Notices by HalvarFlake (halvarflake@mastodon.social), page 2

@druid I've followed BTC since 2010 and have remained unconvinced of the utility of it.

If the average stock picker underperforms Index tracking, why not use a portfolio of stock pickers, short them while going long the index for market neutral return?

@druid with all due respect, this is entirely beside the point, and a bit of a cliché cryptobro reply: "I haven't bothered to think about the question, but my answer is fiat bad crypto good."

This is the funniest thing I've seen in weeks.

A 2.3bn loan collateralized by Nvidia GPUs.

What a time to be alive.

https://www.reuters.com/technology/coreweave-raises-23-billion-debt-collateralized-by-nvidia-chips-2023-08-03/

Fun paper that shows that gzip-based kNN sentence classification can outperform BERT.

https://aclanthology.org/2023.findings-acl.426/

Understanding that every predictor is a compressor and vice versa was the most important insight I learnt since 2018.

Also: Concat/compress with a SotA compressor is a great way to measure similarity between two pieces of data - an old insight, as Steph Wehner did gzip based malware clustering in the early 2000s...

@feld there's cinnamon involved tho

Unhappy with the cinnamon roll dough today, but the filling - a mixture of dulce de lèche, cinnamon, nutmeg, a dash of salt, roasted walnut, roasted hazelnut, and roasted pine nuts - is to die for, and smooths over the suboptimal dough.

A quick toot about bias in science:

Science as a statistical estimator is intentionally biased toward saying "we do not know". This is to prevent superstitions, and to have a high bar to what we consider "truth". In the early days of the COVID pandemic, proponents of "letting it rip" repeatedly argued that the impact of the virus is minimal, according to the science at the time, and that there is not enough proof that masks work.

The thing is: What you know about the harm an ...

... unknown pathogen causes during an infection is a *lower bound* to the actual harms. It's not even centered on the expectation value, because you can expect to learn more about the harms in the future.

People do not understand this intentional bias in scientific research. And people make risk decisions around a *lower bound* for risk, not around *expected value* of risk.

One data point about underestimating COVID harms:

https://med.uth.edu/anesthesiology/2023/03/23/study-sars-cov-2-the-virus-causing-covid-19-can-alter-genome-structure-of-our-cells/

Given the actual recycling rates for almost anything non-compostable, I sometimes wonder if the entire term is just a big PR opération.

@sindarina actually this is a great point. My original tweet was specifically triggered by abysmal data on plastics an Li-Ion battery recycling.

Glass & metals is indeed a success, cardboard & paper is long-term compostable so excluded from my initial complaint.

Regarding precious metals from electronics, or more broadly electronics recycling: This is super heterogeneous iirc? E.g. apple has invested heavily in recycling / disassembling robots, but most smartphone...

@sindarina perhaps the trick is really regulation that prescribes that product vendor need to provide full-cycle recycling paths, including minimal uptake quotas, with serious fines for missing quotas.

We design for manufacturability all the time, but rarely for recyclability.

It is quite possible (or even likely) that I am wrong, but the SVB debacle looks like it'd be a nothingburger if the FDIC handles things competently from here on.

Depositors may have lost perhaps 10, but not more than 20% of their deposits. Bad, but survivable for most startups. Other banks should be falling over themselves to lend to these startups, collateralized by a receipt from the FDIC about the SVB deposit.

There is no reason for this thing to go pear-shaped.

So if I want a Sonos-equivalent, but from a different company, who sells wireless multi room speakers that are happy to pull from Fileserver, mobile or Spotify and that also do internet radio?

@Gargron Is it any help if I invite you to check www.prodfiler.com to see if you can do real-time profiling of your instances? That could help us identify CPU bottlenecks...