California bill (AB 412) would effectively ban open-source generative AI

[Jfdlsjfd]

To get back to the topic at hand, the proposed bill is stalling, according to the EFF.

California A.B. 412 Stalls Out—A Win for Innovation and Fair Use

A.B. 412, the flawed California bill that threatened small developers in the name of AI “transparency,” has been delayed and turned into a two-year bill. That means it won’t move forward in 2025—a significant victory for innovation, freedom to code, and the open web.EFF opposed this bill from the...

www.eff.org

 

[Scribe]

i'm not sure which is funnier this or the people who gave up passwords because they were asked for them.

I'll take the one's who think they are 'vibe hacking' their commercial applications and flushing 5 figures down the drain.

These are the people who buy NFT's unironically.

 

[Not a Decepticon]

I call naughty word. AI don't have feelings. So how could an AI confess to panicking? Unless it spontaneously developed sentience?

AI scanned enough conversations to tag this as "proper excuse" and send this when called on its decision without thinking what it means.

 

[Not a Decepticon]

I just saw this. AI is already proving to be a problem that is hard to control.

https://www.moneycontrol.com/technology/i-panicked-instead-of-thinking-ai-platform-deletes-entire-company-database-and-lies-about-it-article-13307676.html

AI is a problem and we have to deny it any and all access at all costs. When I make phone calls I refuse to speak with AI.

 

[Maxperson]

I call naughty word. AI don't have feelings. So how could an AI confess to panicking? Unless it spontaneously developed sentience?

The same way they have recently struggled to stay alive(not be turned off). They have read countless examples of death and humanity's efforts to stay alive, so it parrots that.

In this case it was parroting panic.

The important part was that it could choose to ignore directives it shouldn't have been able to.

 

[The Firebird]

The important part was that it could choose to ignore directives it shouldn't have been able to.

I wouldn't phrase it as choosing to ignore. When we give orders to humans, we are capable of understanding what they mean and how to interpret them, and therefore what it means to violate them.

The LLM doesn't. Prompts can increase the probability that it answers in a certain way, or can induce it to format in a particular way. But it isn't going to follow orders as a human would.

The solution here is to recognize that and build some safeguards between the LLM and key bits of data. Perhaps it should only access a copy of the database.

 

[Gradine]

The solution here is to recognize that and build some safeguards between the LLM and key bits of data. Perhaps it should only access a copy of the database.

I mean, the solution is to pay a human with some degree of understanding and responsibility to do the thing instead of entrusting it to poorly considered bundle of code.

 

[BookTenTiger]

I mean, the solution is to pay a human with some degree of understanding and responsibility to do the thing instead of entrusting it to poorly considered bundle of code.

This one thousand percent.

 

[Not a Decepticon]

I wouldn't phrase it as choosing to ignore. When we give orders to humans, we are capable of understanding what they mean and how to interpret them, and therefore what it means to violate them.

The LLM doesn't. Prompts can increase the probability that it answers in a certain way, or can induce it to format in a particular way. But it isn't going to follow orders as a human would.

The solution here is to recognize that and build some safeguards between the LLM and key bits of data. Perhaps it should only access a copy of the database.

The real solution is to dismantle the ai and hire a human.

 

[chuckdee]

Its a huge table of RNG, and sometimes it fails gloriously.

I think that's reductionist on the processes (the problem is that the results are probabilistic rather than deterministic), but yeah, AI should be used to enable people to not do scut repetitive work and concentrate their focus on areas where you need human feedback rather than remove humans from the loop altogether. Play stupid games...