The other day I couldn’t find something on a bookshelf so I took a picture of the whole wall and fed it to the robots, then asked the robots where the stupid book was. It told me.
@jouest@Kyeh This is a best-case use of the pattern-matching capabilities. The task can be well-defined, the result can be easily evaluated for correctness by the human requesting the assistance, and the consequences of error are minor. Self-driving car tasks, on the other hand, are not possible to define well, may go wahooni-shaped without warning, and can prove instantly fatal to both the one requesting the assistance and the tourists on the bus going over the guard rail.
@Kyeh@werehatrack my wife just got a new car with very limited self-driving capabilities - glorified lane keep assist and radar cruise. She’s halfway through the convoluted video tutorial and all I could think was “how in the HELL is this preferable to just driving the car??”
@jouest We rented a Corolla in NZ that featured all of that in 2019, and I had to turn all of it off because their narrow lanes and roads just drove it nuts.
@jouest@werehatrack Both of our cars (a Mazda and a Toyota) have lane keep assist features. My wife disabled the feature soon after getting the vehicle (the Toyota) - she said it wanted to stay closer to the center line than she was comfortable. I’ve left it on in my car (the Mazda) so far - it isn’t too aggressive but if I stray too often the car suggests that I take a break.
Ever the rebel, I get some small twisted pleasure from ignoring its advice.
I have been trying to avoid the use of AI, but it appears that all of the search engines are now making it difficult to not use it.
For a discussion of the ramifications of it, I recommend Abigail Thorn’s video on the subject, among many that get into the reality of its current iterations and the implications of its deployment.
The company I work for developed their own version of ChatGPT with a variety of modules that keep growing. They are trying to develop one that will translate clients’ written requirements into programming code. That would certainly cut-down on staff if successful. I have used ChatGPT to help make business requirements a bit more refined or break down into a bulleted list. I also asked it for advice on natural products for getting rid of carpenter ants. I thanked it for its solutions and it wished me good luck with my ant problem. I thought that was sweet, yet scary.
@heartny@jouest Part of the problem is that client requirements read like “Darmok and Jalad at Tenagra.” They only make sense if you understand the entire context on which they are based. The clients, of course, work entirely within that context. So what they say makes perfect sense to them. They might as well be speaking Jovitos.
i use it to explore new topics as I become familiar with them (by reading non-AI literature).
i also use it to bounce ideas off in domains where I am an expert and can tell terrible results from decent ones, and need help in generating and exploring possibilities.
@zippyus it’s always scary to explore a topic that you know a lot about and realize how full of crap people are and wonder how much you’ve read about other stuff that was complete nonsense…
I wish they’d stop shilling AI as a replacement for something we can already do better without it. Example: basic Internet searches. I don’t need AI to tell me what I can see with a telescope from wherever I’m camping, to borrow an example from a current “Meta” (Facebook) commercial. Google search can already tell me how to do that very easily.
The whole charade of companies literally shoving AI anywhere they can find to cram it is getting very taxing. I’m sick of seeing it EVERYWHERE. The fact that there are companies out there right now exploring ways to insert AI into the process of driving a car is legitimately frightening.
@jouest@PooltoyWolf We need AIs with Asimov’s 3 laws of robotics. Then we need to remind the driving software that allowing a vehicle (even one with no living occupant) to crash is committing suicide. And then we will likely need a few robot psychiatrists to deal with suicidal robots, but it’s not a perfect world.
@jouest@PooltoyWolf@rockblossom First we need AIs that can reliably recognize that living things are living, and that this is significant. We have not approached this yet. No one has yet come up with a way to get an AI to “understand” enough to make such a generalization.
@PooltoyWolf to be fair, if I had a trained eunuch live with me for five years and anticipate my needs the next five, that’s a great level of customization. However mass produced AI are trained by corporations first, you 2nd; there is going to be some “unlearning” that might be blocked by the manufacturer. (Also see: smartphone customizations)
There was a vigilante crime-procedural/sci-fi called “Person of Interest” back in 2010 about an infirm billionaire hiring an ex-agent using information gleamed from an omniscient surveillance AI. The AI was a “good guy” in so much that it provided almost Minority Report level of clairvoyance, but could not directly communicate what it sees rather it passed on clues to the humans.
But the AI wasn’t made good overnight; plenty of flashbacks show that morality took place over hundreds of tests over several years with plenty of “rogue” splinters. The heroes often question if the current AI is really on their side or just smart enough to play a really long game to avoid deletion.
It writes problematic code, but it writes it quick… So I get it to generate code for me and I look through it and fix all its mistakes.
I’ve used it to generate lists of names for me.
I’ve used it to help meal plan.
I’ve used it to suggest musicals to listen to based on others I liked.
I’ve used it to do math problems I didn’t want to do. (Calculating volumes of unusual shapes)
I’ve used it for general information look up.
I used it to help randomize settings for an online game I hosted.
I hate AI, that it exists… But since it does exist and you can’t put the genie back in the bottle, I’m going to use it and do use it.
Kinda feel the same way about smart phones and the Internet… Very useful and have revolutionized our lives, but feel like world would be better without them… But they exist, so you bet I’m using them.
The other day I couldn’t find something on a bookshelf so I took a picture of the whole wall and fed it to the robots, then asked the robots where the stupid book was. It told me.
@jouest Wow!
@Kyeh right? Mildly interesting!
@jouest And useful!
@jouest @Kyeh This is a best-case use of the pattern-matching capabilities. The task can be well-defined, the result can be easily evaluated for correctness by the human requesting the assistance, and the consequences of error are minor. Self-driving car tasks, on the other hand, are not possible to define well, may go wahooni-shaped without warning, and can prove instantly fatal to both the one requesting the assistance and the tourists on the bus going over the guard rail.
@Kyeh @werehatrack my wife just got a new car with very limited self-driving capabilities - glorified lane keep assist and radar cruise. She’s halfway through the convoluted video tutorial and all I could think was “how in the HELL is this preferable to just driving the car??”
@jouest We rented a Corolla in NZ that featured all of that in 2019, and I had to turn all of it off because their narrow lanes and roads just drove it nuts.
@jouest @werehatrack Both of our cars (a Mazda and a Toyota) have lane keep assist features. My wife disabled the feature soon after getting the vehicle (the Toyota) - she said it wanted to stay closer to the center line than she was comfortable. I’ve left it on in my car (the Mazda) so far - it isn’t too aggressive but if I stray too often the car suggests that I take a break.
Ever the rebel, I get some small twisted pleasure from ignoring its advice.
I have been trying to avoid the use of AI, but it appears that all of the search engines are now making it difficult to not use it.
For a discussion of the ramifications of it, I recommend Abigail Thorn’s video on the subject, among many that get into the reality of its current iterations and the implications of its deployment.
The company I work for developed their own version of ChatGPT with a variety of modules that keep growing. They are trying to develop one that will translate clients’ written requirements into programming code. That would certainly cut-down on staff if successful. I have used ChatGPT to help make business requirements a bit more refined or break down into a bulleted list. I also asked it for advice on natural products for getting rid of carpenter ants. I thanked it for its solutions and it wished me good luck with my ant problem. I thought that was sweet, yet scary.
@heartny As far as I know, this is still true:
Eventually that may change.
@heartny based on my experience of interacting at the intersection of IT and the rest of the business, your AI might lose its robot mind.
@jouest I think this may be an amusing understatement of the result.
@heartny @jouest Part of the problem is that client requirements read like “Darmok and Jalad at Tenagra.” They only make sense if you understand the entire context on which they are based. The clients, of course, work entirely within that context. So what they say makes perfect sense to them. They might as well be speaking Jovitos.
Just read a story this morning about a NJ kid that is using AI to kill lantern flies.
https://www.yahoo.com/news/high-schooler-develops-innovative-solution-023000005.html
But this is what I hate about AI:
Elon Musk Shares Kamala Harris Deepfake
https://www.rollingstone.com/politics/politics-news/elon-musk-kamala-harris-deepfake-1235069564/
@Kyeh Agreed.
i use it to explore new topics as I become familiar with them (by reading non-AI literature).
i also use it to bounce ideas off in domains where I am an expert and can tell terrible results from decent ones, and need help in generating and exploring possibilities.
@zippyus it’s always scary to explore a topic that you know a lot about and realize how full of crap people are and wonder how much you’ve read about other stuff that was complete nonsense…
@jouest i hear you. realizing that and going “oh, oh no” is a special moment.
I found this broadcast company’s AI policy to seem straightforward and what I’d hope for
https://gray.tv/uploads/documents/Gray-Television-AI-Policy.pdf
OTH, the humans there seem to bring along their own biases and occasional inaccuracies to their reporting.
I wish they’d stop shilling AI as a replacement for something we can already do better without it. Example: basic Internet searches. I don’t need AI to tell me what I can see with a telescope from wherever I’m camping, to borrow an example from a current “Meta” (Facebook) commercial. Google search can already tell me how to do that very easily.
The whole charade of companies literally shoving AI anywhere they can find to cram it is getting very taxing. I’m sick of seeing it EVERYWHERE. The fact that there are companies out there right now exploring ways to insert AI into the process of driving a car is legitimately frightening.
@PooltoyWolf we use pictures of traffic as captchas because robots can’t do it and then let robots drive cars
Please realize CAPTCHA is a fellow robot like you.
@jouest @PooltoyWolf We need AIs with Asimov’s 3 laws of robotics. Then we need to remind the driving software that allowing a vehicle (even one with no living occupant) to crash is committing suicide. And then we will likely need a few robot psychiatrists to deal with suicidal robots, but it’s not a perfect world.
@jouest Very good point…
@jouest @PooltoyWolf @rockblossom First we need AIs that can reliably recognize that living things are living, and that this is significant. We have not approached this yet. No one has yet come up with a way to get an AI to “understand” enough to make such a generalization.
@PooltoyWolf to be fair, if I had a trained eunuch live with me for five years and anticipate my needs the next five, that’s a great level of customization. However mass produced AI are trained by corporations first, you 2nd; there is going to be some “unlearning” that might be blocked by the manufacturer. (Also see: smartphone customizations)
There was a vigilante crime-procedural/sci-fi called “Person of Interest” back in 2010 about an infirm billionaire hiring an ex-agent using information gleamed from an omniscient surveillance AI. The AI was a “good guy” in so much that it provided almost Minority Report level of clairvoyance, but could not directly communicate what it sees rather it passed on clues to the humans.
But the AI wasn’t made good overnight; plenty of flashbacks show that morality took place over hundreds of tests over several years with plenty of “rogue” splinters. The heroes often question if the current AI is really on their side or just smart enough to play a really long game to avoid deletion.
All the time.
It writes problematic code, but it writes it quick… So I get it to generate code for me and I look through it and fix all its mistakes.
I’ve used it to generate lists of names for me.
I’ve used it to help meal plan.
I’ve used it to suggest musicals to listen to based on others I liked.
I’ve used it to do math problems I didn’t want to do. (Calculating volumes of unusual shapes)
I’ve used it for general information look up.
I used it to help randomize settings for an online game I hosted.
I hate AI, that it exists… But since it does exist and you can’t put the genie back in the bottle, I’m going to use it and do use it.
Kinda feel the same way about smart phones and the Internet… Very useful and have revolutionized our lives, but feel like world would be better without them… But they exist, so you bet I’m using them.
@OnionSoup ^ this take right here