ttctciyf

Carroll's episode is worth listening to, not least for it's exceptionally clear exposition of the current physics consensus around QM, qft, standard model, etc., and how they are related, but the point you specifically mention is a weak spot for me.

For one thing he hardly mentions the epic failure of string theory to make good on its initial promises nor the murky waters of anthropic claims and metaphysical notions of beauty, etc., used to keep it suspended like Wile E Coyote after running off the cliff of empirical support.

As non-crackpot physics profs and postgrads (e.g. [1], [2]) have pointed out, this has not engendered public trust and is a key ingredient in whatever "crisis" the discipline is undergoing.

Not that Carroll doesn't make plenty of good arguments to support his views, but his seeming equation of any criticism of the field with crackpotist heresies is a cheap trick for a philosophy prof.

1: https://www.math.columbia.edu/~woit/wordpress/?cat=8

2: https://www.youtube.com/watch?v=kya_LXa_y1E

More info is posted on AMD's community site[1] which dates itself "Thursday", but google which search dates "22 hours ago" which would be Sunday.

They say:

> Which AMD GPUs does the HIP SDK support?

> Everything from high-end workstation and gaming GPUs to laptop cards and even APUs. Below, you can see a list of the GPUs that AMD tests for use with the HIP SDK, but you may be able to support other cards in your own software: for example, Blender's HIP rendering backend supports AMD GPUs going back to the Vega generation.

Helpfully, the article then links to "GPU and OS support"[2] which at the present time redirects to a login-required page at readthedocs.com. Sigh.

1: https://community.amd.com/t5/rocm/available-now-new-hip-sdk-...

2: https://rocm.amd.com/en/update-windows-install-guide-improve...

It's odd how their Linux support documentation is so lackluster. I have an RX 6800 that runs stable diffusion fine with recent PyTorch on Linux, but looking at their Linux support info you wouldn't think this was possible.

The situation (afaik) until a day or so ago was that Linux had official support for datacenter and "pro" cards and unofficially for consumer cards while Windows had nothing.

Considering the dev hours that must go into this it seems odd they don't have better release info, frankly.

Right, but this was new in the last 48 hours or so I think.

Interesting to see consumer GPUs from RX 6800 up getting HIP support on windows, (on the "Radeon" tab) whereas I don't think I've seen them explicitly mentioned under Linux. I wonder if this will now change too?

It's become increasingly difficult to avoid systemd, making Slackware one of the few remaining reasonable options for that. My comment is (half-seriously) imagining that in the future Wayland will attain a similar dominance, in which case Slackware is likely to be one of the leaders of the resistance!

I see this more as a risk for the commercial application of LLMs in that it works against the brand identity LLM operators want.

Happy with Xorg, plan to continue using it as long as possible. Will switch to slackware to avoid Wayland. May switch earlier to dump systemd. And get xv back[1]

1: https://news.ycombinator.com/item?id=36716090

A retnet commit has now appeared in Microsoft's torchscale repo:

https://github.com/microsoft/torchscale/commit/bf65397b26469...

Cheap yoghurt where I buy it tastes fine and has a pleasant consistency. YMMV.

Ordinary, but skimmed milk - mumsnet says you need to boil it for ~10 mins and then let it cool to ~40C before adding the live yoghurt culture. Working well so far!

UHT is usable, and doesn't need to be boiled, but for that added convenience it costs nearly twice as much.

Anecdotally, UK, the cheapest yoghurt (believe me, I have researched!) just dropped from £1.10/l to about £0.7/l - anomalous considering the rampant groceries inflation we have, but there it is.

Milk is about £0.64/l, so roughly equivalent if you add in the cost of heating it.

However, the price for live yoghurt is a different story - much more expensive, which is why I have recently started making my own.

If he's going to roast the logo, maybe stick a fork in it too?

Have to admit I got a tremendous kick out of this!

As a physical model, as opposed to a purely mathematical entity, don't you need to add operators to the wave function in order to have applicability to observable phenomena? I'm no physicist, but applying operators to the wave function seems like a computation to me?

> we can progressively increase the numerical values of the weights—eventually in some sense “blowing the mind” of the network (and going a bit “psychedelic” in the process)

I wonder if there's a more exact analog of the action of psychedelics on the brain that could be performed on generative models?

Andrew Gallimore's online course on psychedelic neurochemistry[1] argues that psychedelics disrupt something like a constant fine-tuning cycle taking place in normal perceptual activity, causing new activations between cortical structures where connections had previously been dormant (see [2] for an in-at-the-deep-end account), causing the generative model in the higher layers of the cortex[3] to adjust in an effort to regain predictive efficacy.

I wonder if some such kind of "dynamic" generative mechanism might be needed to perturb an artificial model analogously?

1: https://www.youtube.com/playlist?list=PLbqdD4EM-aEfmLvbWu8GQ...

2: https://www.youtube.com/watch?v=L45A6XlPRM0&list=PLbqdD4EM-a...

3: https://en.wikipedia.org/wiki/Predictive_coding

> assumption that everything about our physical universe is described by computational rules

If you understand "physical" to mean "stuff that physics studies" then the assumption is not so surprising - physics itself is the discovery and application of models to correctly predict or retrodict (i.e. compute) the simple measurable behavior of matter, so it is intrinsically unable to study anything unamenable to computation.

In fact Wolfram is engaged in an attempt (as he sees it) to enlarge the computational paradigm of physics from models consisting largely of differential equations to a "multicomputational" paradigm (see [1]).

I think your statement is much more defensible if you drop the word "physical": The assumption that everything about our universe is described by computational rules is not a given. A classic example might be our inability to sufficiently specify "qualia". If our experiences were fully describable by computational rules I suspect we would by now have discovered how to specify them so that we could agree (or disagree) that 'the red I see' looks like 'the red you see', but our inability to do this is notorious. Nonetheless my (and your) experience of red is part of the universe.

1: https://writings.stephenwolfram.com/2021/09/multicomputation...

From 1999, the linked article seems an interesting read on the intersection between 'hacker culture' and the libertarian project; dry and a little quaint.

Readers might like to compare the more journalistic (and entertaining, IMO) 2014 piece on 'San Francisco’s tech-libertarian “Reboot” conference' from sadly defunct Pando, for example:

> At first glance it makes no sense to front a rabidly anti-gay candidate like McMorris Rodgers to sell the Kochs’ and the Paul family’s scrubland libertarianism to a Bay Area audience full of hip disruptors and “anarchist” practitioners of bohemia grooming fads.

> But that’s because what Silicon Valley folks think of when they hear the word “libertarianism” actually has very little connection to what the libertarian movement actually stands for, and has stood for since the 1970s.

...

1: https://web.archive.org/web/20141118174216/http://pando.com/...

Yes, thank you!

What's next, a solipsist's guide to social interaction?

Well, not a possibility I considered, but nothing in the argument depends on the regularity being a permanent feature of the cosmos, just that systematic human endeavours, such as mathematics or indeed meaningful debate, depend on it, so when it goes they go. In that sense, if you wish to consider this conversation meaningful you are kind of ceding the point that for now chaos doesn't, in fact, reign. If you don't consider it meaningful, then why are you having it? :)

> built on top of chaos.

I don't think you've taken on the full force of the argument that regularity in human activity (such as building systems) requires a source of order for it not to simply dissolve into chaos itself.

> How can it be that the patterns that we observe are so consistent?

How can we claim to discern consistency (or inconsistency) without the ability to follow a rule correctly? And how can we follow a rule without a source of order or regularity in the cosmos? Wouldn't it be like trying to build a the Eiffel tower out of live slugs?

If you insist on an absence of order in the physical universe, the onus is on you to explain how regularity in human activity (required for mathematics of any kind) can be achieved without it.

This is not an argument for Platonic realism, BTW, or against intuitionism, roughly construed as the view that "mathematics is a creation of the mind" as per [1], or in your formulation that 'mathematical abstractions are not part of the physical universe' (if I understand what you're saying). You can perfectly well believe that mathematics is a mental construct and at the same time acknowledge that it's possible to observe regularities and order in the cosmos. If you want to insist that the regularity doesn't come from physical law, then I find it hard to see how you'll escape from some kind of Platonic belief in a non-physical realm that serves as the source of order :)

In your TV screen dots analogy, isn't it usually thought that patterns appear only because of the structured, generative activity of law-observing physical components, specifically the neurons comprising your grey matter?

1: https://plato.stanford.edu/entries/intuitionism/

> fools us into thinking there is some inherent order -- which there isn't.

Bold claim :)

Even while maintaining a willful agnosticism about Platonic realism, it seems clear that the business of doing mathematics - intuitionistic or otherwise - depends on the ability to state and follow unambiguous rules, or else how to establish a proof within some axiomatic system or other?

But if mathematicians have this ability, even as an asymptotic ideal, and are able to make judgments as to when rules are not being followed correctly, mustn't this in turn depend on some pre-existing regularity (i.e. order) in the universe? Ineluctable physical law seems a natural candidate for this order.

Thanks!

Also xlander ( apparently maintained! https://github.com/ac000/xlander )

It's such a quick and handy tool that I became habituated to xv over a decade of use. Some years ago though, I found it was no longer available to install after an upgrade. I was eventually able to fudge a working version from old code and patches and taking hints (iirc) from some patches for MacOS, but now that has suffered bitrot.

So I'm curious what OS you're running and where your xv is from? Still miss it!

ISTR from my scout days that "ducking" is a fabric surround at the bottom of tent walls that you tuck under the groundsheet - probably from the same root, I guess - though few online dictionaries seem to have heard of it.

Isn't the claimed "truth" based on an appeal to a more intuitive notion of truth than a formal model-relative one?

The Godel sentence for some Formal Axiomatic System asserts, by way of numerical encoding, its own unprovability in that FAS, and since it in fact cannot be proven in the FAS it is "true" in the sense that it asserts (via its numerical encoding) something which is in fact the case, i.e. just the ordinary sense of "true".

The article didn't mention it, as far as I could see, but the source is available[1] at the author's site with an interesting and fairly permissive license.

1: https://80.style/#/hsp/noctis_widgets_and_accessories/noctis...

Essentially, a (fake) Windows instance.