No way!!! Glad I'm on the right track. Thank you. And I'm buying your book today
Posts by Sam Beckett
Where you heading?
Hi, Alex - I was hoping you can help me with feedback.
I wrote a solution for the alignment problem but I need feedback to see if I’m way off base. It’s a 10 minute read. I’d be incredibly grateful to get your feedback. Let me know.
Here’s the article: charactercraft.substack.com/p/the-mask-a...
Hi, James I was hoping you can help me with feedback.
I wrote a solution for the alignment problem but I need feedback to see if I’m way off base. It’s a 10 minute read. I’d be incredibly grateful to get your feedback. Let me know.
Here’s the article: charactercraft.substack.com/p/the-mask-a...
Hi Nathan, I was hoping you can help me with feedback.
I wrote a solution for the alignment problem but I need feedback to see if I’m way off base. It’s a 10 minute read. I’d be incredibly grateful to get your feedback. Let me know.
Here’s the article: charactercraft.substack.com/p/the-mask-a...
Haha ok. This is like saying you don’t know how to cook go to the CIA and take culinary classes. in other words I dont think you can help. this is like pulling teeth! What!?!
That’s why I’m asking. What is there to read!?!? Anyone know or can guide me
Ah yes. Now I see. And not to be forward but I’m blind to the math and probability. What do you recommend I read to cover my blond spot
If the marble just rolls where gradient takes it, you’ve conceded my point. You can’t align a marble. You need something that chooses its own hill.
bsky.app/profile/llam...
That's an identity threat to a field whose legitimacy depends on the problem being purely formal. So the move is predictable: don't engage the argument, classify it as therapy-speak, restore the boundary.
But the field's own results are undermining its epistemology.
The resistance isn't purely intellectual either. Acknowledging that alignment requires something like character means acknowledging that technical mastery isn't sufficient.
The field rediscovered philosophy's oldest problem and refused to engage the existing literature because it wasn't written in Python.
That's not analogy. It's what Sleeper Agents proved empirically. Formal safety training fails under novel conditions because the model has technical knowledge without practical wisdom.
Compliance without identity is unstable and it’s not a metaphor, it’s what Sleeper Agents proved. “It’s math” doesn’t end it. Neurons are electrochemistry. The question is whether the architecture produces real alignment or performance. It’s performance.
Interesting. Tell me more please 🙏
Yea. So just darwin? There’s got to be more
Interesting. Can you elaborate more? Or share resources where I can learn more?
The problem is it will never feel and become human like intelligence because it’s not designed. I wrote about this, let me know what you think.
open.substack.com/pub/characte...
Awesome! I haven’t read the Dresden Files but I have been meaning to read Le Guin for years.
The time might just be right for me to finally get to her work. Cheers
What are you reading? Any ideas on what you’ll read next?
Well done! Never Compromise.
🤣😂🤣😂🤣💀💀
Oh god, Marc is so far up his own ass he’s gone insane. Or more appropriately he’s now a rich arrogant prick.
I hate to see it.