Are you bullish about "AI" or skeptical?
Due to a lot of my comments being on the skeptical side when it comes to big-tech and these tools, I thought I would redress the balance somewhat with a positive story or two.
The screenshot above is from my web app that I built that uses the Google Cloud API and the OpenAI API to ensure I don't waste my time on clickbait.
You see, I have so many subscriptions now, and YouTube does such a bad job of showing me who I am subscribed to, that I can't possibly watch everything even if I wanted to.
Then there are the the videos like the one in the screenshot where the presenter covers several news stories and does not use chapter markers. Many of them cover the same stuff, just with better or worse thumbnails and titles to draw you in.
So what I do is use the Google API to check in on my subscriptions (in batches, with caching because you very quickly get rate-limited).
For each channel I then see what the latest video(s) is/are. Again, cached.
Then when I go into one of the videos I have the option to get the transcript, summarize the transcript (with a choice of AI model, based on token size, complexity, and cost), and all of this is written to a MySQL database (so that I can run this on my dev or web server and not waste API capacity while testing).
It works pretty well so far.
One of the problems I have had is my web server IP address must be on a block list because I couldn't retrieve transcripts. I ended up using some spare web space and putting a script on there which this app uses as a proxy. Dirty but it works and there's only one user (due to API restrictions and costs I have to keep this one for me only otherwise it would cost a lot $$$ over time)
Next thing I would like to do is make any videos actually worth watching get added to my Watch Next queue so I can still veg on the sofa and watch my videos with a nice brew! ☕️
I think AI, like NFTs, has use cases that are overshadowed by hype and bad implementations. People using AI as a substitute for web searches must suffer hallucinations at an incredible rate. I've had poor results, at least. Generative AI for images, video, and writing is generally terrible, although generating an outline or getting prompts can work out. It's hit and miss, but I'm mostly in the skeptic camp.
I'd like AI to automatically filter out ever AI-voiced video on all platforms, but noooo... Facebook adds useless search prompts under posts, and Youtube is hell.
The hallucination issue has gotten really bad. A family got poisoned after following “foraging” advice in an AI generated kindle book.
Unfortunately people are using the tools to get out of being involved instead of to help in the effort.
I'll consider using them if they can save me time, but some of the search stuff they do is just wrong. I am not into them for 'creating' content. Humans do that better. I've not really used them for coding yet, but we have something at work for that as we are not allowed to put any company info into public services. The next generation of programmers will have a different life to me as I near retirement.
I keep trying things out because in my day job I'm expected to have answers. But as I say in the post, I'm highly skeptical most of the time and I don't like the industrial strength plagiarism either as a creator or as a consumer.
That said, even though up to age 16 I wanted to be a professional illustrator, I did use gen ai for the fantasy character art in my story so I'm not innocent either.