Conferences; research updates

[dmm]

This week, Nov 17-18, Thu-Fri, 8am-11:45am Boston time, "Quantum physics and the first-person perspective": www.essentiafoundation.org/quantum-physics-and-the-first-person-perspective/seeing/

JuliaCon 2023, juliacon.org/2023/ the call for proposals is posted, deadline Dec 18: pretalx.com/juliacon2023/cfp


I've spent more quality time focusing of two breakthroughs in understanding the nature and the behavior of machine learning models which came from the "penumbra" of "prosaic alignment" start-ups and which I wrote about in my previous two posts.

"Grokking is (more or less) solved." I took brief notes between Oct 21 and Oct 23: github.com/anhinga/2022-notes/tree/main/Grokking-is-solved

"Generative autoregressive models are similators." I took extensive notes between Oct 5 and Oct 23: github.com/anhinga/2022-notes/tree/main/Generative-autoregressive-models-are-similators

I am continuing to develop thoughts related to these topics, I am going to gradually write more about those topics in the comments.

Comments

[dmm]

Speaking of "Generative autoregressive models are similators", https://dmm.dreamwidth.org/64434.html

Three main directions of thoughts here are:

* How this can be used to control model's behavior better (or to train models better)

* What can happen inside those models (e.g. is it true that the most interesting and consequential things in the AI development will soon be happening INSIDE simulations which unfold when one is running those kinds of models or their successors?)

* Philosophy: how might this new "Janus' paradigm" illuminate the nature of "our reality"

Also, the sociology of this is interesting. How fast and wide the knowledge of this new way of thinking about GPT-3-like models would spread, and how much of short-to-medium-term impact on the AI development trajectory it would have?

[juan_gandhi]

Wow, thank you.

[dmm]

"Quantum physics and the first-person perspective": the first day is about "Wigner's friend": https://en.wikipedia.org/wiki/Wigner%27s_friend

The organization has a YouTube channel, so there is a hope that the recording will be put there eventually: https://www.youtube.com/channel/UCHKZdDf09_8vVHm102fu0sg

[dmm]

One of the papers by the first speaker: "The view from a Wigner bubble", https://arxiv.org/abs/2008.05100

(He is not a QBist, but he has written this paper with his QBist hat on: https://en.wikipedia.org/wiki/Quantum_Bayesianism )

[dmm]

One of the papers by the second speaker: "Qubits are not observers -- a no-go theorem", https://arxiv.org/abs/2107.03513

[dmm]

A paper by the third speaker, https://www.theoryverse.com/: "Thought experiments in a quantum computer", https://arxiv.org/abs/2209.06236

"We introduce a software package that allows users to design and run simulations of thought experiments in quantum theory. In particular, it covers cases where several reasoning agents are modelled as quantum systems, such as Wigner's friend experiment. Users can customize the protocol of the experiment, the inner workings of agents (including a quantum circuit that models their reasoning process), the abstract logical system used (which may or not allow agents to combine premises and make inferences about each other's reasoning), and the interpretation of quantum theory used by different agents. Our open-source software is written in a quantum programming language, ProjectQ, and runs on classical or quantum hardware. As an example, we model the Frauchiger-Renner extended Wigner's friend thought experiment, where agents are allowed to measure each other's physical memories, and make inferences about each other's reasoning."

https://github.com/jangnur/Quanundrum - their open source code

(this is a new fork of a 3 year old https://github.com/Croydon-Brixton/qthought )

[dmm]

Listening to the second day of talks. The first speaker is from U. of Massachusetts, from this research group: http://www.physics.umb.edu/Research/QBism/

This is Boston campus, where I have friends, and where I've even given couple of talks in 2009 and also attended some talks and interacted with people in more recent years...

This was one of my talks there: "Remarks on Neurodynamics: Spikes, Spike Synchronization, and Spike-Timing-Dependent Plasticity", https://www.cs.brandeis.edu/~bukatin/neuro-talk-mar-25-2009.pdf

And this was another: "Partial Metrics, Fuzzy Equalities, and Metric-Entropy Pairs", https://www.cs.brandeis.edu/~bukatin/partial-metrics-talk-jun-2009.pdf

[dmm]

That particular group seems to do research linked with rather interesting math:

http://www.physics.umb.edu/Research/QBism/readings.html

http://www.physics.umb.edu/Research/QBism/solutions.html

https://en.wikipedia.org/wiki/SIC-POVM - symmetric, informationally complete, positive operator-valued measure

This is one of papers from that group: "The SIC Question: History and State of Play", https://arxiv.org/abs/1703.07901

The Wikipedia page says:

"A symmetric, informationally complete, positive operator-valued measure (SIC-POVM) is a special case of a generalized measurement on a Hilbert space, used in the field of quantum mechanics. A measurement of the prescribed form satisfies certain defining qualities that makes it an interesting candidate for a "standard quantum measurement", utilized in the study of foundational quantum mechanics, most notably in QBism. Furthermore, it has been shown that applications exist in quantum state tomography and quantum cryptography, and a possible connection has been discovered with Hilbert's twelfth problem."

In particular: "SICs and Algebraic Number Theory", https://arxiv.org/abs/1701.05200

https://en.wikipedia.org/wiki/Hilbert%27s_twelfth_problem

[dmm]

The first speaker: https://www.jaypienaar.com/

His old blog: https://pienaarspace.wordpress.com/2013-2/

Some papers:

"Unobservable entities in QBism and phenomenology", https://arxiv.org/abs/2112.14302

"QBism and Relational Quantum Mechanics compared", https://arxiv.org/abs/2108.13977

[dmm]

The second speaker is speaking about intersubjectivity.

She has a recent joint paper with Rovelli: "Information is Physical: Cross-Perspective Links in Relational Quantum Mechanics", https://arxiv.org/abs/2203.13342

She publishes quite a lot: https://arxiv.org/search/physics?searchtype=author&query=Adlam%2C+E

https://www.rotman.uwo.ca/portfolio-items/adlam-emily/#tab-fbaca88009969aa65de

[dmm]

The third speaker is a hard-core researcher, publishes a lot: https://www.theorie.physik.uni-muenchen.de/lsluest/oriti_group/members/group-leader/oriti/index.html

"quantum gravity (particularly on tensorial group field theories (group field theories, random tensor models), loop quantum gravity, spin foam models, lattice quantum gravity, non-commutative geometry), gravitational physics, fundamental cosmology; wider research activities in philosophy of science and foundations of physical theories, in particular foundations of quantum mechanics and philosophy of spacetime theories"

[dmm]

He argues that physical laws are purely epistemic, have no ontological status.

He proposes to focus on "agents", and he defines agents as entities performing modeling activity (this is, in some sense, in line with my recent thinking).

But he notes that quantum gravity is the real challenge in this sense, and the real litmus test for any methodology...

[dmm]

Some of the papers by the fourth speaker:

"Relationship between covariance of Wigner functions and transformation noncontextuality", https://arxiv.org/abs/2004.06318

"A mathematical framework for operational fine tunings", https://arxiv.org/abs/2003.10050

He is speaking about

"Why interference phenomena do not capture the essence of quantum theory", https://arxiv.org/abs/2111.13727

[dmm]

A rather lively panel discussion at the end, plenty of disagreement.