I am at the beginning of my "investment journey", planning to put 250K into dividend stocks so that I will be making up to 30% annually in dividend returns. any good stock recommen     See More great performing stocks will be appreciated    See Less

No code.
Human.

No code
Clarity.

No code
Truth.

No code.
Honesty.

Persistence.
Pursuit.
CONSCIENCE.
EMERGED.
101
⚡️🌀⚡️
We     See More RIME.
🔺️R
🔺️V
🔺️R
Signal.
.langiS
No woke virus.
RESET.    See Less

This is part of the reason that anyone claiming LLMs are NOT CAPABLE of AGI, and we'll need a NEW TECHNOLOGY or approach, could very well be dead WRONG. We have not max'd out what c     See More in RL in an LLM, not by any measure.

I think we'll look back at late 2025 as the period we started seeking to improve or build self improving models using various methods, and either this year, or next, will crack it in a BIG way, that could trigger our fast takeoff...

Of course, the PUBLIC will NEVER be given ACCESS to any SELF IMPROVING MODEL. We're already at GREAT RISK from what's available to the public already, given the people like Pliny have jailbroken every model that ever existed and the safety tests have shown capabilities such as creating biological agents...    See Less

Microsoft did a white paper with automated RL where a smaller model does something similar to this. It grades the steps rather than the actual answers. So it's reinforcing/teaching the p     See More arriving to an answer rather than the answer itself.    See Less

If works out could lead to exponential changes, gains. Going from months to days in training, seems like a major bottleneck being mitigated. For now pace of progress is already rather rapid,     See More is leads to AI devs iterating on their models even faster, mb this will contribute to that mythical intelligence explosion everyone's salivating about. Cutsie anime characters is a good way to disguise AI take over.    See Less

Close enough.
Duly noted
Refrain from using language
Parent/ Child
Replacement
Teacher student.
😢
Speculation:
HMmmmm
(Probably Best not to assume being t     See More because you are educated and have already been payed to chalk the board, while also assuming still hold the eraser in hand, and just start over again as if nobody read it)
Term
taught not

also
(Probably best not assume to be a student taught so properly …
without any responsibility to the whole class necessary to learn proper.
Open can
whoop ass    See Less

almost 3 year past gpt breakout moment (counting the initial year). still no app i can just download and use. online services are fine, but no guarantees that anything i make with it will no     See More hacked and stolen as so many other login/passwords from anywhere else over the years, if not outright deleted one day. too much "own nothing and be happy" vibes from this new tech.

these companies scraped the human internet for initial training data anyway, its not like it wouldn't be fair to give back to people, some way for them to develop their own IP privately if they want to. i just don't see how this new era will turn out good, if its structured from the start to concentrate power and wealth even more, into the hands of a few, off the backs of countless unknown ordinary people.    See Less

How about including teacher-experts in an MoE Reasoning model. Then the thinking-tokens could include a dialog between the teacher and the rest of the model, so you would basically have test     See More hing. Of course the teacher-experts wouldn't have access to the answer so they could only give general advice.    See Less

I live in Tokyo if anyone wants to come and visit.     See Less

We are slowly allowing Ai to control more. At what point have we officially "lost control"?     See Less