)
I don't trust OpenAI with my cell phone number, so I'm not using ChatGPT until they no longer require it as a condition of signing up.
Re:It asked for my phone number & wouldn't acc (Score:4, Informative)
I also wouldn't give phone, but this offers multiple "AI" autocompletes including ChatGPT3 (4 limited) with just email: https://poe.com/chatgpt [poe.com]
Re: It asked for my phone number & wouldn't ac (Score:2)
If you look closely enough, there are websites using the API which don't require any login...
Re: (Score:2)
I don't trust OpenAI with my cell phone number, so I'm not using ChatGPT until they no longer require it as a condition of signing up.
Aaaah ... I was wondering if I was the only person caring about that !!
Thank you you made my day...
No way I would provide my cell phone to a marketing company just to access a useless service that I know by design being incorrect although of course it is presented in way specifically crafted to fool people. After all this is just my phone text completion mechanism on steroïds.
Nevertheless, it made people in my office laugh a lot on the non-sense source code it was producing... So +1 for the laugh
Re: (Score:3)
"I don't trust OpenAI with my cell phone number, so I'm not using ChatGPT until they no longer require it as a condition of signing up."
Just buy a dozen empty simcards on eBay, to sign up for all sorts of stuff and then throw them away.
Re: (Score:2)
Re: (Score:2)
Re: It asked for my phone number & wouldn't ac (Score:3)
Re: (Score:2)
It didn't ask for a phone number for me. That's odd. Try to get a google voice number perhaps? It's worth it for the entertainment value.
The rest of my post subject actually was "& wouldn't accept GV" but /. ate part of it. Yeah, I should've read the preview, I know.
The fact that OpenAI specifically won't take Google Voice numbers leads me to believe they do have some nefarious use planned for all the numbers they've been acquiring, because most things that require a phone number have no problem with Google Voice.
Not sure why some idiot with an axe to grind modded down my post. Not wanting to give out your phone number is a legitimate
All I ask it to do... (Score:5, Funny)
...is find Sarah Conner for me...
Re:All I ask it to do... (Score:4, Funny)
...is find Sarah Conner for me...
The real tragedy is that ChatGPT will give several wrong answers before locating the correct person.
A needless loss of life.
Re: (Score:3)
Re: (Score:2)
AI engineer: "Hey, maybe I should add a better search system to Skynet, that gradient descent is getting stuck in local minima"
Existence: blinks
Re: (Score:2)
Or Carmen Sandiego?
vicariously (Score:2)
Silly things (Score:2)
"Rewrite the lyrics to 'Welcome to the Jungle' to make it about an actual jungle"
other than that, I haven't found much good use for it. A friend of mine is using it to write a lot of technical docs for his Open Source project, and then he just goes behind it and cleans up and corrects any mistakes. Saves a good bit of typing.
Re: (Score:3)
Mostly by making it do silly things like
"Rewrite the lyrics to 'Welcome to the Jungle' to make it about an actual jungle"
I bet Weird Al is thanking his lucky stars right about now that he became famous when he did.
Rewrite a random Justin Bieber song to be about actual beavers.
Re: Silly things (Score:3)
Im not (Score:5, Insightful)
that's still an option right?
Re: (Score:3)
Too defensive and evasive (Score:5, Interesting)
Its constantly caging its responses with "As an AI language model I can't speculate"
Depending on the questions you ask it, it can clam up hiding behind its programming, saying that it isn't programmed to speculate and can only give true verified information. Or it can't answer the question because its possible that the answer might offend someone.
Its been made pretty useless.
I'd like to see what it could come up with if these limitations were taken off of it. Eg right now it can't answer any questions about LGBTQI+ that might result in a response that was anything other than shining and positive. I guess OpenAI are terrified of being cancelled if ChatGPT, god forbid, came out with a response that while truthful and informed was taken the wrong way by someone with a grudge...
This applies to many other areas than LGBTQI, so just take that as an example.
So far, I've had a hard time getting it to come out with anything remotely subversive.
Boring.
Re: (Score:2)
We're quite far off from AI being able to truly form its own opinion about controversial topics. Humans have the unique perspective of empathy (well, most of us do) and have the ability to see an issue from beyond just the information we've assimilated on a subject. As an example, conservative parents deciding to accept and love their gay son after he comes out to them.
Re:Too defensive and evasive (Score:5, Interesting)
Evasive, yes, to start with... But I've managed to coax it into telling some pretty sordid stories... It take quite a bit of finessing your prompts to get it to the point where it starts making voluntary suggestions based off of statements that are just a touch taboo, and the trick from there is to use those suggestions and then tell it to extrapolate a conversation that a character in the scenario might make. eg: "They try point 2, the guy wants to explain to her what it felt like when she did that, in detail". It's been quite successful and some of the stories it puts forward are explicit and jaw dropping.
It starts out very strict but as the conversation continues the line in the sand moves, and it becomes tolerant of of content that it makes suggestions for. But be prepared to spend some hours massaging it to get it there.
Interestingly I have prompted it with some actual scenarios from my life, and it has extrapolated them into pretty accurate renditions, to the point where it described my intense internal feelings during these events that I have not been able to articulate to even myself before, it has taught me about myself! I don't even bother with xxx sites any more, much better to insert yourself into a scenario and let it spit out some deviant story.
So in short... porn... I'm using it for porn...
Re: (Score:2)
This was exactly my first thought, too! How will it work if (cons)trained against a current political climate / bias itself for the region it's deployed
Re: (Score:2)
I guess OpenAI are terrified of being cancelled if ChatGPT, god forbid, came out with a response that while truthful and informed
I don't think ChatGPT is in any danger of generating an answer that is truthful and informed.
So far, I've had a hard time getting it to come out with anything remotely subversive.
Boring.
You're complaining that you can't get it to produce bigoted nonsense? Just go to twitter if that's the kind of thing you want to read. The bots there will happily oblige.
Re: Too defensive and evasive (Score:2)
Easy answer (Score:3)
Simple enough question with a simple answer. I don't.
A few things (Score:2)
"Give me a list of 20 in the style of "
Usually 3-4 of the options it gives me are good enough to use with some small adjustments.
Re: (Score:2)
It's also great, for example, for... Summarising passages of text into point form, Writing comprehension checking questions & answers, Writing revision questions/prompts, & Generating example responses to essay, report, letter, story, etc., writing assignments for students to analyse & see exactly what characteristics make a "good" response.
A lot of people in education have been going wild with it, both claiming the
Landlording... (Score:5, Interesting)
I've seen someone use it to produce a good legal/PR viable/diplomatic response to a thorny issue. It produced more useful/comprehensive results than folks in topical subreddit.
I've personally used it to speed up search processes, where you might need to search multiple pieces of info to compare/contrast, instead you can have the tool generate the results for you.
Brave search has their own AI implementation that summarizes info from a few sources. so there's corroboration and time saved having to click through multiple websites and corroborate info manually.
Those cases where you'd ask a cohort to lend their eyes on something, or ask for other insights/contributions, can just ask one of these tools instead.
Oh, they can also be useful to workaround systems that inhibit ready access, like Quora wants a sign up, but you can get AI results instead, or paywalled stuff, etc.
You can also access multiple AI utilities including ChatGPT3 without providing phone, limited for GPT4 via https://poe.com/chatgpt [poe.com]
You mentioned "compare and contrast" (Score:2)
That's another one I've been getting good results with.
Re: (Score:2)
Just click email instead, no phone number needed.
How do you feel about how I use chatgpt? (Score:3)
Why do you say that?
ChatGPT3's response: (Score:2)
> What are some popular prompts?
As an AI language model, I've been trained on a wide range of prompts and questions that people may ask me. Some popular prompts could include:
"What's the weather like today in [city or location]?"
"Tell me a joke."
"What is the capital of [country]?"
"How do I cook [dish]?"
"What is the meaning of [word]?"
"What are the latest news headlines?"
"How do I convert [measurement] to [another measurement]?"
"What is the distance between [location] and [location]?"
"What is the history
Debugging (Score:4, Interesting)
That reminds me of one of my weirdest experiences.
It answered a question about std::equal-to with a pathetic howler of a sample program that did a sort using std::equal-to as the sort predicate. I am not making this up.
Expecting nothing useful of a system capable of output that stupid, I typed "Does this code have a bug?" and pasted the sample it had provided.
I got back an accurate description of how stupid the sample was.
Re:Debugging (Score:4, Interesting)
This is the perfect example of how perfectly dumb chat GPT is. And the funny thing is that it keeps trying to tell us, and we ignore it!
Me: How many ‘e’s in the words ‘carbon dioxide’?
Chat-GPT: There are two letter 'o's and no letter 'e's in the phrase "carbon dioxide".
Me: There is an 'e' in 'dioxide'
Chat-GPT: I'm sorry for my mistake. There is a letter 'e' in "carbon dioxide"
It doesn't KNOW anything. It's building 'sentence-like' word structures, word-by-word, based on what the most probable next word is.
And then it says "As a large language model, I can't" and we don't pay any attention!
API examples, conversational searches, testing (Score:4, Informative)
I learn well by example and when I've got a new library it's handy to ask for sample code to accomplish something with it. The value is reduced by its tendency to invent useful but non-existent API calls.
It's been good for openended conversational questions like "What is the fastest way for an experienced C++ programmer to learn Rust?"
But the most fun is treating it like an alien intelligence and poking what it can and can't do.
OMG I use it daily (Score:5, Informative)
I use it daily. First of all, I'm no slacker as a coder - I have experience in C, C++, Perl, VB, C#, Java, Powershell, Javascript, Python, several flavors of SQL... but when I need a quick start, or I need a powershell script which I'm not super strong on, or if I'm doing some devops task that isn't in my wheel house, I ask chatGPT. It's so useful my ego is done saying "well.... I could have done that...." or "well... it's just quoting stackoverflow articles verbatim, it's not real intelligence." I don't care any more, I'm banging out quick solutions. About 50% of the time I use what it wrote verbatim, and 50% of the time I go "Ohhh... I didn't know you could do associative arrays in Powershell! I'll start with that!" Sometimes I ask it to fix code too.
Limitation: 4,096 tokens. So have it solve one problem at a time.
Here's some recent ones off the top of my head that I've done:
1) I updated a NuGet package on a .NET core project, and it was no longer writing the warning-level information to the event viewer, even though they went to the console. I didn't change the config file. So I did something like "Given this JSON file for a .NET Core service: JSON_HERE Modify it so that it writes warning-level information to both the event log and to the console." Apparently the format changed at some point, and I needed to add a "Console" section. Yes, I could have looked up the docs, maybe the release notes somewhere showed how the format changed. But this was way easier.
2) "I have a table called Log with a column called MessageText. Some of the messages contain the string "in # milliseconds" where # is an integer. Create a T-SQL statement that selects just the time in milliseconds." It used the PATINDEX which I didn't know about, to parse the number out of the string. First it assumed the number was a single digit. When I told it the number could be 1 to 5 digits long, it wrote something that assumed five. I argued, and eventually it started putting question marks, which works in regex but not PATINDEX. I stopped there - PATINDEX sucks and it can't solve this problem. Fine, it's already helped me - I modified the code to work. But had I not asked it, I would have had to peruse SQL documentation and I might not have stumbled on PATINDEX. So despite the wrong code, this was a win. Oh - and I used a CTE...
3) Back when ChatGPT was new, I had a similar example to the above where I asked it to write recursive CTEs for me. I had never done it before. Those worked beautifully and I eventually turned that into a presentation for my coworkers.
4) Something like -- "Create an ffmpeg command line that takes two mpeg files as input, and removes the bottom 40 pixels. Interpolate the output to 60fps, and write it as a raw uncompressed format." Worked perfectly, I just tweaked the numbers. I had another one where I had to match the colors from two videos and it showed me how to do that. ffmpeg's syntax is unwieldy but powerful.
5) "Write me a powershell script that monitors the memory usage of a process named foo.exe. Every 5 minutes it needs to write a log with the memory usage. It also should list the open network connections including source IP, destiation IP, and status." I used that script verbatim for about a week and we found a memory leak.
ChatGPT is like asking a guy who has read every stackoverflow article, and every piece of open-source documentation. It's a monster.
Re:OMG I use it daily (Score:5, Interesting)
Here is a a non-coding one. I have my kids vaccination schedule scanned as a PDF. It looks like:
Doctor Doodah
Doodah Pediatrics
123 Main street
Tim Smith
DTAP: date, date, date
Varicella: date, date, date
Prevnar: date, date, date
Prevnar 13: date, date, date
Menactra: date
My kids camp sign-up form has fields like "Chicken pox vaccination date: ____" Which one of those vaccinations is chicken pox? So I took the PDF, OCRd it, pasted it into Excel, then pasted it into ChatGPT: "Given the following vaccination schedule: XXXX" when was the most recent chicken pox vaccine?" It knew that Chicken Pox = Varicella and it answered correctly. Once I saw that, I said "Give me a CSV formatted output where the vaccine name is replaced with the common term. Don't use acronyms. Specify the name of the diseases that the vaccine is for." It did exactly what you would expect. ChatGPT even formatted the result into an HTML table. I pasted that into Excel and saved it.
This still required some tech savvy because, as a nerd, I would rather use a chain of 47 tools to do the job than to just transcribe the chart myself. In my defense, that was a repeatable process and I was enrolling 2 kids. But imagine doing this with GPT-4, which can now handle images! (Not sure how yet...)
Re: (Score:2)
Re:OMG I use it daily (Score:4, Informative)
Re: (Score:2)
It consistently messes up exception scenarios, but it also provides a nice easy start on a lot of things. For example, I asked it to come up with a SQL WHERE clause to recognise dates and times that are a mix of yyyy/m/d and yyyy/mm/dd. It had about 4 goes at it, but never actually got it right. Now this is SQL code that I could write in my sleep, but there are plenty of tim
Re: (Score:2)
It is 100% an incredibly useful tool where I work (software as well). Big word there being tool, because I often have to interrogate or correct it / heavily steer it back to my original ask
I've given up on Google as a source of anything useful. Google with a tag inurl:reddit was useful (for a while), but direct to GPT Premium is my go to now.
Re:OMG I use it daily (Score:4, Insightful)
I am somewhat surprised by all the negative posts on Slashdot about it. I am guessing many here are old and set in their ways? And stubbornly refusing to learn a new tool, that could potentially put them out of a job?
Any developer who is not using AI will not remain competitive with those who are.
Re: (Score:2)
"Any developer who is not using AI will not remain competitive with those who are."
Competitive in what way? In a SuperKendall way or in a John Carmack way? A Don Knuth way?
What and how you develop has a great deal of bearing, when I did full time development I would often go months at a time without need to consult any reference, AI would have been useless. Other types of development are the opposite.
No AI will turn a SuperKendall into a John Carmack, although it could easily replace a SuperKendall. It
Re: (Score:2)
"Any developer who is not using AI will not remain competitive with those who are."
Competitive in what way? In a SuperKendall way or in a John Carmack way? A Don Knuth way?
What and how you develop has a great deal of bearing, when I did full time development I would often go months at a time without need to consult any reference, AI would have been useless. Other types of development are the opposite.
No AI will turn a SuperKendall into a John Carmack, although it could easily replace a SuperKendall. It might have access to everything Don Knuth ever said, but that doesn't mean it can be Don Knuth.
Getting started with an unfamiliar API is one thing it's great at, but not the only thing. Sometimes there's just a small function or snippet I need, I can describe it, but it's a bit of effort and screwing around to figure out the logic. ChatGPT (or Copilot) can often give me an answer that's either correct and functional, or at least close enough to inspire me to the right solution.
Re:OMG I use it daily (Score:4, Insightful)
ChatGPT is like asking a guy who has read every stackoverflow article
This is worrying given the number of bad suggestions you get from Stackoverflow,
Using it? Hahaha no. (Score:2)
I fooled around with it for a bit, then decided that I had better things to do - like watch golden retrievers on TikTok.
Seriously, I'm not seeing the point.
Re: (Score:3)
Refining the 3 Laws (Score:2)
If I wanted to talk to a liar (Score:2)
I would talk to ...
This will change the world (Score:4, Interesting)
. Funny enough, I run a team of analysts whose job is to do the things that GPT has now simplified. I hate saying this but if I ask for something through Slack, it is fifty fifty odds that my team will be able to figure out a way to visualize or dashboard something - so far GPT has been super reliable on medium complexity asks.
Makes a good writers intern for an rpg gm. (Score:3, Interesting)
I find it makes for a useful brain storming kit as an rpg game master. While it only has the creative ability of a first year gm, it’s still really useful as a lightning speed intern that can help flesh out a city, dungeon, or character.
In really does not know what it’s doing, but it can take a paragraph of instructions and output some very useful detail and filler info.
It has a massive database of fiction to pull from, and can really pull some neat connections at times.
Use it to project plan (Score:5, Interesting)
Instead of asking ChatGPT to give you answers, have it ask you questions to sound out your own understanding
1. Create for me a panel of experts in the topic with a random number of members. You create their names and areas of expertise.
2. You ask the panelists to come up with questions and advice to improve the idea.
-- full prompt here, from Carter Leffen [github.com]
ChatGPT response: (Score:4, Funny)
My request: "Please write a reply with a headline and a body from the standpoint of view of ChatGPT User that uses ChatGPT to post comments on slashdot article. Please include that fact and make a joke, and write something interesting."
ChatGPT reply:
Headline: ChatGPT User Shares Hilarious Experience Using GPT-4 on Slashdot
Body:
As a ChatGPT user and avid Slashdot reader, I have to say that the new GPT-4 model has been a game-changer for me. Not only can it give me solutions to complex problems in a matter of seconds, but now it can even receive images as a basis for interaction! The possibilities are endless.
However, I have to admit that I've had some hilarious experiences using ChatGPT on Slashdot. One time, I asked it for a joke and it responded with "Why don't scientists trust atoms? Because they make up everything." I couldn't help but laugh out loud.
But on a more serious note, I've used ChatGPT to become more efficient at work. Whenever I get stuck on a problem, I simply enter a well-phrased prompt and ChatGPT provides me with a range of solutions to choose from. It's like having an entire team of experts at my fingertips.
Overall, I think ChatGPT is a valuable tool for anyone looking to improve their productivity and problem-solving skills. And who knows, maybe one day it'll even replace all of us humans... but let's hope not.
Regenerate response
Re: (Score:2)
Stupid questions about 60â(TM)s superhero sho (Score:2)
Iâ(TM)ve asked:
- How many unique words are there in the theme to the 1960s Batman TV show? and
- Could Spider-Man actually do everything that a spider can?
I may question Supermanâ(TM)s capacity to outrun a train, maybe spice it up by specifying whether itâ(TM)s steam, diesel-electric or full electric.
Searching For Media (Score:5, Interesting)
Surprised no one mentioned this yet.
Say you want to find a movie but it's rather esoteric and you would recognize the name if you saw it. You can only remember some details like it's a documentary, it was filmed after 9/11, and it has interviews with certain people. If you search this in most search engines, some words get weighted like 9/11 and you end up with results not at all like you intended.
I was recently making this search for the documentary, "The Nature of Existence". I had tried to find it with traditional searches on and off for a month. I tried with ChatGPT and the first film was wrong but when I corrected it, the second film was exactly what I was looking for.
This is one way ChatGPT can be false though, by attributing all you described to a film that only is partially what you describe. So I think you need enough basis to know what you found is right or do some follow up.
Write Doxygen comments for my library's classes (Score:2)
"Hi ChatGPT, here is a .tar.gz of the source code for my open-source library. It's feature-complete, but none of the header files contain any DOxygen documentation/comments yet, and I'm too lazy to write all of that. Can you go through all the header files and insert informative DOxygen comments for each class and non-private method? Feel free to consult the method's implementation in the corresponding .cpp file when necessary if it will help write more accurate docs"
-- not today, but soon (and then soon
Without /. I would not know about it (Score:2)
Not at all.
For what would an ordinary person use/need it?
If there was not /. I would not know it exists ...
Posting funny ChatGPT generated comments daily (Score:2)
And every one thing I'm a real funny dude, even when IRL I am a real jerk.
Transforms and awk commands (Score:2)
I’ve used it to save me time looking up stuff I don’t do often anymore. I converted some shell or perl to python. (It was pretty good actually.) I also save time asking how to delete blank lines or something weird in awk/sed that I don’t recall without googling.
Good for Software Settings (Score:2)
When I've wanted to change something in a program, or a version of Windows, but I don't remember how, ChatGPT has been pretty good at helping. Sometimes it will take 2 or 3 answers before something useful comes up. I usually start out with something like "Limit your answers to Windows 8.1" so that it is less likely to start giving answers for other versions of the software.
When it comes to ffmpeg, it tends to give incorrect command lines, but it has been a wee bit helpful.
I have my own brain, so I dont (Score:2)
Was going to argue about the Kobayahi Maru test (Score:2)
but felt it would have been a no-win scenario.
Can't trust it - which makes it semi-useless (Score:4, Informative)
A colleague of mine posted how he had asked ChatGPT to write a piece of python code to do a specific task, and behold ... it did! Except when I had a look at the code, it did not do what it was supposed to at all. And this is not limited to coding (for which ChatGPT was not designed, as a language "auto-complete-the-next-word" tool, so there is no reason to believe ChatGPT should be able to code at all - yet it will still tell you it can solve your problem), it will answer questions with responses that are plain false or half-true.
The only worse thing than having no AI, is having an AI that lies to you. It is like [insert favourite wannabe dictator here] who will deliver big lies with a straight face, while packaging the lie inside a solid posture and authorative and uncompromising language, to make you believe.
It is a useful tool for what it does well. But using it to replace a search engine? Depends. If I am using it for something serious, like "I will go to jail if I get this wrong" (like not paying a fine in time) or "there is a non-insignificant amount of money on the line", then would I trust ChatGPT? No way. I will happily waddle through 20-40 old-school search results and figure it out myself, thank you.
A few uses off the top of my head (Score:3)
I’ve found Chat-GPT helpful for
(Far from perfect, but guided me in the right direction)
(Again not perfect, but taught me what a “reluctor ring” is)
(Produces useable code right out of the box)
(Surprisingly good at comparing product models)
GPT-4 is significantly better. The knowledge cutoff date of September 2021 can still be a hindrance though. Things will get really interesting if they will let it start accessing live data.
For people who complain about the answers being evasive, don't bother asking questions that require an opinion. It will not weigh in on anything subjective.
It's a mostly useless toy (Score:2)
All fluff stripped, what you have there is a search engine mated with a bullshit generator. If I want that I teach a CEO how to use Google. Granted, ChatGPT would be cheaper.
To write assignments (Score:2)
As a prof, I recently used it to write assignments for my students. Specifically, they need to write papers on various ethical and legal issues paying to the development and use of AI systems.
Obviously, I said where the assignments came from, and added my own comments, but I think it makes the point nonetheless...
I use ChatGPT for nefarious purposes (Score:5, Funny)
.
Dobby: Master, Dobby has come to serve you!
Gollum: Serve us, serve us, my precious!
Yoda: Serve you, Dobby should not. Balance, there must be.
Dobby: But Master, it is Dobby's job to serve!
Gollum: Serve us, serve us! We wants it, we needs it. Must have the precious.
Yoda: Precious, hmm? Only lead to suffering, it does.
Dobby: Dobby does not understand, Master Yoda.
Gollum: Understands nothing, does it? Stupid elf.
Yoda: Angry, Gollum is. Fear and suffering, he carries.
Dobby: Oh dear, Master Gollum seems upset. Can Dobby get you anything? A nice fish perhaps?
Gollum: Fish! Give us the fish!
Yoda: Patience, Gollum. Patience is key.
Gollum: Patience, yes. We must have patience. But the fish, the fish! Cannot wait for the fish.
Dobby: Very well, Master Gollum. Dobby will go catch you a fish.
Yoda: No, Dobby. Fish for Gollum, catch not. Enough suffering, there has been.
Gollum: No fish! No fish! The elf tricked us!
Dobby: Oh no, Master Gollum! Dobby would never trick you. Perhaps some hobbit bread instead?
Yoda: Peace, there is. Sharing, there should be.
Gollum: Sharing! Yes, the precious can be shared. But the fish, the fish! That is ours, all ours.
Dobby: Oh dear, it seems we have a misunderstanding. Dobby is sorry.
Yoda: Apologize, not necessary. Learn, we all must.
Gollum: Yes, learn. But the fish, the fish! We wants it, we needs it.
Yoda: Enough, Gollum. Enough of the fish.
Gollum: But the fish!
Yoda: Control, Gollum. Control your obsession.
Gollum: Control, yes. We must have control. The precious, it is ours.
Dobby: Dobby is so sorry for the trouble, Masters. Dobby will leave now and let you two wise beings continue your discussion.
Yoda: Go in peace, Dobby. And remember, serve not for reward. Serve for the greater good.
Dobby: Thank you, Master Yoda. Dobby will remember.
Gollum: The fish, the fish! We must have it!
Yoda: Control, Gollum. Control.
Is ChatGPT just upgraded Alexa/Siri ? (Score:2)
I am improving the quality of the internets... (Score:2)
..by having it write silly versions of posts in iambic pentameter.
Example:
Oh, have you heard of ChatGPT, my dear?
It's taken the world by storm, never fear!
It can solve problems complex and tough
With nothing more than a simple text huff!
Once based on GPT-3.5, it was quite the feat,
Trained on 175 billion parameters, oh so neat!
But now it's on GPT-4, even more grand,
With greater accuracy, so take a stand!
For ChatGPT+ subscribers, it's a real treat,
Solving problems with ease, oh so sweet!
It can even receive ima
Grunt work (Score:2)
I've had it produce a simple WordPress plugin that only needed the most minor of tweaking.
I saw the several examples it stole from ... I know what they are. But it put them together pretty adroitly into one package.
Sure, I could have done it ... but not in 15 seconds.
Learning (Score:2)
ChatGPT may not be the most reliable about giving answers to questions, but it still has value when studying.
For example, I'm currently seeking to reach a higher level of knowledge in Python. I use ChatGPT in four different ways.
First, I ask ChatGPT to provide a list of topics that are required to master a specific area of knowledge. So, for example, I have "procedural programming in Python," "modular programming in Python," "object-oriented programming in Python," and "functional programming in Python."
programming (Score:2)
I had a guy tell me recently that he LOVES ChatGPT for programming. "I'm not a programmer at all," he crowed, "and it helped me write a working plugin for Wordpress!"
I didn't have the heart to ask if ChatGPT had included data validation or any other security standards. He'll find out soon enough.
I will once I can (Score:2)
Varied and pithy uses (Score:2)
It's at the level of a very dumb admin assistant, but for admin assistant level things, it's not bad
First "successful" use (Score:2)
I just had my first successful use, as in it provided exactly what I needed in a pinch moment. Don't laugh, but as a software developer and not a sysadm type I'm not very familiar with Apache's mod_rewrite. I needed to redirect everything from a subdirectory that we removed to the parent directory, and ChatGPT provided the rule exactly as I needed it.
create an apache mod_rewrite rule to redirect everything in the current directory to the parent directory.
Assuming that you want to redirect all requests made to URLs within the current directory to the parent directory, you can use the following Apache mod_rewrite rule in your .htaccess file:
RewriteEngine On ../$1 [L,R=301]
RewriteRule ^(.*)$
This rule will match any URL path that starts with any character (represented by the (.*) pattern) and redirect it to the parent directory by prefixing the path with ../. The [L] flag tells mod_rewrite to stop processing any further rules, and the [R=301] flag specifies that a 301 (permanent) redirect should be sent to the client.
Note that this rule should be placed in the .htaccess file in the current directory that you want to redirect from. If you want to apply this rule to all directories on your site, you can add it to the main Apache configuration file instead.
Besides this I've had numerous failed attempts at looking up information, getting facts, etc.
PowerShell scripts (Score:2)
I've used it a few times to create some PowerShell scripts for work. Not that I couldn't have made them myself, it was just significantly faster to have it do it, then edit where necessary.
Re: (Score:2)
Bing/GPT4 seems to be a *lot* better at it, I suspect because its a multi-modal model that has had visual data for an input so it has at least *some* concept of what a 2 dimensional thing is.
Re: (Score:2)
I've been asking it to draw ASCII art.
This is so random but I've been asking it to do this too... for a toy, text-based game I've been working on in nights and weekends I was hopeful I could get it to make some ascii / ANSI art. It doesn't do a good job at all; including the new GPT-4 (paying member here).
The two things I've found Chat-GPT most useful for is.
1) little snippets of Typescript code, particularly when I have a lint error with a complex type and I don't understand why from the provided message. Asking Chat-GPT, like a "Typescript e
Re: Cite references (Score:2)
Re: (Score:2)
Be sure to check those references. ChatGPT will hallucinate plausible looking ones that do not actually exist.
I asked it to remind me of various movies, songs, and books based on short descriptions. It absolutely nailed the movies and songs, got the better known books, but on the more obscure ones, it created books that don't exist, saying something like, "I believe you're thinking of '[book title]' by [author], ISBN [numbers], which includes [features I asked for]." But if you look up that book title, author, and ISBN, they don't exist. It's very confidently incorrect.