MusicBee and OpenAI (ChatGPT)

[hiccup]

Or perhaps it's all just for the better.

ChatGPT-like algorithms (are) will be better at solving completely automated public Turing tests to tell computers and humans apart to register at internet fora.
ChatGPT-like algorithms (are) will be better at providing relevant and to-the-point answers to questions.
ChatGPT-like algorithms (are) will be be able to spit out great FAQ's and 'HowTo Tips & Tricks' in seconds.
ChatGPT-like algorithms (are) will be be able to use the phone, use the voice of a daughter to call the father to ask for some urgent money.

So it may be best to learn to stop worrying and love the bomb, sit back and relax, and enjoy the music.

[Mayibongwe]

I fear we are going to get involved in lengthy discussions with GPT spambots on the forum in the very near future.

Or perhaps it's all just for the better...

You know the old adage that says things first have to get worse before they can get better? Here's a classic case from the "very near future":

I think this is just a spambot that adds commercial links to older quotes.

Great pickup hiccup, I hadn't even noticed the hidden links at first. I must say: it's pretty genius what they're doing (amending old quotes).
Have you come across something like this before, or is this the first of its kind on this forum (that we've noticed)?

Up to now, I've only ever seen two techniques:
(1) them inserting a small-sized link at the very end of their posts.
(2) them editing their posts at a later date to insert the links for future readers.

____________


Lots of emphasis on "fiction", but don't be fooled boys. Skynet has become self-aware!!

[BoringName]

Only just noticed this post. About the only thing ChatGPT got right about 3Dbee is it displays cover art.

I've been noticing a few real estate agents clearly using AI to create their property descriptions. The lazy pricks don't even proof read it. The future is looking bleak.

It's not called Skynet anymore, it's called Legion in the latest Terminator instalment.

Not much to worry about, despite the buzzwords getting thrown around, these text predictors are not artificial intelligence.

[Mayibongwe]

...these text predictors are not artificial intelligence.

Artificial Intelligence, in and of itself, is a broad concept. However, I think Alan Turing* defined it best.
In his Turing Test**, he basically said that a machine passes as AI if the human assessing it cannot tell that they are interacting with a machine.

The first screenshot in my previous post: that reply from the spambot is astonishingly comprehensible and precise.
I think that's levels above ordinary text prediction algorithms - it didn't just spit out random phrases aimed at addressing the topic at a broad category.

Were it not for hiccup and his brilliant brains (which thankfully supersede that of the spambot for now :-), I wouldn't have known that it wasn't a legitimate person behind that reply.
Other than the spam links and the fact that it was a newly registered account, I'm wondering how else any one of us would have confidently*** determined it was a fake account.

Interesting question: if I create a new account tomorrow and use it to reply to this thread while advertising the company I work for, will I pass as a spambot?
It's getting to a point where it's improbable to tell machine from human - that's Turing's Test in full display.


Excuse the language, that's a clip from the same film mentioned below.

__________

* hiccup mentioned him a couple of posts above
** There's an interesting movie about it for anyone who cares to watch: Ex Machina (2014)
*** Even at that, pay attention to hiccup's careful use of "I think this is just a spambot", haha.

Don't mind me, I'm just geeking out.

[hiccup]

Artificial Intelligence, in and of itself, is a broad concept. However, I think Alan Turing* defined it best.

Asimov should probably have added a fourth law of robotics:

When you are asked "are you a human?", you shall not lie.

[hiccup]

It's getting to a point where it's improbable to tell machine from human - that's Turing's Test in full display.

It has gotten to a point that I find things can be quite scary.

It's now possible that you could get a telephone call by someone you know (recognising his/her voice), and have a full conversation.
And that could be an LLM model that scanned the voice of somebody you know and found your number and some personal information.
(there are LLM models that only need up to some 5–10 seconds of spoken word to convincingly replicate a voice)

[boroda]

i've heard about recent initiatives from major manufacturers of photo/video/audio recording hardware to embed some kind of "watermarks" into actual photos/recordings, which will be indistinguishable by humans, but must completely prevent AI from any kind of learning using these materials. the results reproduced by AI should look/sound quite different from the sources.

[hiccup]

Even at that, pay attention to hiccup's careful use of "I think this is just a spambot", haha.

That's very well observed.
I have bumped my nose bloody more than once when stating things I was convinced of being true, but was later on proved wrong.
A humbling experience ;-)

not promising I'll stay humble forever ;-)

[BoringName]

Artificial Intelligence, in and of itself, is a broad concept. However, I think Alan Turing* defined it best.
In his Turing Test**, he basically said that a machine passes as AI if the human assessing it cannot tell that they are interacting with a machine.

In your example I think you might be getting mixed up with which part was produced by AI and which was a genuine post. Everything from steagl3 is a genuine post. The only "AI" part is where it describes the process of embedding art which comes across as very AI to me. The bot just copied a genuine post, altered the links and added a very AI response to it.

I expect Turing meant more interaction than just reading a random forum post. In any case id say AI is more plagiarism than intelligence.