When I search this topic online, I always find either wrong information or advertising lies. So what is actually something that LLMs can do very well, as in being actually useful and not just outputing a nonsensical word salad that sounds coherent.
Results
So basically from what I’ve read, most people use it for natural language processing problems.
Example: turn this infodump into a bullet point list, or turn this bullet point list into a coherent text, help me with rephrasing this text, word association, etc.
Other people use it for simple questions that it can answer with a database of verified sources.
Also, a few people use it as struggle duck, basically helping alleviate writers block.
Thanks guys.
Not exactly sure this is the “right way” to use them, but I use one as an autocomplete helper in my IDE. I don’t ask it to code anything, just use it as autocomplete.
Majority of the time, it works well, especially in common languages like Python.
Taking a natural language question and providing a foothold on a subject by giving you the vocabulary so that you can research a topic on your own.
“What is it called when xyz.”
Very basic and non-creative source code operations. Eg. “convert this representation of data to that representation of data based on the template”
I find it’s really good for asking extremely specific code questions
Just rewrote my corporate IT policies. I feed it all the old policies and a huge essay of criteria, styles, business goals etc. then created a bunch of new policies. I have chatgpt interview me about the new policies, I don’t trust what it outputs until I review it in detail and I ask it things like
What do other similar themed policies have that I don’t? How is the policy going to be hard to enforce? What are my obligations annually, quarterly and so on?
What forms should I have in place to capture information ( i.e. consultant onboarding).
I can do it all myself but it would be slower and more likely to have consistency and grammatical errors.
A fringe case I’ve found ChatGPT very useful is to learn more about information that is plentiful but buried in dead threads in various old school web forums and thus very hard to Google. Like other people’s experiences from homebrewing. Then I ask it for sources and most often it is accurate to the claims of other homebrewers that also can be correct or less correct.
I use it to help me come up with better wording for things. A few examples:
-
Writing annual goals for my team. I had an outline of what I wanted my goals to be, but wanted to get well written detail about what it looks like to meet or exceed expectations on each goal and to create some variations based on a couple of different job types.
-
Brainstorming interview questions. I can use the job description and other information to come up with a starting list of questions and then challenge the LLM to describe how the question is useful. I rarely use the results as-is, but it helps me to think through my interview plan better than just using a list of generic questions.
-
Converting a stream of thought bullet list into a well written communication.
-
I have it make me excel formulas that I know are possible, but I can’t remember the names or makeup for. Afterwords I always ask “what’s a better way to display this data?” And I sometimes get a good response. Because of data security reasons I dont give it any real data but we have an internal one I can use for such things and I sometimes throw spreadsheets in for random queries that I can make in plain language.
As a developer, I use LLMs as sort of a search engine, I ask things like how to use a certain function, or how to fix a build error. I try to avoid asking for code because often the generated code doesn’t work or uses made up or deprecated functions.
As a teacher, I use it to generate data for exercises, they’re especially useful for populating databases and generating text files in a certain format that need to be parsed. I tried asking for ideas for new exercises but they always suck.
I am not using it for this purpose, but churning out large amounts of text that doesn’t need to be accurate is proving to be a good fit for:
-
scammers, who can now write more personalize emails and also have conversations
-
personality tests
-
horoscopes or predictions (there are several examples even on serious outlets of “AI predicts how the world will end” or similar)
Due to how good LLMs are at predicting an expected pattern of response, they are a spectacularly bad idea (but are obviously used anyway) for:
-
substitute for therapy
-
virtual friends/girlfriend/boyfriend
The reason they are such a bad idea for these use cases is that fragile people with self-destructive patterns do NOT need those patterns to be predicted and validated by a LMM.
Have they given you anything creative that was good. I also, used it to make a meal plan and make a work schedule as an Excel doc, then it just needed a few edits.
Would you say you are good at creating a meal plan or a work schedule by yourself, with no AI? I suspect if you know what a good meal plan looks to you and you are able to visualize the end result you want, then genAI can speed up the process for you.
I am not good at creative tasks. My attempts to use genAI to create an image for a PowerPoint were not great. I am wondering if the two things are related and I’m not getting good results because I don’t have a clear mental picture of what the end result should be so my descriptions of it are bad
In my case, I wanted an office worker who was juggling a specific set of objects that were related to my deck. After a couple of attempts at refining my prompt, Dall-E produced a good result, except that it had decided that the office worker had to have a clown face, with the make-up and the red nose.
From there it went downhill. I tried “yes, like this, but remove the clown makeup” or “please lose the clown face” or “for the love of Cthulhu, I beg you, no more clowns” but nothing worked.
I once asked ChatGPT how it (AI) works. It gave me the tools needed to get the right results. There were books on prompt engineering free online. But I decided after reading them that it was easier to have AI teach me to use AI…better. that’s the LLMs. On the other hand for image generation, it takes persistence and priority. If the prompt is too complicated, it will do its own thing. If it is too simple, it will do its own thing. After a lot of practice getting to know how it outputs images you will find the right, or close results. Emphasis on close. Leonardo.ai is my favorite.
Edit: if you don’t believe you are creative enough, prone the LLM for ideas. Ask it to make the prompt. They are finnicky
-
I find they’re pretty good at some coding tasks. For example, it’s very easy to make a reasonable UI given a sample JSON payload you might get from an endpoint. They’re good at doing stuff like crafting farily complex SQL queries or making shell scripts. As long as the task is reasonably focused, they tend to get it right a lot of the time. I find they’re also useful for discovering language features working with languages I’m not as familiar with. I also find LLMs are great at translation and transcribing images. They’re also useful for summaries and finding information within documents, including codebases. I’ve found it makes it a lot easier to search through papers where you might want to find relationships between concepts or definitions for things. They’re also good at subtitle generation and well as doing text to speech tasks. Another task I find they’re great at is proofreading and providing suggestions for phrasing. They can also make a good sounding board. If there’s a topic you understand, and you just want to bounce ideas off, it’s great to be able to talk through that with a LLM. Often the output it produces can stimulate a new idea in my head. I also use LLM as a tutor when I practice Chinese, they’re great for doing free form conversational practice when learning a new language. These are a just a few areas I use LLMs in on nearly daily basis now.
I use LLMs to generate unit tests, among other things that are pretty much already described here. It helps me discover edge cases I haven’t considered before, regardless if the generated unit tests themselves pass correctly or not.
Oh yeah that’s a good use case as well, it’s a kind of a low risk and tedious task where these things excel at.
I use it to review my meeting notes.
- “Based on the following daily notes, what should I follow-up on in my next meeting with #SomeTeamTag?”
- “Based on the following daily notes, what has the #SomeTeamTag accomplished the past month?”
- etc.
I’m not counting on it to not miss anything, but it jogs my memory, it does often pull out things I completely forgot about, and it lets me get away with being super lazy. Whoops, 5 minutes before a meeting I forgot about? Suddenly I can follow up on things that were talked about last meeting. Or, for sprint retrospectives, give feedback that is accurate.
To add: I’ve also started using AI to “talk to podcast guests.” You can use Whisper to transcribe a podcast, then give the transcript to AI to ask questions. I find the Modern Wisdom Podcast is great for this.
They work well when being correct doesn’t matter
Well, yes, but then what’s the point? It would be like having Wikipedia filtered through Alex Jones.
When I’m in a hurry I use them for
- longer more complex excel formulas.
- to create powercell scrips to manipulate large csv-files.
I used it to teach me app script and it was 90 percent accurate
They help me make better searches. I use ChatGPT to get a good idea of what better to search for based on my inquiry. It tells me what I am looking for, and then just use a search engine based on that.
Also, taught me some python and appscript. Currently learning and testing its capabilities in JavaScript teaching. And, yes I test out everything it gives me. It is best to output small blocks of code and lice it together. Hoping for the best and then, 3 years later finally create an app lol because that is on my end. Still working on an organization app. 80 percent accurate on following complete directions in this case.
