Sabitlenmiş Tweet
🆕🤖 When "coding agents" can do all the rational parts of software development (possibly that's NOW)... you need to go meta-rational!
⤵️ Link in reply tweet! 🔗
English
David Chapman
63.7K posts
David Chapman
@Meaningness
Better ways of thinking, feeling, and acting—around problems of meaning and meaninglessness; self and society; ethics, purpose, and value.
Katılım Eylül 2010
100 Takip Edilen35.9K Takipçiler
@ArtirKel ☹️ it might be important to contextualize these statements, which pertain to a particular stage in the practice path only, and should not be taken as general advice
English
@AnotherPattern @OortCloudAtlas I'm sorry... but it will come! 🌼
English
<looking at sky> “Bless the p+p+e⁻ → D+ν fusion reaction! Bless *all* the fusion reactions! Bless the sun, bless the stars, bless the galaxy! Bless the hydrogen atoms, each one throughout the universe, that every one of them may be happy!”
English
<take a few steps across the radiant parking lot> “Bless the full SU(3) × SU(2) × U(1) field, that lights our way, without which the sky could not be blue”
English
“Bless the E field and the B field, that afford us miraculous spooky action at a distance” <presses button>
English
David Chapman retweetledi
The center is explicitly modeled on the Berkeley Alembic, but with an emphasis on growing up rather than waking up. In fact, we designed it with significant feedback from leaders there, as well as from spiritual practitioners across the world, from many faith backgrounds.
English
David Chapman retweetledi
That means we'll be offering psychological training on things like shadow work, exposure/courage work, grounding and anti-dissociative training, cognitive security and defense against the digital arts - at scale, for no-to-little cost, to the general public.
English
David Chapman retweetledi
The problem with classic Dharma approaches to Western psychological problems are well known, as are the failure modes of more materialistic Western spiritualities. But in the age of AI, the need for mature practitioners with strong senses of self has never been greater.
English
David Chapman retweetledi
Friends, I'm opening a spiritual center in Boulder!
It's called Nameless Mountain, and it's focused on spiritual maturity rather than enlightenment.
We're trying to raise $50,000 to cover this year's expenses. Please help build this with us by donating or becoming a member:
givebutter.com/nameless-mount…
English
@ahron_maline Ah, good catch. I got distracted by reading about the atmospheric refraction effect, which other discussions I found emphasized.
English
@Meaningness Ok sure enough, "equilux" is based on defining day & night in an asymmetric way. Not very interesting imo
timeanddate.com/astronomy/equi…
English
☀️ Happy Equilux! Today is the day when night and day are the same length. Which is *not* the equinox (for complicated physics reasons that don't make much sense); that's the day after tomorrow.
English
@JeffLadish I guess the point @jon_stokes was making, and I suspect he's right, is that the past couple years advances have mostly come from better training data, not "general intelligence." That tends to imply they will remain incompetent out-of-distribution.
English
@JeffLadish Ah, yes, good point, also that! Plus, synthetic data is probably important too. It's feasible to generate high-quality synthetic data in some domains (programming and math probably included). Not so much some others, and the user interaction data is probably critical there.
English
🤖 This has been my guess for a couple years; but I don't know. I would like to know. x.com/jon_stokes/sta…
Jon Stokes@jon_stokes
I think what's happening is not that the big 3 labs are making some fundamental improvements in model architecture or pretraining. Rather, they just have massive amounts of live user sessions to post-train so they can cover more of the space of likely model interactions.
English
David Chapman retweetledi
@Meaningness I don’t think most of AI companies most valuable data is from user interactions. Eg Claude Code isn’t good because of all the users of Claude Code providing data to Anthropic, but rather because Anthropic engineers have built really good RL environments to train Claude
English
WW27 is supposed to be a bug-fix release. Looking forward to it. x.com/klara_sjo/stat…
Klara@klara_sjo
There will be no WW3. They've abandoned numbered releases and switched to a live service model with seasonal events.
English