Michael Chen

1 followers · 1 following · 872 views

on the atlas — 28

highlights — 110

In Pirkei De Rabbi Eliezar 11:5-6, we are told: “God created humanity from the four corners of the earth-yellow clay, and white sand, black loam and red soil, therefore, the earth can declare to no part of humanity that it does not belong here, that this soil (the earth) is not their rightful home.”
Crossing the Bridge from the Particular to the Universal | Reform Judaism
the Greeks often used allegorization of Homer, etc., as a means of finding ancient textual support for particular doctrines
EarlyChurch.org.uk: Literal and Allegorical Interpretation in Origen’s Contra Celsum by Dan G. McCartney
Literal fulfilment of prophecy is for Origen an important proof of Christianity.
EarlyChurch.org.uk: Literal and Allegorical Interpretation in Origen’s Contra Celsum by Dan G. McCartney
he then goes on in 1.35 to argue for the application of the prophecy to Christ based on the literal sense: "What sort of sign would it be if a young woman not a virgin bore a son?"
EarlyChurch.org.uk: Literal and Allegorical Interpretation in Origen’s Contra Celsum by Dan G. McCartney
Origen dismisses him as irrelevant: 5.54 - The books entitled Enoch are not generally held to be divine among the churches. 5.21 - The Scriptures accepted in the churches of God do not declare that there are seven heavens.
EarlyChurch.org.uk: Literal and Allegorical Interpretation in Origen’s Contra Celsum by Dan G. McCartney
In 2.48 he asserts that the chief value of a miracle is not that it happened but the truth allegorically symbolized therein.
EarlyChurch.org.uk: Literal and Allegorical Interpretation in Origen’s Contra Celsum by Dan G. McCartney
Finally, Origen can appeal to Jesus himself, since he both spoke in parables (which were treated as allegories) and showed the previously ignorant disciples how the Scriptures actually spoke of him and his redemptive program (4.42 - referring to Luke 24).
EarlyChurch.org.uk: Literal and Allegorical Interpretation in Origen’s Contra Celsum by Dan G. McCartney
In 4.49 Origen lists several passages where Paul is reported to have interpreted allegorically (1 Cor 9:9–10 [do not muzzle an ox], 10:1–2 [baptized into Moses], 3–4 [the spiritual rock which was Christ], Eph 5:31–2 [marriage as a symbol of the church]). Further, he finds in Paul's contrast of letter and spirit in 2 Corinthians justification for two levels of meaning (6.70). It is the "veil of ignorance" that keeps people from seeing the allegorical meaning. And Paul's own allegorical interpretation of Sarah and Hagar in Galatians 4, where Paul actually uses the word allgore, provides him with…
EarlyChurch.org.uk: Literal and Allegorical Interpretation in Origen’s Contra Celsum by Dan G. McCartney
Origen refers to the psalmist's declaration "I will open my mouth in parables" (Ps 77:2 [Gr.]) as indication that the prophets intentionally spoke allegorically,[26] and another Psalm (118:18 [Gr.] - "Open my eyes, that I might behold wondrous things out of thy Law") as proof that the ancient prophets regarded the Torah as containing hidden truth
EarlyChurch.org.uk: Literal and Allegorical Interpretation in Origen’s Contra Celsum by Dan G. McCartney
Origen regards it a point in Moses' favor over the pagans that, although his words carry a deep allegorical meaning, even the literal meaning is good and true
EarlyChurch.org.uk: Literal and Allegorical Interpretation in Origen’s Contra Celsum by Dan G. McCartney
Whatever the word "resurrection" meant to Origen, it at least meant that Jesus was seen alive, and that the tomb was empty. It is a point of argument that "there are many evidences of his appearing after death."
EarlyChurch.org.uk: Literal and Allegorical Interpretation in Origen’s Contra Celsum by Dan G. McCartney
We have already seen that in the Hellenistic world at large, divine texts were regarded as bearing an allegorical meaning. Whatever the reasons for this development, the converse was also regarded as true, that for a text to be divine, it must bear an allegorical meaning.
EarlyChurch.org.uk: Literal and Allegorical Interpretation in Origen’s Contra Celsum by Dan G. McCartney
Ernie discovers the surprising fact that one is a member of three. In fact, he generalizes, if n is bigger than m, then m is a member of n. Filled with enthusiasm, he brings this fact to the attention of his favourite playmate. But here, sadly, the budding mathematical collaboration breaks down. Johnny not only fails to share Ernie's enthusiasm, he declares the prized theorem to be outright false! He won't even admit that three has three members!
Benacerraf's identification problem
Set-theoretic method I (using Zermelo ordinals) 0 = ∅ 1 = { 0 } = { ∅ } 2 = { 1 } = { { ∅ } } 3 = { 2 } = { { { ∅ } } } ⋮ Set-theoretic method II (using von Neumann ordinals) 0 = ∅ 1 = { 0 } = { ∅ } 2 = { 0 , 1 } = { ∅ , { ∅ } } 3 = { 0 , 1 , 2 } = { ∅ , { ∅ } , { ∅ , { ∅ } } } ⋮
Benacerraf's identification problem
Since there exists an infinite number of ways of identifying the natural numbers with pure sets, no particular set-theoretic method can be determined as the "true" reduction.
Benacerraf's identification problem
Nominalism (from Latin nomen, "name") says that ideal universals are mere names, human creations; the blueness shared by sky and blue jeans is a shared concept, communicated by our word "blueness". Blueness is held not to have any existence beyond that which it has in instances of blue things.
Theory of forms
At the summit of existence stands the One or the Good, as the source of all things.
Platonism - Wikipedia
The concept of the anima mundi (Latin), world soul (Ancient Greek: ψυχὴ κόσμου, psychḕ kósmou), or soul of the world (ψυχὴ τοῦ κόσμου, psychḕ toû kósmou) posits an intrinsic connection between all living beings, suggesting that the world is animated by a soul much like the human body.
Anima mundi - Wikipedia
world-soul, the copy of the nous
Platonism - Wikipedia
the nous, wherein is contained the infinite store of ideas
Platonism - Wikipedia
"The lovers of sights and sounds like beautiful sounds, colors, shapes, and everything fashioned out of them, but their thought is unable to see and embrace the nature of the beautiful itself."
Platonism - Wikipedia
The only true being is founded upon the forms, the eternal, unchangeable, perfect types, of which particular objects of moral and responsible sense are imperfect copies.
Platonism - Wikipedia
In the 3rd century AD, Plotinus added additional mystical elements, establishing Neoplatonism, in which the summit of existence was the One or the Good, the source of all things; in virtue and meditation the soul had the power to elevate itself to attain union with the One.
Platonism - Wikipedia
Platonism affirms the existence of abstract objects, which are asserted to exist in a third realm distinct from both the sensible external world and from the internal world of consciousness, and is the opposite of nominalism
Platonism - Wikipedia
acute, circumflex, and grave ⟨´ ῀ `⟩
Ancient Greek phonology - Wikipedia
The acute represented high or rising pitch, the circumflex represented falling pitch, but what the grave represented is uncertain.
Ancient Greek phonology - Wikipedia
The accented mora is marked with acute accent ⟨´⟩. A vowel with rising pitch contour is marked with a caron ⟨ˇ⟩, and a vowel with a falling pitch contour is marked with a circumflex ⟨ˆ⟩.
Ancient Greek phonology - Wikipedia
standard Attic dialect of the fifth century BC
Ancient Greek phonology - Wikipedia
loss of vowel length distinction, the shift of the Ancient Greek system of pitch accent to a stress accent system, and the monophthongization of diphthongs (except αυ and ευ).
Koine Greek phonology - Wikipedia
from about 300 BC to 400 AD. At the beginning of the period, the pronunciation was close to Classical Greek, while at the end it was almost identical to Modern Greek.
Koine Greek phonology - Wikipedia
The doctors treating his final illness had asked him if he was in great pain, to which he replied: "There are people who suffer much more than me."
Carlo Acutis - Wikipedia
linear parameter decomposition
On Optimism for Interpretability
we can recover a species' place on the tree of life from the activations of the Evo 2 genomics model
On Optimism for Interpretability
We can peer inside these models to extract and use the conceptual frameworks they've developed – frameworks that may transcend current human understanding in their respective fields and lead to breakthroughs across those domains.
On Optimism for Interpretability
interpretability should enable us to fix it. This means developing capabilities to modify or remove specific features and mechanisms within trained models. It means allowing us richer control over the training process itself, and more efficient distillation methods that preserve only desired behaviors.
On Optimism for Interpretability
interp-based “model diffing”
On Optimism for Interpretability
Then we evaluate whether these conditions hold for three separate feedback loops via which AI will improve AI: A software feedback loop, where AI develops better software. Software includes AI training algorithms, post-training enhancements, ways to leverage runtime compute (like o1), synthetic data, and any other non-compute improvements. A chip technology feedback loop, where AI designs better computer chips. Chip technology includes all the cognitive research and design work done by NVIDIA, TSMC, ASML, and other semiconductor companies. A chip production feedback loop, where AI and robots b…
Once AI Research is Automated, Will AI Progress Accelerate?
Better measurement: comprehensive measurements and forecasts of the pace of software progress.
How Can AI Labs Incorporate Risks From AI Accelerating AI Progress Into Their Responsible Scaling Policies?
Pro-social governance: a governance structure that empowers the lab to act in the public interest even if this is opposed to shareholder value.
How Can AI Labs Incorporate Risks From AI Accelerating AI Progress Into Their Responsible Scaling Policies?
AI development speed limits: a Responsible Scaling Policy that explicitly evaluates how fast the lab could advance its overall AI capabilities while still being safe – given the lab’s current processes – and commits not to advance faster than that.
How Can AI Labs Incorporate Risks From AI Accelerating AI Progress Into Their Responsible Scaling Policies?
Info security that prevents a state actor from stealing the weights and accelerating their own AI development. Alignment, boxing, and internal monitoring that are good enough to ensure that misaligned AI doesn’t “poison” the AI development process**.** External oversight: a publicly legible mechanism to ensure a responsible 3rd party signs off on high stakes decisions, including any decision to rapidly improve AI capabilities.
How Can AI Labs Incorporate Risks From AI Accelerating AI Progress Into Their Responsible Scaling Policies?
Today it takes ~3 months to train a SOTA AI system.
Will the Need to Retrain AI Models from Scratch Block a Software Intelligence Explosion?
Retraining means that we’re unlikely to get an SIE in <10 months, unless either training times become shorter before the SIE begins or improvements in runtime efficiency and post-training enhancements are large.
Will the Need to Retrain AI Models from Scratch Block a Software Intelligence Explosion?
Our evaluation agent, which builds behavioral evaluations for researcher-specified behaviors, successfully discriminates models with vs. without implanted test behaviors in 88% of runs. The agent's failures are concentrated in a small set of subtle or rare behaviors that the agent struggles to evaluate.
Building and evaluating alignment auditing agents — AI Alignment Forum
Our tool-using investigator agent, which uses chat, data analysis, and interpretability tools to conduct open-ended investigations of models, successfully solves the Marks et al. auditing game 13% of the time under realistic conditions. However, it struggles with fixating on early hypotheses. We address this limitation by running many agents parallel and aggregating findings in an outer agentic loop, improving the solve rate to 42%.
Building and evaluating alignment auditing agents — AI Alignment Forum
Testing whether methods such as GFlowNets, max entropy RL, KL-regularized RL, and diversity games can be used for unexploitable search.
Research Areas in Evaluation and Guarantees in Reinforcement Learning (The Alignment Project by UK AISI) — AI Alignment Forum
Are there regularisation methods (entropy bonus, KL constraint, pessimistic bootstrapping) we can use to prevent exploration hacking without hurting sample efficiency too much?
Research Areas in Evaluation and Guarantees in Reinforcement Learning (The Alignment Project by UK AISI) — AI Alignment Forum
Deliberately train models to exploration hack, creating a 'model organism' (Hubinger et al., 2024) of exploration hacking. These 'model organisms' can then be used to examine possible ways of preventing exploration hacking.
Research Areas in Evaluation and Guarantees in Reinforcement Learning (The Alignment Project by UK AISI) — AI Alignment Forum
Does exploration hacking occur in practice? Are there ways to detect and prevent this behaviour?
Research Areas in Evaluation and Guarantees in Reinforcement Learning (The Alignment Project by UK AISI) — AI Alignment Forum
Baker et al., show that certain process oversight protocols induce increasingly hard-to-spot reward hacking.
Research Areas in Evaluation and Guarantees in Reinforcement Learning (The Alignment Project by UK AISI) — AI Alignment Forum