time_to_smile

> Analogies are great for explaining ideas, but aren't that great as logical arguments.

In this case it is hardly an analogy. With respect to Information Theory, compression and machine learning are essentially the same thing [0]. If you can understand why taking a hi-res photograph of a lo-res image does not create more information (and therefore cannot be used to make a hi-resolution poster) then it's essentially the exact same logic explaining why outputs of a model cannot be used to improve that same model.

More to this point, which other comments have already pointed out, what's interesting about this post is it asks whether or not very, very large data sets can be reasonably compressed by models such that they can simulate training another model while requiring less data.

From an information theoretic perspective this latter case is both completely sound and of potentially big practical importance.

I do have to say, I'm a little surprised that basic information theory is no longer common knowledge among the HN community (more-so surprised at a resistance to it), especially among people interested in Machine Learning. Out of curiosity, do you have a background in ML and/or Comp Sci?

0. https://en.wikipedia.org/wiki/Data_compression#Machine_learn...

This comment reminds me of a time, many years ago, when someone wanted an the internal design team to make a big poster out of a low res image.

The design team explained that they couldn't blow the image up to the size of a poster because it would look too grainy as it was extremely low resolution.

The individual demanding the poster replied with a 'brilliant' solution: "Why don't you just take a picture of the low-res image, and then use that hi-resolution picture to blow up the image!"

The same applies here. You need more information to learn the signal better, by definition there is no new information in the model.

Good product people have always been visionary, lead by a strong sense of design and understanding of how to meet the user needs.

Product Managers are the modern day bureaucrats created to reduce the individual impact on product and make PMs hot swappable like engineers (who I also think are better when they they have the aforementioned skills).

Google should be considered in object lesson in the problems of trying to bureaucratize product creation.

You're just describing a "community" and the "shared values that define them".

So I think we agree: tech people came in, attempted to displace the local community and trampled on their values (and claimed them "xenophobic" for resisting).

And now that the tech community is withdrawing what's left is a broken city without the community and shared values that once defined it.

The tech community wanted SF to be a shopping mall for rich tech workers, some parts of the community resisted being turned into a shopping mall, they lost and now SF is rotten.

> Xenophobic natives

This claim isn't backed up by the demographic info of SF [0], and SF has historically been a very diverse, immigrant friendly city.

Percentage of US born (as well as specifically California born) residents of SF started to rise with the tech boom. The lowest number of US born residents of SF was 2000, it rose slightly by 2010 and continued to increase into 2020.

0. https://en.wikipedia.org/wiki/Demographics_of_San_Francisco

Anyone who spent considerable time in SF before the most recent tech boom would likely argue that it was the tech industry that caused the rot in the first place.

The foundation of safe cities is vibrant local communities with historic roots. Parts of the city where people choose to have and raise kids. Places with a sense of shared community where neighbors look out for each other.

The rapid influx of extremely high income individuals transformed SF over the course of the last decade. Uprooting those communities and replacing them with largely transient techworkers with no real interest in forming communities that don't help them to increase their TC and level up with their next new role.

Then when tech declines you really see the impact this has.

SF was treated as basically a luxury mall for rich tech workers to get whatever they want whenever they want. As tech workers start to leave it's no wonder the city feels like an abandoned mall.

Just wanted to second this experience.

I'm one of those people that has perpetually wanted to use linux as my main desktop, but inevitably find myself dual booting Windows for gaming, and then just making my desktop a Windows gaming box and then using a Macbook for dev work. I think I've had about 3 major runs of trying to be Linux desktop only, and that typically lasts just a month or two.

For the past year I've been using exclusively Linux (Pop!_os) as my desktop for everything I do, and that has involved quite a bit of gaming. It's really insane how giant a leap forward Proton (driven by the need for Steam Deck compatibility) is as far as Linux gaming goes. It does occasionally take a bit of tweaking to get things to work just right, but nothing too crazy.

It's also worth mentioning that, at least for me, System 76 has really solved the problem of "just open the box and turn it on" support for Linux. Extremely high build quality and most important, seamless support for Nvidia drivers. It is a blast to spend the day building deep learning models and then make use of that nice GPU to play AAA titles at max settings.

Any gamer who has been wanting to go full Linux Desktop for a while, it's really quite possible now.

There are plenty of bigger "human ending" concerns on the table right now than AI and we certainly aren't pausing anything for those.

I think this is more marketing copy than virtue signaling.

They don't want you to use the app because it's better for you, but because it allows them to collect much better and more fine grained information about your usage (to sell to advertisers and use the target you) and you can't block ads.

I suspect the only reason they keep old.reddit.com alive is the is enough added content from users of that subdomain, while these users are also, at this point, the most likely to just stop using reddit altogether if it was remove.

I assume eventually that math with work out so that it makes more sense to just get rid of 'old' entirely.

If BNPL was the same as credit cards, then why not just use credit cards?

BNPL often ignores credit check or specifically targets groups that credit cards would deem too risky.

For me this doubles the problem with the BNPL space: not only are the predatory, but I have serious concerns about their own risk management.

I knew of a start up that was experimenting with running their own BNPL service with zero credit checks. The users that were being targeted had a 30+% default rate (because other users had credit cards) but the upper bound of interest that could be charged was around 25%... the company seriously didn't understand why they weren't able to make profit on this product. But hey, revenue did grow which is all VCs cared about so it worked for them.

What about Peter Ward's research do you find unconvincing?

Of course it's absurd to treat it as a certainty, but the geological evidence of the destruction to life caused by past period of rapid CO2 escalation are at the very least a bit concerning.

> would lower quality of life beyond what global warming will do.

Source on this? Climate crisis could quite possibly lead to extinction of humans and most life on Earth [0].

Now I do want to emphasized the word could. Extinction events are certainly "on the table", but certainly not a certainty.

However this is the true essence of the reason why we'll never stop climate change. The immediate cost of stopping or even slowing is high, and certain. The cost of not stopping it, is often considered higher, but is always probabilistic and less certain in nature.

Compounding this is that the more we rely on non-renewable hydro-carbons, the more painful it is to reverse or even slow our march towards catastrophic climate change, while at the same time the cost of not stopping it becomes less uncertain.

For example in 1900 it would have been incredibly easy to slow economic progress in order to prevent climate change, but the uncertainty around the cost of not slowing progress was very high, so the perceived expected value of stopping our march toward climate change was much lower than the present time expected cost.

It is strange to see calls for moderate pragmatism when our current state has been accelerating CO2 emissions. We're, quite literally, stepping on the gas to meet whatever future awaits us.

0. See Peter D. Ward's "Under a Green Sky" for one example scenario.

You must have a smarter toaster than I do. My toaster is a completely deterministic mechanism that knows nothing about my past toasting experience, nor makes any inference about future toast.

A spam filter:

- Is provided historic information about the problem

- Learns from this information to construct a basic model of language

- Is then provided unseen information

- Then makes a probabilistic decision based on a degree of uncertainty.

The exact model behind the spam filtering can be extremely simple (naive Bayes) but could easily (and probably will without you even realizing it) expand to include things like GPT.

A spam filter is making decisions under uncertainty with new information based on patterns it learned from previous information. If this doesn't fit your definition of "AI" then I think if you understood what was happening under the hood I don't think you would consider GPT to be AI either.

If you toaster does learn your toast preference over time from your toasting behavior (unlike mine) then I would consider that AI as well.

> AI is a feature, not a product

Part of the reason for this is that very often the best applications of AI are really invisible to the user, making everything a little bit better, quietly behind the scenes.

The same is true for software in general. The best software products automate loads of stuff behind the scenes so from the user perspective they just click a button and the thing they want to happen, happens.

Perhaps the most successful and important AI product is a great example of this: the spam filter. I bet most younger email users don't even realize how much value this old AI tool provides them unseen.

But the trouble is making effective AI products is not "sexy" right now, and if your team ships one it won't get any credit in a big company. I've had my PM instantly shoot down plenty of interesting applications of AI that wouldn't be very visible to the customer or larger product-org.

Well the real truth is they were right 60 years ago. That was probably our last realistic chance to avoid catastrophe. But we did nothing and as much as doom and gloom might sell newspapers, true despair does not.

Every article since they has made use of increasingly unrealistic solutions to create the illusions that there is more time.

And again, the "apocalypse" is not a single day's event in the case of climate change. Given that fossil fuel usage continues to climb as well as CO2 (and now methane) emissions it's clear we have chosen the worst case path. The only limiting factor is how much hydrocarbons we can suck out of the ground.

We have committed to destruction of our ecosystem and transformation of the biosphere in ways not seen in millions of years on this planet.

You don't need to read the news, you can just look at outside and watch it continue to happen at an escalating rate for the rest of your life.

> about 1 out of 100 people i interviewed actually know how computers work, we still hired about 10 out of 100

This is very reminiscent of the dotcom bubble. If you knew how to turn a computer on you could get a relatively high paying tech job.

The quality of tech talent has dropped significantly in recent years, even (maybe especially?) among the leetcode grinding crowd. Software engineers have already replaced themselves with AI by turning into interviewing machines that know how to survive in a large tech company but if asked to build and ship a product from nothing would be completely frozen.

I really miss programming in the post-dotcom shadow. Most people in tech where in it because they loved writing code and solving hard technical problems. TC was literally not a term at that point, virtually nobody had RSUs, comp was fine but nothing amazing. There were plenty of jobs that paid better than writing software so all the people looking for a quick buck fled there and left coders alone.

> tools like GPT potentially disrupting our careers on a timescale best viewed on the scale of months.

GPT is cool, and has a lot of potential uses, but don't be scared by the hype.

The biggest impact GPT is going to have on most people's software careers is that everyone will have to include "Prompt engineering" on their resumes in the very near future, and data science/ML folks are going to have to answer a lot more questions about transformers by people who have never heard of "Attention is all you need", even though the deepest they'll have to dive into the subject in practice is a few api calls.

That's because this bubble is much larger than the dotcom and has impacted far more than just tech (lookup "Everything Bubble"). A major part of this is companies still have a lot of cash.

When you pull up the financials for many newly IPO'd companies you'll see that many of them have questionable ability to ever generate profit, and most certainly not at the size they are at. However all of them have pretty healthy cash reserves from the (obviously overvalued) IPOs they just went through.

Facebook/Meta was the poster-child for "don't worry about profit, just worry about growth!" For years it was assumed that Facebook proved this strategy made sense, but even today we're seeing this giant falter. Even if Facebook pulls it off, it doesn't mean that many of the "growth is all that matters!" startups following in its shadow will.

We'll see companies fail, but this bubble will take a lot longer to unwind than the dotcom bubble.

We have been getting warnings like this for years and for years things have been getting visibly worse.

I don't know how you "have to force myself quite hard to take this seriously" when the environment around us has been degrading faster than scientist have been claiming for years.

Maybe you're confused by how events related to climate catastrophe unfold? It's not like a nuclear bomb where one day everything is fine, then the next every thing is gone. That case is different in two major ways: the break down is instantaneous and a new normal is almost immediately established.

The final warning for climate is about a process that once started can no longer be stopped.

You don't have to "trust the science", you can literally watch it happen as the arctic heads rapidly towards a blue ocean event (see "arctic death spiral"), crop failures have been increasing, California has suffered extreme years of drought followed by this current season of incredible floods, lake mead reaches a lower level than it has since it was filled, Texas is subjected to extreme weather (in both directions) causing year after year failure of the power grid.

These things will continue to happen, will happen more frequently and at greater magnitude for the rest of your life.

If you feel unconvinced it's because you're in a state of denial (which in part of grieving so not terribly surprising or uncommon).

Very interesting paper, thanks for the link!

The other major thing missing from Chat GPT is that it doesn't really "learn" outside of training. Yes you can provide it some context, but it fundamentally doesn't update and evolve its understanding of the world.

Until a system can actively and continuously learn from its environment and update its beliefs it's not really "scary" AI.

I would be much more concerned about a far stupider program that had the ability to independently interact with its environment and update it's believes in fundamental ways.

Well I for one, as a person with sophisticated tastes, find the emperor's clothes to be absolutely fabulous, none finer have I ever gazed upon.

Thanks for sharing this.

Quantitative finance really is a fascinating field, not because it will make you rich, but because you can dive much deeper into understand exactly what the market believes about the probability of different events.

Of course what the market believes doesn't have to be correct, but nonetheless very interesting to dive into.

> there is no meaningful risk that these startups will not be made whole in the coming weeks.

Wow, two very bold claims here:

1. they will be made whole

2. in a few weeks

For 1. I think this is very unlikely to happen. Absolutely account holders will get some money back, but I would be pretty surprised if it was 100%. Regarding number 2. I would be even more shocked if anything more than the FDIC insured amount was returns within "a few weeks".

However you are claiming that there is "no meaningful" probability these will not both happen. Do you care to elaborate on this more because I have heard nobody with any experience in this space making claims like this?

Though it is widely publicized that he was pretty aggressive at getting YC startups to move their money out yesterday.

It's not altogether unlikely that other groups of startups, without an aggressive push to pull out of SVB, are in more trouble.

That is also a lot of startups in a very short window of time. Startups don't exist in isolation, and, for better or worse, often actively try to purchase each others products when possible. It is an ecosystem.

I work at an IPO'd SaaS company, but a lot of our customers are startups. If 30% of our startup customers suddenly can't pay their bills that impacts us in a big way. On top of this, like most companies, we were already concerned about how we were going to meet profitability goals.

And again, the short time frame means not a lot of time to respond and adapt.

> Felt like a lot of dumb panic.

A lesson everyone should have learned from the toilet paper scarcity crisis during the pandemic is that it doesn't matter if it's rational or smart, if you want to keep your ass clean make sure you're in the front of a line when people are panicking.

You could make a pretty compelling argument that the last decade the entire tech sector has been propped up by "dumb exuberance", but people aren't complaining about the money they made during that period.

When was the last time you picked a random number from an actual real number line?

Try sampling random floats in your programming language of choice and see how long it takes to get an integer (you will eventually get one).

Then consider that floating point numbers represent only a finite subset of the (countably) infinite rational numbers. And then consider that the set of rational numbers is unimaginably smaller than the set of irrational numbers (which is the other part of the reals).

The fact that integers have measure 0 in the set of real numbers only seems confusing if you're thinking of everyday operations rather than thinking about mathematical abstractions.

It's remarkable how deeply ingrained in our worldview mind/body dualism is, even though I think on paper most people don't accept this as an axiom.

If anxiety is not created by the body where does it originate from? extra terrestrial beings? the immortal soul? purely in our surrounding environment?

All of these (except the third option) require pretty impressive assumptions to even be a workable hypothesis.

> It's no good just to do meditation and calm your mind if your body and nervous system is still holding on to so much stress.

I just want to point out that for nearly all of its history yoga is primarily about meditation, with the asanas being preliminary, quickly covered material in most classic texts.

The Yoga Sutras of Patañjali are a good reference for anyone curious (though the sections on the siddhis is... interesting).