HN user

StackOverlord

110 karma
Posts1
Comments85
View on HN
www.quantinuum.com 3y ago

Quantinuum H-Series quantum computer

StackOverlord
24pts60
The Cynical Genius Illusion (2018) 3 years ago

Academia selects for conformism first closely followed by cognitive abilities ([paste link to paper here]). But who cares ?

I certainly don't I'm more interested in the groundbreaking implications of this thesis:

"Contribution to the characterization of three texture descriptors: crispness, crunchiness, and brittleness through acoustic and sensory approaches."

https://www.theses.fr/1992DIJOS040

0ThreadHN
Show HN: I automated half of my typing 3 years ago

I think it matters a great deal too, and not just for typing code but for reading it too, and this goes beyond simply typing fast but also writing short concise code, especially for parts of the code that are closer to defining program architect than implementing the details: it's better to have everything fit in one page of code rather than being dispersed throughout multiple files.

The reason for that is that a codebase will condition the complexity and thus time it takes to add features to it, in a way that is similar to algorithmic complexity and big o notation, except we as human can't even afford polynomial complexity and constant factors matter a lot.

Imagine you're developing an API, both server and client. You can cut your time in half by automatically deriving the client code from the server side specs. Of course you may have to develop that tool yourself and it takes time. The point is that the time invested developing it will be repaid each time you implement a new endpoint in your API, cutting development time in half:

n * (t(s) + t(c)) versus n * (t(s))

0ThreadHN
Discovery of spherules of likely extrasolar composition in the Pacific Ocean [pdf] 3 years ago

"additional data available to the Department of Defense" which I assume is classified.

Spot on

The 2014 meteor was originally identified as an interstellar object by Siraj in 2019 when he and Loeb were studying Oumuamua. The pair posted their findings as a preprint and submitted their results to an astronomy journal, but the paper was not accepted for publication because they used data from a NASA database that used classified information that could not be verified.

The snag started a three-year process as Siraj and Loeb worked through a bureaucratic logjam to receive government confirmation on their findings, working with scientists and officials at NASA, Los Alamos National Laboratory, and other offices. They eventually connected with Matt Daniels, assistant director for space security at the White House’s Office of Science and Technology Policy, to get an analysis from Shaw and Mozer.

https://www.cfa.harvard.edu/news/scientific-discovery-gets-k...

0ThreadHN
How was the size of Earth first measured? (2015) 3 years ago

"The pyramids give us the dimensions of our planet on a scale defined by the planet itself." - Hancock

It's a bit as if we had never seen a poem and stumble upon some text that harbors rhymes. You'd argue there's no intrinsic link between, say, pickle and tickle, you'd dig in each word etymology to show they wouldn't rhyme in their past forms. And beyond that you would deny the fact words rhymes because they are not in succession, and when I'd point out it's because the rhymes are crossed, you'd laugh it off.

It's not about convincing you (of what ? of some secret intent ? I'm not even sure this is the case). It's about how convincing it is. For that you need to model the cognitive process that interprets these coincidences.

Algorithmic Simplicity and Relevance - Jean-Louis Dessalles

https://telecom-paris.hal.science/hal-03814119/document

4.1 First-order Relevance Relevance cannot be equated with failure to anticipate [16]: white noise is ‘boring’, although it impossible to predict and is thus always ‘surprising’, even for an optimal learner. Our definition of unexpectedness, given by (1), correctly declares white noise uninteresting, as its value s at a given time is hard to describe but also equally hard to generate (since a white noise amounts to a uniform lottery), and therefore U(s) = 0. Following definition (1), some situations can be ‘more than expected’. For instance, if s is about the death last week of a 40-year old woman who lived in a far place hardly known to the observer, then U(s) is likely to be negative, as the minimal description of the woman will exceed in length the minimal parameter settings that the world requires to generate her death. If death is compared with a uniform lottery, then Cw(s) is the number of bits required to ‘choose’ the week of her death: Cw(s)  log2(52×40) = 11 bits. If we must discriminate the woman among all currently living humans, we need C(s) = log2(7×109 ) = 33 bits, and U(s) = 11 – 33 = –22 is negative. Relevant situations are unexpected situations.

s is relevant if U(s) = Cw(s) – C(s) > 0 (2)

Relevant situations are thus simpler to describe than to generate. In our previous example, this would happen if the dying woman lives in the vicinity, or is an acquaintance, or is a celebrity. Relevance is detected either because the world generates a situation that turns out to be simple for the observer, or because the situation that is observed was thought by the observer to be ‘impossible’ (i.e. hard to generate).

In other contexts, some authors have noticed the relation between interestingness and unexpectedness [9, 16], or suggested that the originality of an idea could be measured by the complexity of its description using previous knowledge ([10], p. 545). All these definitions compare the complexity of the actual situation s to some reference, which represents the observer’s expectations. For instance, the notion of randomness deficiency ([8], ch. 4 p. 280) compares actual situation to the output of a uniform lottery. The present proposal differs by making the notion of expectation (here: generation) explicit, and by contrasting its complexity Cw(s) with description complexity C(s).

0ThreadHN
How was the size of Earth first measured? (2015) 3 years ago

You can get to the absolute dimensions of the pyramid from the crackpot equations though:

https://pasteboard.co/cIMzn8KktEZy.png

https://tobeornottobe.org/the-great-pyramid-intro/math-const...

https://pastebin.com/TsUtcWAa

I just found a great recap on this topic that also fits nicely with the points your link brings up, shedding a better light on the number 43,200 in other cultures (the scaling factor between the pyramid and the earth dimensions for those who TL;DR).

https://fossana.medium.com/the-pyramids-of-giza-have-propert...

It also addresses the fact the speed of light in m/s is encoded three times in the pyramid, twice in the dimensions, in cubits and meters, and another times via the GPS coordinates (yes 3 fucking times !). But how would they know about the duration of a second ? I don't know ! What I know is that 43,200 * 2 = 84,600, the number of seconds in a day.

The thing goes deeper with this man and his investigations of the first edition of Shakespeare's Sonnets.

https://www.youtube.com/watch?v=0xOGeZt71sg

https://www.youtube.com/watch?v=nIS-hNrr0-c

0ThreadHN
AI isn’t good enough 3 years ago

For the first time with GPT4, OpenAI as been able to predict model progress with accuracy:

A large focus of the GPT-4 project has been building a deep learning stack that scales predictably. The primary reason is that, for very large training runs like GPT-4, it is not feasible to do extensive model-specific tuning. We developed infrastructure and optimization that have very predictable behavior across multiple scales. To verify this scalability, we accurately predicted in advance GPT-4’s final loss on our internal codebase (not part of the training set) by extrapolating from models trained using the same methodology but using 10,000x less compute:

Now that we can accurately predict the metric we optimize during training (loss), we’re starting to develop methodology to predict more interpretable metrics. For example, we successfully predicted the pass rate on a subset of the HumanEval dataset, extrapolating from models with 1,000x less compute:

We believe that accurately predicting future machine learning capabilities is an important part of safety that doesn’t get nearly enough attention relative to its potential impact (though we’ve been encouraged by efforts across several institutions). We are scaling up our efforts to develop methods that provide society with better guidance about what to expect from future systems, and we hope this becomes a common goal in the field.

Source: https://openai.com/research/gpt-4

0ThreadHN
Ruby's hash is a Swiss-army knife 3 years ago

You can also recursively define fib as a lazy sequence that is the pairwise sum of fib and fib shifted by one. (Clojure, from Rosetta Code).

    (def fib (lazy-cat [0 1] (map + fib (rest fib))))

    => (take 10 fib)
    (0 1 1 2 3 5 8 13 21 34)


    Explanation:

       0 1 1 2 3 5   ;  this is fib
    +  1 1 2 3 5 8   ;  this is (rest fib) 
    ---------------
       1 2 3 5 8 13  ;  this is (map + fib (rest fib))
                     ;  and the sequence needs to be initialized with (lazy-cat [0 1] ...
0ThreadHN
Why do old books smell so good? 3 years ago

Birds are susceptible to a respiratory condition called "teflon toxicity" or "PTFE poisoning/toxicosis." Deaths can result from this condition, which is due to the noxious fumes emitted from overheated cookware coated with polytetrafluoroethylene (PTFE).

0ThreadHN
Managing difficult software engineers 3 years ago

You will be the judge.

I was tasked of enabling us to backtrack on our decision of building a microservice rather than a library. The service was using postgres, and was connected to multiple website, one of which was targeted to use the service as a library. The challenge lied in the fact we used JSON to store some metadata as a shallow tree of depth 2-3, and the targeted website used mysql (which did not support JSON data at the time). Concretely it required to introduce edits in more than 300 files and manually take care of parsing a couple datetimes. Maybe there were other gotchas I have forgotten but it was hard enough with respect to the test suite I decided to write an ORM adapter to add support for json in mysql (dump it as a string, retrieve it, parse the string as native datastructures, walk the tree to convert datetime strings). I know this sounds crude, but it was enough for these tiny JSON data we only looked at when investigating problems at the command line. I took the steps to isolate this adapter as an independent library so that we could take the same decision of transforming the microservice as a library dependency on other websites we maintained ... basically at the cost of adding a library dependency in the project config.

After a code review that astutely pointed out my adapter ran at every loading of an object from the database (which I corrected by running it lazily upon accessing specific fields), it was decided my code was a gas factory and the CTO proceeded to do the right thing, get rid of that perky – but working and ready to go in production ! – library (took me 2 days to write) and just update the damn code. 3 days later he still wasn't finished with the 300 files edits, and he ended up abandoning and splitting the library in two. One would be used by the microservice, and the other as a library on the website. He traded treewalking our datastructures with treewalking our codebases, program efficiency with programmers efficiency (we were 15 on the team while our competitors had 50 devs). This way any modification brought to that service/library: - is not transferred to the other version, leading to organizational complexity - is transferred and the time it takes to dev a features is multiplied by a number between 1 and 2. - or there are complex surprises under the surface and this factor is greater than 2.

0ThreadHN
Open Challenges in LLM Research 3 years ago

A type system is for internal consistency though. Facts are about external consistency with real world data. And even then facts are always a social augmentation in that they are always captured in a given social context, and by that I include the lenses of theoretical frameworks and axiomatics. They always have a spin they can lose when considered from another standpoint and at the very minimum they are conditioned by attention and relevancy, and it has everything to do with our current representation of the world and nothing with the world itself.

0ThreadHN
Open Challenges in LLM Research 3 years ago

I don't get your argument about the frame problem. Maybe it's like squeezing a big pillow inside a small bag. A bulge forms that won't fit. It's the frame problem. Turn the pillow around, squeeze it into the bag again: a bulge now forms on the opposite side: it's the hallucination problem. I can see how one could be the solution to the other. Hallucinations as a lack of rules.

0ThreadHN
Ode to the M1 3 years ago

I upgrade my macbook once per decade. I changed the battery, the ssd and the bluetooth card about 6 month ago (8 years into owning my second macbook). I need to change the screen (1 cm black strip on the left because of a shock) and the speakers are broken. Anyone has a tip for replacing the screen cheaply ? I spotted some going for 30$ on Aliexpress, it looked like the seller had a rather important stock.

0ThreadHN
Judge sends Sam Bankman-Fried to jail over witness tampering 3 years ago

Alain Juillet, former head of the French CIA (DGSE) recently claimed in a youtube interview that "if you do not know what is happening in the African freemason circles, you can't understand the current geopolitics of Africa". He was the head of several lodges/masonic confederations.

https://www.jeuneafrique.com/137763/politique/franc-ma-onner...

discusses the power play between American and French controlled lodges in Ivory Coast.

https://www.jeuneafrique.com/141986/politique/franc-ma-onner...

In the neighboring countries of Mali and Guinea, the same expansion phenomenon is observed. Amadou Toumani Touré, overthrown on March 21st, and Alpha Condé oversee the destinies of the Grand Lodge of Mali and that of Guinea, respectively. As for Blaise Compaoré, he was - until he gave way to Djibrill Bassolé, his head of diplomacy - grand master of the Grand Lodge of Burkina, which counts among its ranks numerous ministers, diplomats, and businessmen, notably a part of the management of the national chamber of commerce. Further south, Beninese Thomas Boni Yayi, a known evangelist, has always denied his affiliation to Freemasonry but maintains close relations in the field. Togolese Faure Gnassingbé keeps people guessing, causing some of his brothers to smile: "This young president quickly understood the means to control his elite," they note.

0ThreadHN