Useful Applications for AI

[amirm]

I know many of us worry about the negative consequences of AI, and worthiness of the responses it provides.

In the last few weeks, I have been extensively using both Gemini and my paid ChatGPT Pro subscription. Between the two, I have managed to get a lot done that would not have been easy or possible without them. So I thought I create this master thread on true usage scenarios that are positive. Please no complaints about AI or arguments. Just post what you used it for and how it was useful.

Here is a simple but shocking example. Was watching a Scifi TV series from 1970s on youtube called Space:1999. They have guest actors/actresses on it. I paused the video I was watching in Chrome and clicked on "Ask Gemini." I then typed in who the actress was. It instantly gave me the name which I verified online using image search. Finding that name in credits and searching each one to see if it is the right person would have taken an infinity.

Even more shocking, although created a misfire at first, was watching a reality show about Airport Security. This again dates back to 1970s/1980s. There was a scene with custom agent in it. I asked who that was and gave me a name. Did an image search and wasn't correct. I asked again and this time it identified the person correctly!

The fact that Gemini knows you are watching specific video helps a ton with context and not having to type more. Giving the same link for example to ChatGPT resulted in it complaining that it could not search youtube videos.

I have since routinely used the same method to get more data from the video I am watching.

So post what positive applications you have seen for AI.

 

[drodgers]

I work as a software engineer, and the applications there are wild. I haven't written code myself in more than a year. My team's delivery rate is up by ~400% over 12 months, with about 50% of it manually supervised (interactive claude code), and 50% fully automated human-just-says-yes-on-the-change-request.

The biggest effect I'm seeing is raised ambition - we're taking on more, much bigger projects because we know we can.

Personally I'm building tons of personal and home automation software that I previously would never have finished or wanted to maintain.

 

[Chr1]

Circle to Search.

Incredibly useful. Particularly when in combination with the ability to quickly photograph anything and then circle...
With video: Just pause the video you are watching and circle within that too. Also good for plain old text too.

Very useful indeed.

 

[101geo]

Got a working version of Camilladsp and some cnc cutting files for a sub project thanks to Claude pro. Still have no idea on how to use CLI and any Cad apps.

 

[audio_tony]

I use it frequently for PHP, Python and even electronics stuff. I can upload a schematic of an amplifier for example, and explain the issues I'm having and it'll provide suggestions (which are always correct).

I use Gemini mostly as I have a subscription.

I also use it to fix old family photographs - you have to be careful with your prompts, but it mostly does a fantastic job of renovating them.

 

[NTK]

Check out the handiwork by @pierre, the gentleman who also brought us the marvellous spinorama.org. His SotF (Sound of the Future) project is about 1 million lines of code now. One great guy assisted by AI.

AutoEQ for rooms and headphones

Release is available at https://github.com/pierreaubert/sotf/releases/ Currently available: - MacOS (arm): native app, tui app, systemwide - MacOS (x86): built but not tested - Linux (x86): appimage format (and tar.gz but untested) - Linux (arm): currently building - Windows (x86): .exe...

www.audiosciencereview.com

 

[SoonToBeRetired]

I've been using Gemini to "colorize" my old black and white negatives that I have scanned in, sometimes with wonderful results. Remarkably, it reproduces flesh tones very well, but otherwise it tends to favor blue and will use it every chance it gets. You also have to remind it to vary its color choices. But it is useful and above all, entertaining.

 

[davmol]

For an easy and optimistic watch, there was a great bit of TV late last year, following a UK comedian called Chris McCausland, who is blind, and he went to find out about some of the latest advancements in technology, including AI, and understand how it might benefit humanity. For example, the Meta AI glasses are (quite rightly) getting a lot of backlash due to invasion of privacy and users hiding the recording light on the frame, but this short documentary shows them in a very different light, highlighting the life changing benefits for someone like Chris.. It's a BBC show, so hopefully you'll be able to find it wherever you are (or possibly watch it via VPN):
Chris McCausland: Seeing into the Future

Also I'm two years into a PhD (looking at assurance of LLMs) and I recently went to a talk about how humans pass down wisdom through language, and what is happening in that arena today, with LLMs. I'm usually in the science end of talks, but this talk from the arts side was really enjoyable - a couple of my favourite gems:

Thamus expressing his dissatisfaction with the Theuth's invention of writing:
"What you have discovered is a recipe not for memory, but for reminder. And it is no true wisdom that you offer your disciples, but only its semblance, for by telling them of many things without teaching them you will make them seem to know much, while for the most part they know nothing, and as men filled, not with wisdom but with the conceit of wisdom, they will be a burden to their fellows.'"
Plato (c. 429-347 B.C.E), "Phaedrus" (c. 360 B.C.E.), 274c-275 b, R(eginald) Hackforth, transl., 1952.

Erasmus showing that "slop" was a problem long before AI:
"Yet the pursuit of speech like this involves considerable risk…. We find that a good many mortal men who make great efforts to achieve this godlike power of speech fall instead into mere glibness, which is both silly and offensive. They pile up a meaningless heap of words and expressions without any discrimination, and thus obscure the subject they are talking about, as well as belabouring the ears of their unfortunate audience.”
Copia: Foundations of the Abundant Style, 1512

I particularly enjoyed the speaker referring to Erasmus as the "original Large Latin Language Model".

 

[audio_tony]

I've been using Gemini to "colorize" my old black and white negatives that I have scanned in, sometimes with wonderful results. Remarkably, it reproduces flesh tones very well, but otherwise it tends to favor blue and will use it every chance it gets. You also have to remind it to vary its color choices. But it is useful and above all, entertaining.

I've also found that these prompts can be useful:

Restoration prompt:*

Restore this family group photograph while preserving the exact faces, clothing, poses, ages, body shapes, expressions, background, and composition. Remove visible cracks, dust, speckles, digital noise, compression artifacts, and discoloration. Improve clarity and facial detail naturally without changing identities. Correct the warm/yellow indoor color cast, balance skin tones, recover contrast, and slightly sharpen the image. Reduce glare and overexposure near the top-left light area while keeping the original room lighting believable. Keep the wallpaper pattern, table, flowers, clothing colors, boutonnières, and all people accurate. Make the final result look like a clean, natural, high-quality restored family photo, not AI-generated or overly smoothed.

Negative prompt:*

Do not change faces, do not make people younger or older, do not alter clothing, do not add or remove people, do not change the background, do not replace the wallpaper, do not create fake facial features, do not over-smooth skin, do not make it look painted, plastic, cartoonish, or modernized.

*obviously don't include those headings. I find these prompts to be quite useful - as some of my images were just completely made up from what I provided.

 

[youpassbutter]

I've also found that these prompts can be useful:

Restoration prompt:*

Restore this family group photograph while preserving the exact faces, clothing, poses, ages, body shapes, expressions, background, and composition. Remove visible cracks, dust, speckles, digital noise, compression artifacts, and discoloration. Improve clarity and facial detail naturally without changing identities. Correct the warm/yellow indoor color cast, balance skin tones, recover contrast, and slightly sharpen the image. Reduce glare and overexposure near the top-left light area while keeping the original room lighting believable. Keep the wallpaper pattern, table, flowers, clothing colors, boutonnières, and all people accurate. Make the final result look like a clean, natural, high-quality restored family photo, not AI-generated or overly smoothed.

Negative prompt:*

Do not change faces, do not make people younger or older, do not alter clothing, do not add or remove people, do not change the background, do not replace the wallpaper, do not create fake facial features, do not over-smooth skin, do not make it look painted, plastic, cartoonish, or modernized.

*obviously don't include those headings. I find these prompts to be quite useful - as some of my images were just completely made up from what I provided.

This is great, thanks for sharing. I would appreciate it if folks would share prompts they find useful for various tasks (or links to useful prompts). Thanks again.

 

[Martin]

I found myself using AI for more and more tasks and spent a few days last year comparing various AI implementations and settled on a ChatGPT Pro subscription. I have since found a myriad of uses for that subscription both in business and in my personal life. From simple image manipulation to vacation planning to product research to writing and interpreting technical documentation. I find it necessary to have at least a basic understanding of the subject your are tasking AI with to catch the occasional mistake / hallucination.

What I found AI most useful for is reading and summarizing government produced documents. ChatGPT is excellent at distilling hundreds of pages of verbose bureaucratic mumbo-jumbo into a concise document and salient points. If you limit it to documented information you give it I find it doesn't hallucinate or make mistakes.

AI is a great tool for processing huge amounts of text. Try giving it a terms and conditions text the next time you have the opportunity and asking it to make sense of it. I have found that interesting.

Martin

 

[Erici]

I have cancer and the ability to use AI to search for Clinical Trials of different drugs has been amazing! Compared to researching at the National Instute of Health website, the effort has been cut by 90%. I've found that there have been minimal hallucinations or strange results when the data exists somewhere online. When I load a blood report and ask a question, the results have been good.
Asking for a the comparison of different therapies, the results have been mostly adequate.
Asking for advice, not so good, as I would expect.

 

[NTTY]

End of last year, my boss asked me to look at the complains received from one of our subsidiaries about a cloud provider who updated their software and generated business disruption in few factories, mainly in US and Australia, with failure to resolve.
I only asked to receive the email thread, I did not know about the supplier nor the software. I asked the AI to find the master agreement we signed with this company and requested an analysis of the potential claims and remedies correlating the contract and content of the email, asking the AI to document its findings in a table, pointing to the exact section of the contract for the remedies, so I could check by myself. It suggested me to prepare the email to send to the supplier, and I said ok.
My boss initially gave me 1 week to do that, and I did it in 1 hour.

Claims and associated contractual remedies reached $1.5M.

If the AI makes my life easier, it also makes it way busier…

 

[JPA]

I know many of us worry about the negative consequences of AI, and worthiness of the responses it provides.

In the last few weeks, I have been extensively using both Gemini and my paid ChatGPT Pro subscription. Between the two, I have managed to get a lot done that would not have been easy or possible without them. So I thought I create this master thread on true usage scenarios that are positive. Please no complaints about AI or arguments. Just post what you used it for and how it was useful.

Here is a simple but shocking example. Was watching a Scifi TV series from 1970s on youtube called Space:1999. They have guest actors/actresses on it. I paused the video I was watching in Chrome and clicked on "Ask Gemini." I then typed in who the actress was. It instantly gave me the name which I verified online using image search. Finding that name in credits and searching each one to see if it is the right person would have taken an infinity.

Even more shocking, although created a misfire at first, was watching a reality show about Airport Security. This again dates back to 1970s/1980s. There was a scene with custom agent in it. I asked who that was and gave me a name. Did an image search and wasn't correct. I asked again and this time it identified the person correctly!

The fact that Gemini knows you are watching specific video helps a ton with context and not having to type more. Giving the same link for example to ChatGPT resulted in it complaining that it could not search youtube videos.

I have since routinely used the same method to get more data from the video I am watching.

So post what positive applications you have seen for AI.

Did you try IMDB? Using it I found a list of the full cast, with thumbnail pictures, in less than 30 seconds. I suspect that's the page ChatGPT Pro scraped to give you its answer.

 

[Turambar]

Documentation for a historical novel I'm writing. It speeds things a lot, though sometimes there are strange things that i need to check out (and sometimes find that they were strange, but true!)

 

[davmol]

This is great, thanks for sharing. I would appreciate it if folks would share prompts they find useful for various tasks (or links to useful prompts). Thanks again.

Small tip is you can get very good results by asking GPT (or whichever LLM you are using) to generate the prompt for you e.g. "I want to improve the colours and clarity in an old photo so please provide an effective prompt that I can use with GPT / Gemini / any-other-model".

 

[mpw]

OpenClaw to wipe out your whole harddrive and delete all you email while giving your banking account routing number to hackers !

 

[davmol]

I have cancer and the ability to use AI to search for Clinical Trials of different drugs has been amazing! Compared to researching at the National Instute of Health website, the effort has been cut by 90%. I've found that there have been minimal hallucinations or strange results when the data exists somewhere online. When I load a blood report and ask a question, the results have been good.
Asking for a the comparison of different therapies, the results have been mostly adequate.
Asking for advice, not so good, as I would expect.

Tools trained with academic text, such as as Elicit and Consensus, are extremely effective for research purposes like those you mentioned. I tend to use Elicit for open questions and detailed research reports e.g. "what are the most effective treatments for breast cancer", and I use Consensus for closed questions e.g. "is drug x an effective treatment for prostate cancer" (it literally tried to find the consensus if you ask closed questions). If you generate multiple reports from these, or have a lot of literature, you can attach it all in Google's NotebookLM and have it generate incredibly authentic podcasts, so you can listen then choose what to read in detail.

 

[marlin]

I've been using NotebookLM to help make sense of and answer questions about the DaVinci Resolve video editing software - https://notebooklm.google.com/notebook/fd870600-26ae-400c-940e-3879cbd7b0c2 - I figure it has saved me hours of going through incredibly dense reference materials. I also have a Rotel RX-504 receiver and I used NotebookLM to supplement my research into it's care, maintenance & repair - https://notebooklm.google.com/notebook/01786268-4214-482c-b9b2-64a8810ce71e

 

[Audiofire]

OpenAI GPT is great at translation as a large language model (LLM). Google Translate has more problems with context, slang and idioms from what I've seen. Duck.ai (from DuckDuckGo) is also powered by OpenAI and requires no login, but is bad at OCR. I have used ChatGPT OCR for electronic circuit analysis, which works with some hallucinations, so critical thinking is needed.