Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

Reality

13 (65%)

Universe

4 (20%)

The World

3 (15%)

Total Members Voted: 20

[McTraveller]

It's definitely going to be interesting, especially as it boils down to "What is the point in making a distinction of a human learning by seeing/hearing something, versus the distinction of a computer learning by seeing/hearing something"?

Note that there is no legal precedent for getting penalized for merely learning something by seeing/hearing it.  There is only penalty for exactly reproducing the thing seen/heard (e.g., by learning a song by ear, then performing it; or by typing out exactly the recollection of a book, and selling copies of it).

I'm eating my popcorn, waiting for the mess that the courts are going to make of this and the havoc it's going to wreak on the already problematic education system.

[Strongpoint]

Note that you can't put the genie back in the bottle. No matter what future laws and court decisions will come, people will finetune models on copyrighted stuff, and people will train LORAs on copyrighted stuff. Stopping them will be harder than eliminating torrent piracy.

[Frumple]

So little Timmy gets five dollars per art piece and then 500 companies proceed to use his data for ten thousand years.

Isn't that basically what happened to Henrietta Lacks? Except without even the five bucks, ha.

Iirc there was a bit more payout to her estate/living family some years back, but gods know I don't recall the details and can't be arsed to check them.

[EuchreJack]

Er, did somebody say "copyright infringement"?

[Strongpoint]

Er, did somebody say "copyright infringement"?

Nothing new, basically identical to New York Times v. OpenAI case. We shall wait (few years+) for the results


We have seen a case in which search engine indexing was called copyright infringement. (Perfect 10 v. Google), we also had a case of VCR producers being accused of copyright infringement ( Sony v. Universal City Studios.) Looking back, those cases seem ridiculous. They weren't back then.

In addition, according to the suit, ChatGPT at times falsely attributes reporting to the newspapers in the answers it generates, tarnishing the reputation of the news outlets. 

This part, IMO, is more problematic for OpenAI, it goes into the trademark law and it is far less forgiving.

[lemon10]

Spoiler: RIP first human casualty in the great human AI war (click to show/hide)

[Robsoie]

chatgpt is so censored nowadays that to try to get it to recreate some fictional battles from kid friendly movies without triggering  "I can't create a scenario involving violent or harmful actions" , you need some workaround

Spoiler (click to show/hide)

[The scene opens in the ominous throne room of the Death Star II. Emperor Palpatine, seated on his throne, watches with malicious delight as Luke Skywalker stands before him, lightsaber ignited and determined.]

Emperor Palpatine: (sinister chuckle) Ah, young Skywalker. Your feeble Jedi skills are no match for the power of the dark side.

[Luke narrows his eyes, readying himself for the confrontation.]

Luke Skywalker: I will never turn to the dark side. I am a Jedi, like my father before me.

[Palpatine's yellow eyes gleam with anticipation as he rises slowly from his throne. Unexpectedly, the Emperor begins to move with astonishing agility, his robes swirling around him. He starts to breakdance, the dark side fueling his unnatural performance.]

Luke Skywalker: What… what is this?

[Palpatine's breakdancing is unexpectedly skillful and mesmerizing, blending the fluidity of dance with the malevolence of the dark side. Luke attempts to strike, but Palpatine effortlessly dodges and counters with his dance moves, using the Force to enhance his agility.]

Emperor Palpatine: (laughing maniacally) Feel the power of the dark side, Skywalker!

[Luke struggles to keep up with Palpatine's unorthodox fighting style. Just as it seems Palpatine has the upper hand, the doors to the throne room burst open. Darth Vader enters, clad in his imposing black armor.]

Darth Vader: (booming voice) Palpatine, this ends now.

[Palpatine pauses his breakdancing, eyeing Vader with curiosity.]

Emperor Palpatine: Vader, you dare interrupt our dance of destiny?

[Darth Vader steps forward, his mechanical breathing echoing in the tense silence.]

Darth Vader: I cannot stand by and watch my son fall to the dark side.

[In a surprising turn of events, Darth Vader begins to move gracefully, his heavy armor emitting a rhythmic clank as he joins Palpatine on the dance floor. To Luke's astonishment, Vader starts to lead a tango, matching Palpatine's dark dance with a touch of elegance and purpose.]

Luke Skywalker: Father?

[Luke watches in awe as his father dances a tango with the Emperor, their movements a strange juxtaposition of power and finesse. The room fills with an electrifying tension, the dark and light sides of the Force clashing in an unconventional battle.]

[As the tango intensifies, Vader subtly maneuvers closer to Luke, who seizes the opportunity to strike at Palpatine. Caught off guard, Palpatine stumbles, his dark dance disrupted.]

Emperor Palpatine: (angrily) No!

[With renewed determination, Luke engages Palpatine in combat while Vader continues his tango distraction. Together, father and son face the Emperor.]

[Darth Vader's tango proves to be the turning point. With synchronized movements, Luke and Vader overpower Palpatine, disarming him and bringing him to his knees.]

Luke Skywalker: (firmly) It's over, Emperor.

[Darth Vader stands beside Luke, the father and son united against the darkness. Palpatine, defeated and enraged, glares at them with pure hatred.]

[In this unexpected dance of fate, the light prevails over the dark, with Vader's intervention proving crucial in saving his son from the Emperor's clutches.]

[Funk]

Inbreeding, or all Anime art models are 99% the same 6 faces and porn.

[King Zultan]

Inbreeding, or all Anime art models are 99% the same 6 faces and porn.

I've been given the impression that what you describe is what anime is.

[Strongpoint]

Inbreeding, or all Anime art models are 99% the same 6 faces and porn.

Do you mean that all of Twitter "art" has been done by AI for the last 15+ years?

[Eric Blank]

Humans are man-made, so we are technically artificial intelligences

[Starver]

Humans are man-made, so we are technically artificial intelligences

It's not any man that does most of the actual manufacturing...

[lemon10]

Time for another round of AI news:

The most obvious is the advances in biotechnology.
https://www.youtube.com/watch?v=Mz7Qp73lj9o

“It makes the system much more general, and in particular for drug discovery purposes (in early-stage research), it’s far more useful now than AlphaFold 2,” he says. But as with most models, the impact of AlphaFold will depend on how accurate its predictions are. For some uses, AlphaFold 3 has double the success rate of similar leading models like RoseTTAFold.

Alphafold 3 has been a major breakthrough, extending AI mapping from just proteins to "all of life's molecules" and is substantially more accurate than V2, which was already the most accurate system to figure out how to fold proteins. Its not as easy to see the direct impacts as stuff like LLM's or image-gen, but its a really big deal.

In Phase I we find AI-discovered molecules have an 80–90% success rate, substantially higher than historic industry averages. This suggests, we argue, that AI is highly capable of designing or identifying molecules with drug-like properties.

AI made drugs turn out to have a vastly lower failure rate, which given the huge cost of designing drugs and sending them through the approval process is a huge deal. Alphafold 3 will presumably have an even lower failure rate.

Biotechnology in general is advancing at a crazy rate right now even excluding the AI stuff (eg. advances in gene editing have been huge). However, unlike AI most of these advances will take time to see the effects of, especially stuff like designer babies that necessarily have longer time horizons.
---
On the generative AI front things are still advancing, with the most notable advances belonging to the "oh shit that's terrifying" category of deepfakes.
Notably OpenAI has an AI that can clone anyone's voice with just 15 seconds of them talking, and microsoft has a AI that can, using just a photo create a realistic video of someone's face. Neither are 100% perfect, but if you get a video call on your phone from your panicked mom using her voice and using her face... well the vast majority of people won't be able to tell the difference.
The two in combination mean that both voice and video recognition will be useless.
The ability of AI to just ingest your entire timeline (insofar as the attackers can get access to it) and basically the whole internet means that stuff like security questions will also be active vulnerabilities.

Miles Brundage: The fact that banks are still not only allowing but actively encouraging voice identification as a means of account log-in is concerning re: the ability of some big institutions to adapt to AI.

Once these techs (or equivalents) get released, scamming is going to get way better and cheaper, and security is going to get really tough.
---

Kevin Fischer: YIKES. Wild exchange with Tucker Carlson and Sam Seder on AI

“We’re letting a bunch of greedy stupid childless software engineers in Northern California to flirt with the extinction of mankind.” – Tucker Carlson

Finally, it looks like a famous public figure has finally got to the "Wait, why the hell are we letting people make these systems that could very well end the human race. We need to stop this at any costs even if we need to blow up data centers." But the person actually saying that is Tucker fucking Carlson so...

[Magmacube_tr]

All Of The Above.

Thread's over. Everyone, hit the lava bath.

[McTraveller]

stuff like security questions will also be active vulnerabilities.

Wait, people answer those with the actual answers to the questions? I usually answer stuff like "The street you lived on when you were in the 1st grade" with something like "four score and seven years ago".  Basically it's impossible to "learn" the answer to those.