IKantRead

The old crypto chants were that crypto would lead us away from the ills of traditional markets/finance.

Glad to see that it's been watered down to "just as bad as" arguments.

Who in the world is still long on crypto?

In the past few years we've seen ample evidence that crypto is largely a complete scam, and likewise strong evidence that none of crypto's hoped for value will come to fruition (we didn't see it useful for fighting inflation, it's not being used to avoid sanctions, it certainly isn't being used as a currency, etc).

The fact that crypto still has any market value, and that companies like coinbase not only exist but have had a stellar year defies the imagination.

I get a few years back when there was still a lot of speculation/optimism, but clearly today everyone see that it is just a con. Today even my most cynical view of markets seems naive.

Not to my knowledge, which is why I used the word "suspected" since I think this falls on the "makes intuitive sense, but would not surprise me in the least if it turned out to be completely incorrect" category of hypotheses.

I consider "suspected" to be the least level of evidence while still taking something into consideration as a potential cause. A suspected murderer might not even have been arrested, let alone convicted.

We do know that sulphur emissions have a global cooling effect, and we do know that sulphur emissions recently were reduced, so it's a reasonable hypothesis from first principles.

To be clear, I'm absolutely not promoting increased sulphur emissions as a solution to our climate problems. Moreso pointing out that all those emissions are potentially masking the true severity of our current predicament.

I never claimed that they "require a traditional publisher", in fact I explicitly point out that you can pay for these yourself (though I can't imagine putting together a good team of editors without having prior publishing experience).

My point was that, in response to the parent claiming there's nothing traditional publishers offer, these are things that traditional publishers do in fact offer an author. If you write for a traditional publisher you mostly have to just worry about writing, and, unfortunately, marketing these days.

Which is fascinating because sulphur emissions counteract (mask might be a better term) global warming. Reduction in sulphur emissions is suspected to be one of the main culprits of this years sudden rise in Earth sea-surface/land temperature this year.

Wild when you see just how much emissions are still being released and still presumably cooling the Earth, meaning the effects of climate change we're seeing now are still likely a dampened version of the true long term impact.

I know plenty of authors and none of them are subsidized by wealthy families. All of them do it part time in the evenings out of a labor of love.

It is worth pointing out that there's nothing particular odd if it were the case that writing was subsidized by wealthy families. For the vast majority of the history of writing, writing was subsidized an left to monks, philosophers or aristocrats. It's only been in the relatively recent time period that writing was a potential occupation for anyone interested with enough skills/talent.

only reason for using a traditional publisher is the cash advance then?

A few really important things come to mind:

- Editing. I'm not talking about mere copy editing which you can get done reasonably cheaply, but rather having an editor that is reading through everything and giving feedback is hugely important.

- Layout and printing of the book There's a lot that happens between writing and having a polished book in your hands. You can contract all this out but it adds a lot of work.

- Distribution. While the burden of marketing a book has increasingly fallen upon the author these days, if you want your book to be on the shelf at your local Barnes & Noble, then your much better off going with a traditional publisher.

- Prestige. Like it or not, the vast majority of people on Earth still look down upon self publishing. For some types of books this is less important: technical books and fantasy fiction books can go without in many cases (but if you want to use your book for credibility in something like consulting you'll still want a traditional publisher). But if you want to write on a serious topic it helps a lot to have an academic press publish your work, or if you want to really pursue writing literature you at least want some publisher that is recognized in your relevant community.

Currently I think the only really good use cases for self publishing are the fantasy fiction and niche technical book markets assuming you already have an audience. And even in those cases there are plenty of reasons to go with traditional publishers over self publishing.

I've always found it fascinating that geophysicist and earlier advocate for Bayesian methods, Sir Harold Jeffreys, didn't believe in continental drift and plate tectonics because he felt there was no known source of energy on the Earth massive enough to explain this movement. [0]

He remained an opponent until death (at which point continental drift was widely accepted) which is both a testament to the literally unbelievable energy behind seismic activity and the importance of updating your Bayesian priors as you gain new information.

0. https://en.wikipedia.org/wiki/Harold_Jeffreys#Opposition_to_...

This quote in particular struck me as relevant:

And now Google is unusable: using LLMs even just as a compressed form of documentation is a good idea.

Beyond all the hype, it'd undeniable that LLMs are good at matching your query about a programming problem to an answer without inundating you with ads and blog spam. LLMs are, at the very least, just better at answering your questions than putting your question into to google and searching Stack Overflow.

About two years ago I got so sick of how awful Google was for any serious technical questions that I started building up a collection of reference books again just because it was quickly becoming the only way to get answers about many topics I cared about. I still find these are helpful since even GPT-4 struggles with more nuanced topics, but at least I have a fantastic solution for all those mundane problems that come up.

Thinking about it, it's not surprising that Google completely dropped the ball on AI since their business model has become bad search (i.e. they derive all their profit from adding things you don't want to your search experience). At their most basic, LLMs are just really powerful search engines, it would take some cleverness to make them bad in the way Google benefits from.

A lot to be said for not defaulting to data frames, in both r and python

I would even add especially in Python. The main issue I have found is that pandas heavy code is just not as easy to integrate into other Python tools/features/abstractions as code using mostly numpy, dictionaries and various comprehensions to do the vast majority of your work.

As a heavy pandas user for several years, I decided about a year ago to not import pandas by default and instead treat most data problems like regular python problems. I've been genuinely surprised as how much easier it is to create useful abstractions with the code I've been writing, and also how much easier it's been to onboard non-DS devs into the code base.

There are a few obvious cases when Pandas is very helpful, and I'll pull it out in those places, but I've been able to do a tremendous amount of data work in the last year and used very little pandas. The result is that I have an actual codebase to work with now rather than a billion broken notebooks.

Recent experience at a fairly young startup has shown me that open office culture has also started to breed a very different type of programmer.

People will often be pairing nearly all day long, any claim that you need a moment to focus and think about a problem is met with perplexity, every idea should be shipped to prod asap, while tests exist the idea of performing basic QA/manual testing on your own work is only used in the most extreme cases.

Contemporary startup engineering culture is best described as frenetic. It certainly feels hyper productive (if not extremely exhausting for a more traditional, introverted programmer), but I've started to notice a fairly large amount of that "productivity" is fixing mistakes a more focused programmer would have avoided.

I suspect the long-term impact of open offices my be even more deleterious than it's impact on the focus of individual programmers.

It's worth pointing out that most of the best science happened before peer review was dominant.

There's an article I came across awhile back, that I can't easily find now, that basically mapped out the history of our current peer review system. Peer review as we know it today was largely born in the 70s and a response to several funding crises in academia. Peer review was a strategy to make research appear more credible.

The most damning critique of peer-review of course is that it completely failed to stop (and arguably aided) the reproducibility crisis. We have an academic system where the prime motivation is the secure funding through the image of credibility, which from first principles is a recipe for wide spread fraud.

I'm pretty sure Sam Walton didn't establish Walmart with the hope of being acquired.

It's bizarre that we live in a time where we can't even fathom a business that is fundamentally very profitable, we just envision growing the company until it's attractive enough for someone else to take on the unsustainable cost of running the business: either get acquired by a large company or hoist your debt onto the public market.

Investment really did used to be about more than a complex "greater fool" game.

As an ex-smoker, I also used to just toss cigarette butts anywhere without thought. Decades after quitting it's still surprising to me that I ever thought this was normal (I've never, otherwise, been a big litterer).

At least when I began smoking in my early teens this was just what smokers did. When you were done smoking you just tossed the butt on the ground and stomped it out. I remember being annoyed when someone scolded me for littering, at the time it somehow felt different but I admitted the logic didn't quite work out.

I suspect a lot of this behavior originated from behaviors established before filtered cigarettes were the norm. Just tobacco rolled in thin paper, especially when burnt to the end before it burns your fingers, will likely break down after the first heavy rain. Filtered butts however, hang around for a long time.

I can see how it feels like cheating to coax the model to produce the answer you want. But... it's not!

If it's for a single example, it is absolutely cheating. As an AI engineer this is a particular point of frustration where people complain because a large system can't return the result they want, when they were able to get the answer they wanted on their own with a lot of prompt hacking.

Each prompt is basically a point in latent space, and if you're "tweaking" the prompt what you're really doing is just re-rolling the dice until you land in a neighborhood closer the answer you want. You're not better at prompting, you just got lucky and are confusing that for insight.

Now if you're specific prompting trick works across a suite of evaluations, then you are probably on to something. But what people are doing in most cases is equivalent to performing some ritual before pulling the handle on a slot machine and then, when they finally win, claiming that they finally stumbled upon the correct ritual.

A good QA person is to a software developer as a good editor is to a writer. Both take a look at your hard work and critique it ruthlessly. Annoying as hell when it's happening, but in my experience well worth it because the end result is much higher quality.

I might just be too old, but I remember when QA people didn't typically write tests, they manually tested your code and did all those weird things you were really hoping users wouldn't do. They found issues and bugs that would be hard to universally catch with tests.

Now we hoist QA on the user.

Working with younger devs I find that the very concept of QA is something that is increasingly foreign to them. It's astounding how often I've seen bugs get to prod and ask "how did it work when you play around with it locally?" only to get strange looks: it passed the type checker, why not ship it?

Programmer efficiency these days is measured in PRs/minute, so introducing bugs is not only not a problem, but great because it means you have another PR you can push in a few days once someone else notices it in prod! QA would have ruined this.

don’t waste your time writing your own neural net and backprop.

I don't think you should be combining writing a neural network with doing backprop since I don't know anyone working with serious ML who is not using some sort of automatic differentiation library to handling the backprop part for them. I'm not entirely sure people even know what they're saying when they talk about backprop these days, and I suspect they're confusing it with gradient optimization.

But anyone seriously interested in ML absolutely should be building their own models from scratch and training them with gradient descent, ideally start with building out your own optimization routine rather than using a prepackaged one.

This is hugely important since the optimization part of the learning is really the heart of modern machine learning. If you really want to understand ML you should have a strong intuition about various methods of optimizing a given model. Additionally there are lots of details and tricks behind these models that are ignored if you're only calling an api around these models.

There's a world of difference between implementing an LSTM and calling one. You learn significantly more about what's actually happening by doing the former.

I have a few friends who had major liquidity events happen to them and they each own 3-5 homes as investment homes. I believe these properties were paid full in full and in cash, so no mortgage.

This small proportion of my friends is mortgage free, but they own multiple houses so a fairly large proportion of my friends houses are mortgage free.

As this other comment pointed out [0] 40% of homes are mortgage free, but that says nothing about home owners.

Anecdotally I know that none of my neighbors are mortgage free. However I also have a few friends that had major cash windfalls things like startup exits or time working in hedge funds, each of them owns 3-5 houses outright as investments.

So, with this small sample I can easily see that of the total number of homes owned by friends/neighbors, 40% being mortgage free sounds about right, but far less of my friends/neighbors are mortgage free.

0. https://news.ycombinator.com/item?id=38612670

Of all the ex-FAANG people I've worked with, I've found the ex-Amazon ones to be, by a wide margin, the least impressive group.

I'm not huge fan of Google, but will readily admit that, on average, the ex-Googlers I've worked with have tended to be great engineers and very sharp people.

So, at least from that anecdote, there does seem to be a difference in each of these hiring filters.

I feel the same way and would also never work at a FAANG.

That said, I suspect this is would be viewed as a benefit of the process since I'm fairly confident Amazon and the like don't want people who feel this way working for them regardless of the talent of those individual people.

I think they're saying the interaction effect of doing drugs, drinking and considering oneself a victim is negligible on overall success.

It doesn't seem like a bold claim to me. I've known plenty of poor, struggling working class people. The only impact drinking or claiming they're a victim have is influencing how much fun they have at parties.

I've happen to know plenty of wildly successful people who also drink and/or think they're victims. Again, primary effect I've noticed seems to be how much fun they are at parties.

But is there is something inherent to Lisp that makes it "clunky" to work with matrices?

s-exps are basically coding at the level of the AST, which for many programming tasks is a powerful level of both abstraction and control.

But this interface tends to work much better on tree-like structures (the most simple of which is the list, which in it's simplest case is a cons cell, Lisp's most fundamental particle).

Most programming tasks can boil down to tree manipulation, but large matrices feel a bit out of place with this interface because you are basically to working with arrays (which in it's simplest cases is a pointer to a memory address, which is C's fundamental particle). C like languages (I know, technical ALGO-like), tend to feel more natural for these tasks.

But of course if you can manipulate trees easily, you can manipulate code easily which gives you...

Lisp can be at least as high level and expressive as any other language.

In practice, for Lisp this is also a weakness. Again, I love Lisps in general, and of course any Lisp and especially Common Lisp (with it's exceptionally excellent macro system) can do anything you want.

However, my (and many others) experience has been that this leads to is it becomes very easy to create code that is amazing for the original developer, but very difficult for new developers to get their head around. Hence the adage "Lisp is optimal for team sizes of one."

You could undoubtable build the worlds most elegant interface to working with matrices in Lisp, but now you essentially have a new language.

As a huge Lisp/Scheme fan, I don't find it surprising that Python became the language of choice for neural networks over Lisp.

NN in general just boil down to doing lots of linear algebra, which is a lot about mutating very large matrices. For this you really just want a wrapper around BLAS/LAPACK, so you can leverage existing optimized libraries.

Working with matrices in Lisp always feels a bit clunky, and doubly so when you want to do a lot of stateful operations on them. The real compitetor in this space to Python is something like Matlab, which has probably the best interface for doing linear algebra (but is worse at everything else).

The one area where Lisp and Python both shine is the ability to perform automatic differentiation. Lisps are great for this task since it's all symbolic manipulation. However, this is only important once you have a solid interface for working with matrices.

If you want to get a feel for the difference I would suggest reading through the surprisingly excellent The Little Learner. I think you'll find that while it really demonstrates the power of Scheme (Racket in this case) in areas where it excels, you wouldn't want to use the framework in that book for anything other than toy examples.

trying to solve innovative business problems with incremental improvement.

This describes pretty much every AI startup I know, both good and bad.

I suspect the places where AI will really succeed will be places that entirely change existing workflows and user experiences. This requires a fair bit of research and experimentation, but most startups don't have the appetite for that, and VCs are putting on the pressure to do AI stuff fast.

A lot of AI potential is being squandered now due to a simple lack of patience and planning in the rush to market.

Expensive restaurants, fancy restaurants, and 3-star Michelin restaurants are not the same things.

I've eaten at a few 3-star Michelin restaurants and the experience at each was worth living imho. 3-star means not only is the food world class, but the service is as well. I still remember fondly each thing I ate, and just as importantly, the interactions I had with the chefs throughout the evening.

It happens to be the case the excellent food and service does come at a price, so Michelin 3-star places tend to be expensive. Likewise people tend to dress up when they spend a lot so they tend to be "fancy", but those are things that happen as an artifact of the quality of experience they provide.

I've been to plenty of places that are expensive and very "meh", and I absolutely loathe "fancy" restaurants that pretend they are offering you some privileged experience (which is usually a laughable facsimile of the experience of real high-end dining).

I'll add that there are also diminishing returns: The experience of sitting at the counter of at Masa, with only two other couples why your personal chef hands you some of the best sushi in the world is incredible, but there are other omakase places in NYC that offer a nearly equally amazing experience (especially if you go on a quiet Tuesday night) for 1/3 the price, but I'm very glad I had that counter experience at Masa.

A world where we let people's failings and faults dominate their success and virtues is a world inherently destined for incredible mediocrity and the creation of only tepid works.

Provided we can keep riding this hype wave for a while, I think the logical long term solution is most teams will have an in house/alternative LLM they can use as temporary backup.

Right now everyone is scrambling to just get some basic products out using LLMs but as people have more breathing room I can't image most teams not having a non-OpenAI LLM that they are using to run experiments on.

At the end of the day, OpenAI is just an API, so it's not an incredibly difficult piece of infrastructure to have a back up for.

I would add that this creates the problem, which I think most tech companies have forgotten about: if you try to attrition out staff and the market is competitive, you'll only attrition out your best employees.

Last place I was at everyone was starting to get really frustrated with the way management was handling projects and all but one person on my team wanted to quit. The problem is that the talent on the team was largely so-so. I was the only one that found work and left, but that only served to further decrease the morale of the remaining members who felt both more trapped and now without coworkers they liked. Across the org I noticed pretty much every other team started to experience similar issues.