a_136_chiffa

LLMs were invented by AI2, before Transformers were a thing - with RNN-based ELMO.

Alan Kay on Dijkstra, (1997 OOPSLA keynote): 'I don't know how many of you have ever met Dijkstra, but you probably know that arrogance in computer science is measured in nano-Dijkstras.'

Was this article written by Dijkstra himself?

I am a computational biologist with a heavy emphasis on the data analysis. I did try Jupyter a couple of years ago and here are my concerns with it, compared to my usual flow (Pycharm + pure python + pickle to store results of heavy processing).

1) Extracting functions is harder 2) Your git commits become completely borked 3) Opening some data-heavy notebooks is neigh impossible once they have been shut down 4) Import of other modules you have in local is pretty non-trivial. 5) Refactoring is pretty hard 6) Sphinx for autodoc extraction is pretty much out of the picture 7) Non-deterministic re-runs - depending on the cell execution order you can get very different results. That's an issue when you are coming back to your code a couple of months later and try to figure what you did to get there.

There are likely work-arounds for most of these problems, but the issue is that with my standard workflow they are non-issues to start with.

In my experience, Jupyter is pretty good if you rely only on existing libraries that you are piecing together, but once you need to do more involved development work, you are screwed.

> But history has judged him as one of the worst presidents.

Nah, not if you talk to any craft brewers or beer enthusiasts. They are basically inches short of having his portrait hung in every craft brewery and pub due to his H.R. 1337 bill. If it wasn't for political partisanship, this might have already happened.

> The baseline percentage of the population which experience depression (and other mental illness) each year is known to be pretty high -- and highest in for people in their 20's. One third sounds fairly normal; at any rate, there is a burden here to show that this is an exceptional proportion.

There has been a study in Belgium that showed that Grad students had a relative risk of psychological disorder scored according to a questionnaire used to decide whether psychological/psychiatric care is needed were about 5x times more at risk than the general population, even when matched for the age, education attainment and several other known common confounding factors for mental health disorders. https://www.sciencedirect.com/science/article/pii/S004873331...

> If you are starting from an undergraduate degree, you probably need at least two years to take the PhD intro classes. One year to start a research program might work in a subject like the author's where your papers are chats about social implications, but there are plenty of subjects where even the data collection is going to take longer than that.

In Germany, you need a master's to start a PhD, as such your classes load is minimal...

> Being a professor is only one possible goal and for some fields, it isn't even the primary one.

Not according to the professors mentoring you. That's changing, but the normal consideration is that if you are not on a tenure track, you are a failure. If you acknowledge it, quite often you will lose the support of your PI and your peers.

One where you don't get sustained funding to maintain it. In compbio even major resources, known to everyone in the domain only have funding from one two-year grant to another.

Absolutely not. In case it's someone you don't know a colleague, you open a communication channel allowing them to switch to a different topic and indicating you are not currently busy or annoyed (in which case you would have added it to the "ca va" - "ca va, mais un peu presse/occupe en ce moment").

If it's someone you know well or who knows you, in the timings of responses and the expression of the face you are able to pick up how they're feeling, if something is going on in their lives or is bothering them.

It's much less scripted than the American "hi, how are you", which can only have one answer and always need to have the same expression.

You can change your credit score, social circles or what you search. Your genome is frozen in time and is passed down to your children.

Whether you like it or not, US tends to lead the world. Tools developed here tend to implanted elsewhere, even if they get regulated after about a decade.

> I really doubt this. I've been genotyped by 23andMe and the most interesting information I've seen from their health reports are a handful of disease probabilities and some fairly useless-to-me traits (like, for instance, a probabilistic view of my hair color).

This is about to change. The academic research going on is immense and GWAS studies coming in the 10 years will be characterizing everything - from your chance to develop cancer, to car crash or dropping out of school.

You need to be aware of what analyses of your data is being shown to you compared to what are possible and can be run in the background. It's a little bit like the Facebook telling things about you in ~2011 (your best friend is X) and then Cambridge Analytica pouring on it five years later to serve you ads that would best affect your voting pattern.

If they don't find your data, they will likely use your relative's data to infer the risk. I doubt your cousins, uncles and aunts would think twice before signing up with their real names to one of the "find where your ancestors came from for $69.99!" adverts.

This industry should be heavily regulated and have engineered security layers making sure you always know where your data is, who has or had access to it and how it can be used.

For now.

The issue with such software is that it's mission-critical. In other terms, it needs to be zero-downtime, bullet-proof, audited and certified by external actors and to be supported for the next 30-50 years.

Historically, all those limitations make the "move quickly and break things" approach popular in the current wave of startups impossible and the code with this kind of requirements has historically been implemented by large corporation already working in the industrial domain.

If anything, this feels a domain where GE and large contractors will be chosen to write mission-critical code over startups.

The issue is not to repeat the experiments, but rather to avoid putting the logs in the wheels of people who want to re-analyze your existing data (including yourself a couple of years later) and avoid losing thousands of working hours doing forensic data analysis to point out shitty science. Keith's Baggerly 2010 talk on the hoops he had to jump through to get to Anil Potti (https://en.wikipedia.org/wiki/Anil_Potti) is a great demo of the application case: https://youtu.be/7gYIs7uYbMo.

And as for "doing real science" vs. trying to make it more reproducible, there is an excellent analogy with "doing real programming" (aka adding features) vs. refactoring and architectural adjustments. Telling that you consider the second as a waste of time tells more about yourself than about the subject.

That's why free/leisure time to figure WTF is going on politically/economically socially and an education not to get lost on the way of getting there have been considered as pre-requisites for a functioning democracy.

They are military personnel, writing code for military applications. They are under the obligation to refuse to execute orders that go against international conventions and have an obligation to respond to Congress/Senate inquiries. Their superiors, who might want to pressure them into doing things that are illegal or don't align with the US stance can be court-marshaled.

Good luck getting anywhere near that level of responsibility from Google execs or managers.

Not really - the inflation is currently reported on a "reference consumer basket", which doesn't really represent different segments of society.

In other terms, if you are spending on health care of education you are in a really bad place - prices for those services have risen at about 8x the inflation rate over the last 40 years. IF you are purchasing or building a house you will face prices 3x inflation price. However, if you are mostly buying electronics and software, the prices have actually fallen by about 50-80% over the same period.

Hope not. In the US, at least, the craft beer came to be dominated just by a couple of types that are easy to make - notably the IPAs. Some other types are present as well, but the selection and the quality are nowhere near. While it is a welcome change from a quasi-universality of the American Pilsner (Bud-cough-weiser-cough), a bit more diversity and more forays into more technical and difficult to succeed wines would do a world of good.

In the same way, I am just hoping that the transition towards Bio wines won't lead to a predominance of a new single style that would come to replace Bordeaux-like reds.

As a French - you can actually taste the difference between a 5$ and a 50$ wine, even despite the difference itaste. In France. Not in the US.

For some weird reason, the US wine is 100% posing, and statement about your social position. Almost no attention is given to the actual quality of the wine and a crapton of of 5$ wines taste significantly better than 10$-20$ or even 50$ bottles. In the same spirit, the "sommeliers" and wine vendors are eager to sell you the most expensive wine in their stock, not in discovering underappreciated wines.

Not surprising that no one develops a proper taste for wine here.

PS: And yes, don't chill reds. Ever. They are to be served at ~18C or ~65F. Otherwise, you won't be able to taste anything in them, even if they are good.

Wound you mind elaborating?