On the topic of chatbots...

[crfriend]

I don't know if this was changed just for me, but the time I have to wait to make a second post is ridiculously long. The site now makes me wait from 10-11 minutes:

That was set to attempt to quell an entirely and inappropriately "overenthusiastic" new user who was flooding the place with largely content-free posts. The moderation staff opted to sit on the matter for a few days and then came in for a blast about "censorship!" at which point the posts were summarily deleted and the "user" disabled.

This happens from time to time. The mods are used to it; the community never even sees it. This is by design.

[moonshadow]

don't know if this was changed just for me, but the time I have to wait to make a second post is ridiculously long. The site now makes me wait from 10-11 minutes

I've noticed that too. I figured it was in response to the chatbot issue. Hopefully the measure is temporary.

There have been a few times I was not able to make a second post and actually forgot what I was going to say.

[bikesaurus]

I'm glad it's not just me with the 10 minute wait . Like you moon, I forget what I'm going to write in between replies.

[crfriend]

I've since pulled it back to five minutes.

[Fred in Skirts]

I have not noticed it as it normally takes me 6 to 7 min to write a reply and get it right. You know fix errors and the like and sometimes just to give an answer that is not getting someone mad at me.
So as I never make posts closer than 5 min It has not bothered me at all!

[Uncle Al]

I received this info from AVG.com, in my email. I use AVG.com products.
I got side-tracked and didn't get it posted to this thread.
I hope the info helps.

Is ChatGPT's use of people's data even legal?

Uncle Al

[moonshadow]

I have a feeling these A.I. chatbots,, and even A.I. in general might wind up becoming something we all wish we could uninvent.

It's going to be damned near impossible to tell fact from fiction before long, hell we're practically there ready.

[Uncle Al]

It's going to be damned near impossible to tell fact from fiction before long, hell we're practically there ready.

I totally agree %

Uncle Al

[rode_kater]

Is ChatGPT's use of people's data even legal?

Ugh, this article (deliberately or otherwise) confuses two meanings of personal data: data produced by a person and data about a person. The latter is what is protect by things like the GDPR. However, your personal website about trains is not personal data and since you published it on the internet, you really deliberately gave up all control. You don't publish something if you don't want people to read it.

If the makers are smart they'll have scrubbed the input for anything looking like addresses, phone numbers, CC numbers, etc. Also, people seem to be under the impression that LLMs are actually storing copies of their input, which is incorrect. LLMs hallucinate all the time, how do you distinguish between a hallucination and something you didn't want on the internet?

I think people need to get used to the idea of a computer having a memory as fallible as human memory. It's just not how we've experienced computers so far. Just like the "right to be forgotten" doesn't erase things from people's brains.

[crfriend]

I think people need to get used to the idea of a computer having a memory as fallible as human memory. It's just not how we've experienced computers so far. Just like the "right to be forgotten" doesn't erase things from people's brains.

If the things are not "better" at things than we as humans are, then what's the point in them existing?

My main worry is that the damned things are going to dumb the population down even more than it's already been. If the damned things are in use, then said use should be prominently flagged; otherwise they're the equivalent of the human imbecile at a cocktail party who interrupts every conversation and adds precisely nothing save interrupts for the conversants.

[rode_kater]

I think people need to get used to the idea of a computer having a memory as fallible as human memory. It's just not how we've experienced computers so far. Just like the "right to be forgotten" doesn't erase things from people's brains.

If the things are not "better" at things than we as humans are, then what's the point in them existing?

Well that's the interesting question isn't it. The only thing they have going for them is that they're faster and infinitely customisable by lay-people.

I guess it's a matter of perspective though. I've been following the whole AI development for decades and the whole problem is how to deal with natural language has been one of the hard problems all that time. So for me, this is amazing stuff. Being able to examine a piece of text and determine who is the subject or object of a verb, who the pronouns, word like "this", "that", "there", etc has always been a big research topic and now all of a sudden we have something that can actually "understand" these kinds of things.

This is super important, because people use these words all the time. And so having a computer be able to "understand" these kinds of sentences is super useful. However, by themselves they are just a step up, a tool to improve our interaction with computers.

They can also generate text, which is nice and has all sorts of uses but for me, that's not the important part. There's a lot of research going on now to understand why LLMs are able to understand language like this. And yeah, people are going to use them for all sorts of good and bad purposes. That's technology for you.

[crfriend]

Well that's the interesting question isn't it. The only thing they have going for them is that they're faster and infinitely customisable by lay-people.

Raw speed is questionable in applications where accuracy is paramount -- and vital where an ability to discriminate between truth and fiction is important. Since chatbots cannot perform the latter facet, that makes them inherently prone to spreading misinformation or, worse, disinformation with a "veracity" that the layman might misinterpret.

They're also not very good at interpreting some forms of writing -- even archaic constructs that remain valid grammatically and syntactically. In "interacting" with the one that spawned this thread I quite deliberately used such a construct and when the 'bot regurgitated my writing completely blew the interpretation.

Chatbots also lack emotion and creativity -- traits which make us human. This shows up in their "output" which is as sterile as the tools in a surgical operating room. They're boring to read, and generally time-wasters for the astute human. This is why they need to be prominently identified when in use.

So, for as much hype as ChatGPT is getting, it''s nothing but an advanced toy with very little practical use other than to pollute dialogue in meatspace.

[bikesaurus]

I've definitely found use for ChatGPT in my field of DevOps. I can use it get started on a script to do something. For example, I need a python script that enumerates AWS IAM users with attached security keys greater than 90 days old. I can write that script myself, but ChatGPT can write me an almost entirely functional script to do it much faster. It's great for boilerplate type stuff and has sped up some of the work my team does.

Having said that, it's almost a hindrance to our junior engineer, as it takes away a lot of the discovery work and thinking about how to plug things together. While he can do things quickly with it, he's not learning why it works.

I think LLMs can be used responsibly, but we also need to understand their limitations and avoid using them where real human interaction is desired.

[crfriend]

Having said that, it's almost a hindrance to our junior engineer, as it takes away a lot of the discovery work and thinking about how to plug things together. While he can do things quickly with it, he's not learning why it works.

This sums it up perfectly and points where we as humans are going. Nobody bothers to learn the "why" any longer, and this dumbs the humans down. I routinely talk about CS (Computer Science) topics that directly relate to problems we're having at work and find the conversation going over the heads of the youngsters (and lots of the Management as well). It's extremely uncomfortable.

I think LLMs can be used responsibly, but we also need to understand their limitations and avoid using them where real human interaction is desired.

They can be used responsibly by responsible adults, but those seem to be dwindling in number. For the non-astute, they represent a real and present threat.

[Coder]

I’ve been using them sparingly for my own programming work. One time I had it convert a python script to JavaScript/node - I had written the original python and didn’t feel like converting it to JS. A few minutes later I had a mostly working script - had to fix some areas where it got a few asynchronous calls wrong.

Besides that, it had invented a node module that didn’t exist, and was named exactly like the python library I was using. After a bit of encouragement (asking it to use “sharp”), it was mostly working, yet still had to make some other fixes.

It saved me a ton of time - and it gave me a solid example for when I’d use it again.

Also, sometimes I don’t have time to dig into the source code to learn how to use some obscure library, searching SO or using google doesn’t help - and the library itself provides useless examples - ChatGPT can be really instructive, if used properly.