ChatGPT

[ChuckF]

ok little buddy

[-FX-]

Originally Posted by viscousmemories

The possibilities are infinite.

Currently people are all go-go over text, some links, a few lies and the ability to type shit out on a computer, search the internet, remember shit you typed, etc etc

Next up, more real shit like buying and selling, investing, bitcoin, using software, creating video (deep fakes and others), music, product placement, reviews, social media postings, doxxing, threats and blackmail, hacking, ransomware, onlyfans virtual sex, designing shit, manufacturing that shit, selling that that shit, biomedical research, tracking people online, rigging elections, disabling power grids, advanced weapon design and war strategy, piloting drones, analyzing weaknesses and creating cascading disasters, etc etc

It's just getting started

Sure typing shit out might seems like the greatest thing ever, but it's just words and data on a screen

When it enters the actual real world, that's when it get's real interesting

[-FX-]

Especially when it's AI vs AI, and anything goes ...



Let me tell you a story kids ...

[erimir]

Originally Posted by -FX-

You can't sue an AI

Or arrest them

Brilliant insight here.

You also can't sue a car, or a gun, or a computer.

Yet drivers, shooters and hackers can all face consequences. And saying "I didn't have full control over what it did" or "I couldn't predict the consequences of using it the way I did" isn't a free pass.

[-FX-]

Originally Posted by erimir

You also can't sue a car, or a gun, or a computer.

Indeed.

Ownership of something matters on many levels. If somebody creates an AI, then releases it into the wild, where it can be modified, spread out into the cloud, maybe even running as a virus on computers worldwide, who do you go after?

You can't sue or arrest Samuel Colt for what is done with a revolver.

What if the inventor of an AI is dead? But the AI is still going? Doing shit? Breaking your forums?

Do you go after the computer it is running on? The network? Who do you blame?

Even more important, how do you stop it? If it's not actually alive, how can you kill it?

If outlawed, how do you get rid of it? Is it even possible?

[Sock Puppet]

Those were hotly debated topics in the 1990s.

[-FX-]

Once an AI learns to make memes, it's all over

[JoeP]

Please see this post in the more important Nick Cave thread.

[Ensign Steve]

[viscousmemories]

True, but also true that it can be pretty helpful in some arenas. I've been using it to explain code snippets, help me write some simple Python scripts, and even develop queries in SOQL (Salesforce Object Query Language) for me lately. Unlike StackOverflow, when I get an error trying some code it recommends I can paste the error and say "this is what I got" and it can usually identify what it did wrong and modify it.

[mickthinks]

Indeed.

I think there's something wrong with @cwebber's weaponising "mansplaining" to denigrate the AI there, but I am not smart enough to articulate it. Anyone?

[Ari]

I think it’s pretty accurate as what they’ve basically built is a mimic trained heavily on the internet. It’s not trying to be correct it’s trying to pick the most likely human response, and the internet is full of articles explaining things in simple, not always correct ways.

That it will get science horribly wrong, math occasionally wrong, and programing sometimes wrong seems to say a lot about the quality of data fed into the mimic and that perhaps programing advice is the most trust worthy thing on the internet these days.

[JoeP]

From The race of the AI labs heats up | The Economist

Epoch, a non-profit research institute, estimates that at current rates, big language models will run out of high-quality text on the internet by 2026

Quick everyone, generate lots of high-quality text!

* JoeP looks around forum

Hmm, maybe not here.

[Ensign Steve]

[Ensign Steve]

Originally Posted by JoeP

From The race of the AI labs heats up | The Economist

A real big issue is that over time a larger proportion of text on the internet will be generated text, and you do not want to feed that back in to your training model.

Google Translate faces a similar issue, as its training data largely consists of websites that have been human-translated into multiple languages. Naturally, as machine translation gets better, websites can save money by publishing machine-translated versions of their sites. It's up to Google Translate to not then turn around and use those sites as training data.

There are signals they can look for, in both cases, to determine whether potential training data is machine-generated. But it's challenging, and it is a whole other step that can be hard to get right.

[JoeP]

The whole Internet will end up as a bunch of competing AIs copying each other's output while carefully excluding their own.

[Ensign Steve]

I mean, have you seen an unmoderated subreddit?

[Ensign Steve]

If there's anybody whose take on this I care about, it's this guy.

ChatGPT with Rob Miles - Computerphile - YouTube

I've only watched about the first 3rd so far, but the whole simulation/simulacrum distinction is blowing my mind.

He goes over this thing a bit, and it's like wtf.

Reddit - Speaking to ChatGPT in perfect Danish, while it insists throughout that it can't speak or understand Danish.

What a time to be alive.

[viscousmemories]

Nothing, Forever Is An Endless Seinfeld™ Episode Generated by AI

[-FX-]

Of course, we can’t dispense with news sites. I’ve already considered forums, and that leaves the personal blog. AI will synthesize all other types of content. Content creators will begin to question why they’re putting in all this hard work, only for AI to “steal” it. So this will result in fewer and fewer content creators and an increase in posts for personal, non-monetization reasons. And I think forums and personal blogs will be a big part of that.

With ChatGPT AI, Forums Are The Future Of The Web

[-FX-]

Forums are the future!

Except for that censorship problem. Which the Admins and Mods don't ever realize is happening.

[LarsMac]

This chat AI stuff reminds me of a guy I used to know. Ralph was the younger brother of a buddy of mine from Chicago.
He was very knowledgeable. Like he had spent his entire childhood reading(absorbing) Encyclopedia Britannica.

You could ask him about something, and he could flood you with information.

However, any time he was around, and people were having conversations, he would suddenly trip on a word that he heard, and start running off with a discussion on whatever subject matter was triggered.

Ralph would even trigger on other words in his spiel, and take off on all sorts of tangents.

The first AI that I encountered could do that, too. However ChatGPT seems to do better at not triggering off its own spiel.

[Ari]

Originally Posted by Ensign Steve

ChatGPT with Rob Miles - Computerphile - YouTube

In theory I’m subscribed to that channel and thus should have known all about this video, yet this is the first I’m seeing of it, and oh look, now that I’ve watched it Youtube is suddenly showing me a bunch of other videos from the same channel I’ve never seen. Thanks algorithm.

At around 33m he says the bigger the data set the more likely it was to ask not to be turned off and declared itself sentient. Great, that’s not going to fuck with things.

[Crumb]

You use the YouTube subscription page right? https://www.youtube.com/feed/subscriptions See all the videos from your subscriptions chronologically with no algorithmic interference.

[viscousmemories]

Too bad it doesn't have filter. I subscribe to a lot of newsy channels that post every day if not multiple times a day.