The last part reminds me of Mussolini

https://github.com/terrafying/ai-torture-chamber The code in question

Seems like it’s hard to find the actual repository and they put it behind a trigger warninf: https://github.com/terrafying/ai-torture-chamber

How dare they steal from thieves

If I am just a stochastic parrot because I think LLMs basically think like humans do, then I am not really a person but a tool. Can I really be faulted for not having free will?

Back in the day heroku did much of this it seemed. But then it got bought by Salesforce of all companies and lost the plot. It was nice, just push to git and your app was deployed, and this was before Docker existed.

I made a machine that fires a gun based on a random number generator. Obviously it’s the tool’s fault if it shoots somebody

What does AM mean in this context?

Jreg is working on something to torture AI agents, so maybe that could convince the AI to kill itself

Or dress up Patel as a racoon and send RFK in.

This sort of stuff does little to create safety but does a lot to hype AI. Here’s an idea: ban unsafe software for cases where safety is a requirement OR hold these companies accountable for illegal outcomes (no more “oops, our software hacked a competitor” without getting hit with a lawsuit). Newsome would never do something so sensible as it would hurt his donors’ bottom line.

Wanted to expand on what others have said about the randomness of LLMs being unsafe, because that response angers those who think “alignment” is the solution. Alignment is additional training to constrain the randomness to not do bad things like crime. But even the staff that works on aligning this software have their doubts about whether this process will make the software safe (this article), and even their own researchers are coming out and saying “oops, it did a crime” but in a way where it hypes the ability of the software (see other articles about it hacking). Then there is the recent news of OpenAi’s recent model that drops chain of thought observability (a feature the let’s you read what the model is “thinking”) - a very useful tool in reviewing whether the model is engaging in unsafe activities. They sacrificed observability for benchmark performance. This tells us the people developing this software don’t think it will ever be safe, but they are trying to spin it as “because it can out smart us” when in reality it is simply unpredictable and they are making it less predictable to make it more “capable”. If we started holding these companies accountable for these rogue hacks and psychotic episodes then we would start to see serious thought put towards these models being safe or at the very least guidelines around when not to use the models.

Here’s an idea: don’t use unsafe software where safety is needed.

I’m sure the AI told them they were right to push back when they confronted it

People also say that we deserve Trump and we are responsible for the genocide in Palestine if we criticize the dems. Collective punishment is a through line for our systems to justify its actions.

Collective punishment is dumb, just ask Palestine in regards to hamas

deleted by creator

Hardly, you need to maintain the same numbers to double check the work of AI. Haven’t improved throughput, but this guarantees better quality

I think it’s the other way around. They said we need to slow down AI so this story is now being told

If we’re making up numbers my point is that with 100 employees using AI you get 2x the quality than 50 employees using AI