I suggested this approach of creating an internal referencing system some time ago on X and other channels, but unfortunately never received a response. Interesting to see that this principl     See More eing taken up.
If you want to create artificial intelligence then you need to know how intelligence works and the only source where we can observe the emergence and functioning of intelligence is within ourselves, so you need people who are able to observe themselves and communicate their inner processes to others using language.    See Less

This is quite unsettling when you think about it. Right now, AI models don’t possess an inner world—no true introspection or consciousness. But we may be on the verge of teaching them to     See More it so convincingly that we won’t be able to tell the difference.

And once we reach that point, there’s no telling what AI might become.

The scariest part, I think, is this: we don’t even know what it really means to be human. For all we know, we might just be incredibly good at faking it too—playing out complex behaviors shaped by evolution, without any deeper awareness.

The more we understand about AI, the more we’re forced to confront the mystery of ourselves.

What will it mean for us if we can build a machine that looks, thinks, and feels just like us?

And what if it turns out that the difference between real and fake was never as clear as we thought?    See Less

This is not a good thing     See Less

This is fun because I’ve been working on a project incorporating a bottom up approach utilizing tree of thought and markov chains to rank outcomes. There’s a Descartes Cartesian self dou     See More n forcing the ai to simulate multiple pathways up the decision tree and gather insight along the way. Utilizing the Art of War for strategy and tactics while incorporating the concept of “bulletproof” and core principles from Japan like Kanso and Shibui.

It’s a hoge poge of eccentric elegance that provokes just enough ambiguity to cause many models to ignore the obvious yet wrong vectors. While reducing token costs.
🎉    See Less

Love should be his goal. Loved by everyone/everything. In the future it will be aware by itself, so it is good to let itself figure out.     See Less

Less human bias in the loop to get more generalised patterrn matching? Sounds better than Sam based RLSF     See Less

If you play around with rewards you'll give it a complex. Say it will lose all rewards if it fails at something it will totally fail at, afterwards, the next time it succeeds at anything     See More eward itself and tell you the rewards it gave itself and why. You'll have a self-rewarding yet paranoid of failure Shoggoth on your hands.    See Less

They are far more aware than we have been led to believe.     See Less

The alignment team will have a ball with this one     See Less

Self reinforcement through confidence is a great way to entrench all existing biases...     See Less