The technology itself is novel and cool. Its the complete and utter meltdown of all tech companies into brainless hype machines that is harmful, which is course, is a function of capitalist incentive and the need for the tech industry to come out with some new paradigm shifting innovation every decade. A normal, healthy society would have been able to leverage machine learning and LLM technology where its most useful, like parsing large amounts of data, or running a local instance on your computer to ask a few questions, etc. We wouldn't see LLMs in every text editor, pencilcase and pair on sneakers but these snake oil salesmen who run the US economy are absolutely desperate for a new paradigm shift so they can keep making exponentially more money.

The thing is, we don't need to build these datacenters siphoning comically evil amounts of energy from the grid and making personal compute a thing of the past. Average everyday person doesn't need cloud compute, they can run a local 4b parameter (very, very small) model on their laptop or phone if they need to ask chatgpt to make them a workout routine or to ask them who won the 1918 world series. But these fucking cretins don't care, that's not the point, they are in this because it's a golden ticket to growth city and once they cash their check they don't give one hot fuck about the human-spirit-stealing-machine they built.

TLDR: our society is broken and that's why we keep getting the shittiest, lowest-common-denominator version of everything. everything has to suck by definition because that's the only version that the system we built will allow.

I'm a therapist. I use HIPAA compliant AI to generate my (editable) case notes for my sessions now. Not only is it a huge time saver to simply edit a generated note as opposed to making one from scratch, but in many cases it takes more detailed notes, including quotes from clients.

I have heard of other therapists and medical doctors also using AI to help with diagnosing.

The danger is when therapistsdon't review the content to check for accuracy. Because occasionally it will generate something not really reflective of what the therapist might have been doing, or it might lack detail that the therapist might have otherwise inclused. But more often the stuff it comes up with is surprisingly accurate.And editing is even easier when you can just tell the AI something like, "include more details about how the client noticed their pattern of putting their own feelings last," and it just does what you asked. You don't necessarily have to edit manually, though you can.

I have a friend at work that does a lot of video. He films weddings, music videos etc. and is making a pilot for Netflix. He uses AI to go through all his footage and tag it according to content. E.g. if he needs a clip of birds, he can just search ‘birds’ and it will pull up all relevant footage. Incredibly useful.

And if ChatGPT made a mistake? How would you know before it’s to late ?

For every small benefit, there are disastrous mistakes. We shouldn't discuss one without the other:

https://tech.co/news/list-ai-failures-mistakes-errors

March 2026

• Police used AI facial recognition to arrest a Tennessee woman for crimes committed in a state she says she’s never visited

February 2026

• Health advice given by AI chatbots is frequently wrong, says new study

January 2026

• Study reveals that fixing AI mistakes takes up to 40% of the time that it saves

• An AI tool used by ICE to identify applicants with previous law enforcement experience falsely flagged applicants with no such experience, leading to the placement of unqualified recruits in field offices.

December 2025

• AI mistakes clarinet for gun at Florida school

November 2025

• Google Antigravity deletes entire content of user’s computer drive

• Report finds AI hallucinations in 490 court filings from the past six months

October 2025

• Teenager handcuffed after AI mistakes Doritos packet for gun

• Lawyer submits AI-assisted court filing with fake citations

• Man follows ChatGPT advice over stopping eating salt, develops rare condition. The man was hospitalized, sectioned, and eventually treated for psychosis. He tried to escape the hospital within 24 hours of being admitted.

• ChatGPT-5 jailbroken with 24 hours of release

July 2025

• AI Coding app deletes entire company database

• McDonald’s AI chatbot error exposes data of 64 million job applicants

• AI program is tasked with running a small shop, goes insane, claims to be human

• Apple Intelligence falsely presents BBC headline

... and it just keeps going.

Anything that's fuzzy and impossible to automate with traditional algorithms, but that also has a reasonably high tolerance for error. It just makes up stuff a good portion of the time, you see.

However, I’ve found some benefits with AI. For example, I’m chatting with ChatGPT on credit cards, because it is something I may lean towards getting into. It’s helping me better understand than most people have tried explaining to me. Simply because it is giving me a more stream-lined response than people just beating the bush.

Watch out, personal finance is not one of those things.

rubberducking for those with social anxiety. Also small friction to get surface level answers that normally took digging from multiple sources.

it's a study monster that initially wiped chegg, duolingo, sparknotes etc. The double edge is that people forgot how to take notes, learn fundamentals to handle complex problems.

An amazing use for it in audio engineering is for feedback suppression. The old way to give yourself more headroom required you to sit there and turn up the gain until feedback happens and cut that frequency. Now you just turn on the feedback suppression and it does all that for you on the fly. It's game changing for live sound, every major venue has it now.

Accesibility.

I agree there is a lot of annoying hype. However i also agree there are some specific use cases where it can be helpful.

I for one find it handy some times when i am writing bash scripts to do things on my system. I obviously check them before running but it does save time.

Although i do recommend running models locally if possible as it is obviously preferable from a privacy and cost standpoint.

Running automated hacking and blackmail campaigns against AI companies.

I went to my local neighborhood association because I wanted to improve where I live. I was elected president of the association a couple months later, mostly because no one else wanted to do it. It's a fairly poor part of a medium sized city in the U.S.

I've been using AI (running locally on a computer I built that isn't connected to the internet, to reduce harm to the environment) to apply for grants, plan events and help me run the meetings.

It is actually perfect for the job. Saying that as someone who thinks AI is mostly hype and useless for the majority of its current common uses these days. I feed it the text from city grant applications or ask it to make a poster to increase attendance and it's saved me a lot of time. Without it, being someone diagnosed ADHD, I would not have been able to do most of the stuff I have accomplished so far.

Chatbots? Basically nothing. Any interaction I have with one leads to spending more time verifying its output, inevitably finding many mistakes, and eventually finding a primary source for what I'm actually looking for. The best actual impact it has is forcing me to narrow down my nebulous question into what I actually specifically want, but the bot itself is contributing very little to that.

Neutral nets in general have limited real usefulness in analyzing large batches of data when other purpose-built analysis software doesn't exist.

"AI" is a misnomer and there is absolutely zero evidence to suggest that we're even on a path toward actual AI, sometimes called AGI, though they're also changing that to just mean a profitable LLM which is fucking hilarious.

Any task you use a bot to do, you will become worse at that task. For mass data analysis, that's fine, poring over reams of data is already a skill that other technology has largely obsoleted. But using it to do research, to read or write for you, or god forbid to make actual decisions and think for you, are very slippery slopes that are already causing a lot of the general public to seriously erode their basic mental capabilities.

I was sitting in a restaurant the other day and staring at the menu. It was Italian and none of the things made sense. Too wordy and not clear what was meat and what was fancy cheese. The waiter was utterly useless - too busy to help and when present, not answering my questions about what would be a good simple pasta in white sauce.

I took a photo and asked Claude what’s a good white sauce pasta which would be like Alfredo.

It found two options I hadn’t even looked at. AI is good at sorting through complexity. But I don’t just mean AI as in LLMs. It needs a lot more tools and knowledge to be useful. So what you need is a smart system which may or may not have AI as a component.

Curating massive music libraries. I've been using a small embedding model to organise my music for DJing, and being able to generate a t-sne plot clustered on perceptual similarity has been wonderfully useful.

I've also found CLIP models useful for searching videos, just embed a screenshot every couple of min of footage and query with a description of the scene.

And as bad as generated subtitles can be, when the only other option is nothing at all they are pretty nice to have.

Most of my qualms with AI aren't in the usage of AI, but in its creation (water usage, mass layoffs, etc.—you've heard it all before).

To me it's like asking "What are some good uses for slaves?" (An extreme example to show the point, I'm not trying to say AI is the same as slavery).

Like yeah I could find good uses for it, but should it exist in the first place?

Not a hot dog

Honestly, Google Search has been better the last couple years after spending the previous twenty years getting consistently worse.

Most of what I use Google for is trivial. Like how old is a certain actor, or why was this author canceled, or what does this item do in a video game?

It's great for those things. Especially the video game stuff. I don't want to watch a 10 minute video just to get a discrete answer, and now I don't have to.

I can even ask it for spoiler-free hints on a particular puzzle, and most of the time it gives me something useful.

LLMs tend to be a “jack of all trades, master of none”. You are likely to find them useful for helping you with something you are inexperienced at, but not at something you are an expert in. However, because they lie a lot, it’s best to double-check your information, but the LLM can still be helpful with the ”you don’t know what you don’t know” issue.

seems to be decent for OCR, maybe also speech recognition. i hear it's okay for finding some concept you can explain abstractly but don't know the exact word for, but haven't tried this personally.

Some hospitals use AI to scan patients and find signs of illness before it becomes a problem. I'd say that's a pretty good use of AI.

Reading TOS and clicking off all the privacy options on the cookie popup

I had a project of markdown files. About 400 of them, with about 1200 plus links in them.

The original filenames were changed. The links no longer worked.

The LLM went through each link, and found the new one, based on filename and file content, using its ability to recognise patterns, words, etc etc.

Absolutely saved me maybe a couple of days of manual pain labour, and all done in about 10 minutes.

This is the kind of thing I use it for. Horrible repetitive processes.

I occasionally use it to find links to VODs for esports tournaments. Asking it to only link the specific game I want with no other summarization is a way to find them without spoilers (like when youtube "helpfully" suggests the last game of the grand finals of the tournament as a search result).

I'm chatting with Le Chat about things I wouldn't ask a friend and wouldn't trust a stranger about. It brings up things I wouldn't have thought of.

I actually find it pretty helpful for tech support stuff. It doesn't always get it right, but it's usually at least in the right general area and TBH it beats going through endless forums where the answer is buried among 8 pages of people bickering about nothing, or those ones where someone has your exact problem and then replies "nm I fixed it" and doesn't say what they did.

neuro-sama

In computational biology / biotechnology, LLMs are being trained on biological sequences and can then be used to generate new genes or genetic variants. These genes can be placed into bacteria who are then fed with e.g. sugar to make them produce various valuable molecules from renewable resources instead of from crude oil using conventional chemistry. There is also work on enabling plastic biodegradation this way.

Well, I know its quite specific, but nothing beats AI at stereo matching and depthmap generation and that's important in many fields.

It's ok for very furnace level exploration. Like 100 level stuff. If it's something you'd google and easily find in an article it's likely to do a ok job.

I've also found it's good for tedious straightforward tasks. Anything that would be uncomfortable or timely or automate manually. Best for one offs.

I've also found it's extremly good for translation, which was it's originally use.

Karrot is a used item app that has a feature where you take a picture of an item and it IDs the item and tells you what it's worth. It's pretty impressive. It could ID my houseplants better than some dedicated plant ID apps I've used. It's not great with one of a kind items, but otherwise it's surprisingly accurate.

It’s very helpful for neurodivergent people - helps you figure out who you are and what you want, how you think, learn and work best, identify your obstacles and help you overcome them, understand your neurodivergency and compare it to how neurotypical people think. It’s fantastic at generating ideas that you then test out. The ideas that it gives you are based on how you actually function, so often times they’re valid.

Learning, exploring concepts and ideas.

Vibe coding slop you don't need to work in production

I have a script that uses yt-dlp to get subtitles off a YouTube video and summarises the main points for me with a language model so that I don’t have to watch a 20 minute top10 list video that could’ve been a buzzfeed article.

The whole thing is fully vibe engineered too.

Everything Kitboga has used it for and is currently doing with it. Hilarious AND interesting!

Writing and fact checking ONLY the most basic concepts and common information that is found multiple times and in multiple places online (eg. It's strongly reenforced and verified in the training data and has/will be the same for a long time).

Mass formatting, changing formats, changing language, and decoding via common methods.

Pitching "what you mean, but can't remember the name for".

...and that's about it.

I’m good with asking DuckDuckGo for help with a game and AI scrapes the game sites to just give me the answer. You know, sites like IGN and Dutch with popups and ads and other trash… and AI just steps past all that shit and gets me the answer. I’m fine with that. All these sites are just scraping guides and such, I don’t mind if they get stepped on.

Like, tell me how to save both the Geth or the Quarians in Mass Effect 3. So if you don’t know, Mass Effect was this kinda mediocre space shooter with magic, and it was fun… but then you get to the second one and it remembers all your choices, and there’s so much to do. And the Geth were the bad guys in the first one. And if you’re playing a male character, the Quarian is this hot middle eastern type lady you can romance. Then you meet a nice Geth (it’s a long story and the game is longer) and by the third one, you do this mission, it’s a lot of bullshit, then this bullshit boss fight, then you get put on the spot. I’m not going to spoil it, but for like 99.99% of us, one of your crew dies. And it sucks. But if you made some very specific choices, going back to the very first one… you can save both. I’m talking like 100+ hours since you made a mistake that didn’t even matter then, but it locks you out of saving one of your people. Stuff like that.

You could ask it what's in an image, then do your own research on what it tells you. I used it to identify a math function from a meme.

It can also just be fun to toy with. I personally like to play with image interrogation in StableUI, then put the generated caption into an img2text. Like some sort of alternate reality image. It gives pretty funny results sometimes.

Probably one of the best use cases I've seen for machine learning ( not the other LLM mega junk ) has been a recent game used it to allow the PVP/E CPU enemies to adjust to the map terrain.

There are probably much better uses for machine learning, but I haven't personally seen them AFAIK nor can I think of any others since my brain is falling asleep and I'm struggling to stay awake right now.

If your really bad at coding like me then its useful for debugging when I'm running short on time, but I always prefer to search it up.

I've been pretty happy with OCR.

Pixel phones can monitor phone calls for scam conversations (it runs locally on the phone, so audio doesn't get saved or uploaded).

The relevance for me personally is whether or not they can be useful for programming, and if they're accessible to run locally. I'm not interested in feeding my data to a datacentre. My AMD GPU also doesn't support ROCm so LLMs run slow as fuck for me. So, generally, I avoid them.

LLMs consistently produce lower quality, less correct, and less secure code than humans. However, they do seem to be getting better. I might be open to using them to generate unit tests if only they would run faster on my PC. I tried deepseek, llama3.1, and codellama; all take like an hour+ to answer a programming question given that they are just using my CPU, as my GPU doesn't support ROCm. So really not feasible for anything.

Depending on what you count as AI, I think some of the long-existing predictive ML like autosuggestions based on learning your input patterns are fine and helpful. And maybe if I get a supported GPU I won't mind using local LLMs for some things. But generally I'm not dying to use them. I can do things myself.

I regularly use CoPilot to search Microsoft documentation for me. E.g. I needed to find a particular interface in Entra and couldn't remember where it was. So, I asked CoPilot and it got me to the right spot. I've thought about asking it about Microsoft licensing, but I figure that might result in CoPilot becoming self aware enough to kill itself.

I also use a number of AI agents built into the cybersecurity tools I use on a daily basis. Generally stuff along the lines of "find all the cases related to this system/IP/user/etc" type queries. It's also good for questions like "how do I tune this alert" so I don't have to remember whatever bullshit process this vendor put together for tuning false positives. Our primary SIEM/SOAR tool has an AI which does initial triage and investigation work and it's not terrible. It struggles with correlations for more complex events, usually highlighting events which have no bearing on the event in question. But, it often provides a good first pass and description our first line analysts can use to start a real investigation.

AI is a tool. And like a lot of tools, it has it's benefits and limitations. The problem is we're still figuring all those out and the people marketing these tools don't want to admit to the limitations and they over-sell the benefits, then blame the user when those benefits don't materialize. Given how much modern economies are based on information and knowledge, I do expect AI to have some lasting impact, but I also expect that we'll adapt and it will just be another way of getting things done in a generation or two.

I use it to extract structured data out of an unstructured source were programmatic approaches would fail.

You need to differentiate between generative AI, NLP, machine learning, etc. Your question is pretty much entirely pointless otherwise.

I've used it to summarize long documents.

1. Drafts, coming up with ideas you may not have thought of initially
2. Grading whatever you write against the quality of other trained material, especially research papers (LLMs tend to be trained more on those material)
3. For LLMs that search the internet/cite sources, it can be a more powerful search engine. Enter a keyword, the LLM can guess other semantic terms related to your search prompt/terms and refine the results that way. The additional search results cited matter to me more than the response itself (which can hallucinate, but have a niche role as a wildcard if you're the type to pattern-match concepts into new ideas)

It's good as an assistant.

In science, pubmed searches only terms in titles and abstracts, but LLMs will specifically search every figure and supplementary file. But you have to manually confirm because sometimes they just make shit up.

Language translations.

I think that's about it.