What is AI actually useful for right now?

[disneysteve]

I know there is a lot of buzz about AI and everyone is investing in it and companies are adding AI features to their products but so far, every use of it I've seen has been simply awful.

ebay has a feature to let sellers use AI to write their item descriptions. It's HORRIBLE. The AI descriptions are ridiculous, filled with nonsense and flowery language and completely lacking in the details that actually belong in an item description like brand and size and material and condition. Instead you get "this must-have item is crafted of high quality plastic and cardboard and is a perfect addition to any collection".

Now they've also introduced a feature to add AI-generated backgrounds to your photographs. I saw one for the first time a couple of days ago. It was a woman selling an article of clothing. Her photo was taken on a mannequin but the AI made it look like it was floating in midair and the background looked like the wall of an abandoned building. There were a couple of large smears of some sort of dirt or something on the wall and the floor looked like it was falling apart. The color gradient of the image totally detracted from the item itself as it was much too dark so the clothing sort of blended in too much. All around a terrible way to try and sell one's items.

The Facebook app has now changed the search function to an AI search and they've added AI content to people's posts. For example, I posted a restaurant review the other day in our local foodie group. Now when you go to my post, at the very top is an AI-generated summary of the responses and on other posts, right under many posts are several AI-generated links to possibly related information. It's distracting, irrelevant, and there's no way to turn it off.

So tell me, right now, today, is there anything that AI is being used for that is actually beneficial and helpful rather than making things worse like the examples I keep encountering. I have no doubt that some day AI will be used in good ways, but so far I only seem to be coming across the awful ones.

[ua_guy]

The revenue side of healthcare is increasingly using automation for routine tasks, but has yet to really dive into implementing predictive and generative AI. There's a couple of projects on the horizon for larger organizations which are actual AI. One is responding to payers when they request additional information about a healthcare claim; AI can compile the information based on what it knows about the type of request and the information needed and will produce it from the patient's chart. Another one is drafting responses for appeals to claim denials. AI is trained to produce support for those appeals. It is also being used to navigate payer phone trees so claim follow-up staff don't have to do that. Based on the status and information about the claim, it decides which options to press or what to say to either gain the information or connect staff to a live rep after it waits on hold. And that's all relatively boring...the clinical side I hear is pretty wild with the range of AI uses.

I can't think of anything AI at home that's useful. It all reminds me of Clippy, the artificially-intelligent paperclip assistant from Microsoft, 30 years ago. "It looks like you're writing a letter, would you like help with that?!" I haven't found a useful place for AI at home yet.

[kork13]

Most AI type systems are pretty poor if critical thought isn't applied to assess what it produces. The most meaningful use of it that I've come across is to use it as a prompt to help generate ideas to build upon.

For example, my wife has been in a doctoral program for the past couple years, but hates writing. She has used generative AI language models to help her to kickstart her writing or pull her out of writer's block, and done very well as a result. She needs something to help her get started, and the AI can do that for her.

With a well-crafted prompt to the AI, it can collect ideas & throw them together in a new way that provides you with a different way to look at a topic. So if you use AI as a helper, it can be valuable. If you try to gave it do your work for you, you'll be disappointed.

[myrdale]

All your personal fantasies of Hilary Clinton with Abraham Lincoln are now possible.

There is some really creepy stuff out there. Is that for the betterment of society? Absolutely not.

We are too the point now you should question everything you see. Photoshop has always been a thing, but now it's not even doctoring a photo, it is creating a photo from scratch.

Five years ago, I could hire a wax model sculpturer and a photographer to create the scene I listed above at a cost of probably $20,000, today I could ChatGBT it for free.

From a Freedom of Speech angle, political cartoonist have every right to write what ever they want. But in a world of hyper realistic AI I do question the limits of what you should be allowed to publish.

[ua_guy]

Nobody is above the law, including AI. Businesses are finding out the hard way when they're held liable for things their AI does. There's been a lot of discussion on AI and Copyright law...will be interesting to see how that evolves in the coming years.

[disneysteve]

We are too the point now you should question everything you see. .

And hear. The tech now exists to take a sample of your voice and use AI to recreate it to say whatever they want "you" to say. No longer do they need recording of you actually speaking each word that they can digitally chop up and put back together to say something else. Now just a sample of your voice saying something is enough to then make you say anything at all.

[disneysteve]

I will say that one cool use of AI was in the last Indiana Jones movie. They used AI along with other tech to "de-age" Harrison Ford for the flashback segment. I'm sure we'll see more and more stuff like that as the tech gets better and faster and cheaper.

[scfr]

I've started a bit of brainstorming for a road trip I want to make in the next year or two. I want to reach a certain destination, driving different routes there and back in order to see different scenery and sights.I had come up with what I thought was the "best" route but wasn't thrilled about it, so decided to try ChatGPT to see if it would come up with something better. My first attempt was a total fail (it took me WAY out of my way) so I started refining and after about 3 revisions (clarifying what I really wanted in a way that the AI could "understand", in part by de-emphasizing certain things) it hit on a route that I'm excited about. It will mean a little bit more driving overall but I'm much more excited about the scenery that we'll be able to take in and the towns we'll be able to visit. And it has allowed me (on my own without Chat GPT but thanks to its route suggestion) to possibly incorporate doing a small segment by rail for a minimal fee, adding another layer to the adventure. I'm a bit embarassed that I didn't come up with that route on my own! I feel like it served as a quite useful "collaborator."

[kork13]

I feel like it served as a quite useful "collaborator."

I think that's the perfect description of most AI capabilities right now -- used as a collaborator, they can be very useful.

[LivingAlmostLarge]

to generate ideas for dinner or shopping or anything

[crazyliblady]

I am a librarian and also publish articles in peer reviewed journals in my field. A colleague referred me to a tool called chatPDF. You can take any random PDF and just drop it into the tool and it gives you a basic summary. You can even ask it questions. I do not use the summaries to help me write my articles, however. I only use it as a time saving tool to gauge whether a particular article is worth pursuing as a source. If I decide it is worth looking at further, I then actually read the article.

[QuarterMillionMan]

"Images of future cities (meh)."