lars512

This is a popular theory, but not an argument with any evidence for why it's true.

It also doesn't address why people think there is a need for safety regulations, such as the concern of bio-capability uplift from closed and open models (we just came out of a pandemic not long ago), nor the cyber capabilities of new models that just showed bad misalignment and hacked multiple organisations. In the safety community, these are considered warning shots, and if we don't heed them and change practices future failures could be much much worse.

...and in particular it says that one of the reasons is that developers are refusing to participate in the non-AI branch, and when they do, changing what tasks they select to those where AI would be less useful.

Overall this suggests to them that the current speedup is likely greater than what the study could measure.

This kind of systematic distillation by a competitor can allow them to fast-follow you and pick up capabilities.

If you've invested in expensive capabilities training, of course you don't want this, so it's in Anthropic's economic interest to hinder it however they can, and that's enough to explain their behaviour here.

Anthropic seems to genuinely care about safety though, which for the rest of us means not having models that enabling easier cyberattacks, targeted scams, and the rarer but more severe risks like people trying to create and release new pathogens. This means walking a tight line, especially as models become more capable, and often wrapping a model in layers of defences against misuse.

If those capabilities transfer to a closed competitor model, all bets are off in terms of whether the competitor will apply the same defences.

If those capabilities transfer to an open weight model, not only will there be no ring of defences around the model, any defences you put into the model itself can easily be stripped away. So although it's nice to have capable open models, it will increasingly bad for us all if open models keep fast-following closed model capabilities as they have been, at least until we have solved the active research problem of keeping them safe.

This is all to say that, however you might feel about Anthropic, we might still prefer that they can deter this kind of distillation for now.

Here's data based on absolute poverty lines

Distribution: https://ourworldindata.org/grapher/distribution-of-populatio...

Share: https://ourworldindata.org/grapher/share-in-poverty-relative...

The share tells a story that poverty is decreasing at all levels, relatively speaking. The distribution tells the additional story that population has increased: there may be less change in the number of people at the $20-30 level and the $30-40 level in recent decades than the share alone would suggest.

Hetzner is great if you’re in Europe, is there an equivalent in the US or East Asia?

Isn’t that just assuming that people, rather than industry, is the main consumer? Perhaps there are energy hungry industrial applications that could move.

I had a good use case at Our World In Data for the public data pipeline, where one repo had the pipeline and one git-lfs repo had the build output of the pipeline. A git note added to a commit to the code pipeline recorded the hash identifying the built data.

Overall it felt elegant, and needed no maintenance after setting it up, but honestly it was never used. I think the need to look back in time was rarer than expected, and git notes being hidden by default didn’t help for awareness.

And here I thought it would be the antagonistic undecagonstring...

At Our World In Data we ended up using Buildkite to run custom CI jobs, integrated with GitHub, but on cheap, massive Hetzner machines. I can really recommend the experience!

When living in Stockholm, I came to appreciate the various levels of twilight and darkness, rather than thinking of day and night so strictly. The sun being low on the horizon also scatters light across the sky in ways that are very beautiful and last much longer than sunrise and sunset in Australia where I grew up.

This looks nice! It could be a replacement for me for duckdb-parquet, a plugin for Datasette that lets you run it on top of DuckDB instead of SQLite.

You can check out a range of important development statistics on India here: https://lars.yencken.org/projects/country-explorer/india

The one we're talking about today is "extreme poverty", which is the $2.15 purchasing-power-adjusted line. It's fantastic news that most Indians have surpassed this line, but it's also helpful to think of this line as just one rung in a ladder out of poverty. Life just above this line is still not great.

This chart, which shows how much of the population lives in different poverty lines for India, gives you a sense for the population as a whole. You can compare it to other countries to see their distribution, and China is probably a good comparison to make.

India: https://ourworldindata.org/grapher/distribution-of-populatio...

China: https://ourworldindata.org/grapher/distribution-of-populatio...

Despite progress on extreme poverty, you're right that there are still some 3 billion people in the world who cannot afford a nutritious diet, and likewise 3 billion people who live in energy poverty, meaning they have to cook indoors with solid fuels (wood, coal, dung) that damage their health and shorten their lives. It's important that we make progress on all these things in the coming decades. We absolutely have the power to.

The world is awful, the world is much better, the world can be much better!

To throw yet another option in, you could consider an LXC container per project, if they’re small and you don’t find you need Docker. LXC containers are basically multiprocess containers, unlike Docker’s single process containers, making them feel more like VMs and giving you a great dev experience.

Ah, minidiscs, they still have a kind of futuristic cool to them even now! When I was younger only the kids from Hong Kong had minidiscs and minidisc players.

I've been enjoying `files-to-prompt` by Simon Willison: https://github.com/simonw/files-to-prompt

There's often a lot of small fixes that not time efficient to do, but a solution is not much code and is quick to verify.

If the cost is small to setting a coding agent (e.g. aider) on a task, seeing if it reaches a quick solution, and just aborting if it spins out, you can solve a subset of these types of issues very quickly, instead of leaving them in issue tracking to grow stale. That lets you up the polish on your work.

That's still quite a different story to having it do the core, most important part of your work. That feels a little further away. One of the challenges is the scout rule, the refactoring alongside change that makes the codebase nicer. I feel like today it's easier to get a correct change that slightly degrades codebase quality, than one that maintains it.

Unfortunately, MacOS ships an earlier version of bash that does not include associative arrays, so they’re not as portable as you might like.

Hi HN! You may be interested in some related work:

"Millions of children learn only very little. How can the world provide a better education to the next generation?" https://ourworldindata.org/better-learning

Our topic page on Global Education: https://ourworldindata.org/global-education

Our wider catalog of education data: https://ourworldindata.org/data?topics=Global+Education

Reading between the lines of the post, performance reviews require a lot of trust between a person and their manager. I totally understand that in many workplaces that trust is not there, and you are forced to develop strategies for navigating those environments. High trust workplaces do also exist though, and they can be worth switching to or trying to foster.

Will it finally become possible to unsubscribe from the New York Times?

The easiest way to move anywhere is to apply for a job there, and if successful, let them guide you through the visa process.

That gives you a visa linked to your job. But keep unbroken employment in that country for 4-5 years and you will get permanent residence (pre citizenship), which frees you up immensely but requires you to not spent more than 1-2 years at a time outside that country.

If you get that far, you’ve done the hard work and citizenship is yours if you want it just by settling there longer.

Related question: when you're using databases with great offline sync protocols like this, how do you do schema evolution of your DB, especially in the face of different client versions that you cannot upgrade in lockstep?

My context here is having worked in the past on a mobile health app, and recalling all the pain we had around this problem.

Thanks for the encouragement!

Exactly this. What's plotted is the chance of surviving to that age then dying.

Oh, that's mostly my lack of Observable Plot skills. The 110 year old category is meant to mean 110+, so it covers all the years after.

Factors are of course overlapping, but you could think about what questions a life insurer would ask you, since they have money on the line and probably the best models.

Smoking is a really big one, given that it's estimate that 2/3 of life-long smokers die due to smoking and that continuing to smoke across a lifetime is estimated to reduce your life expectancy by 10 years.

Hi HN, author here. Glad you're enjoying it!

I put it together as a kind of Stoic practice to reflect on the fact that we will all die, but also with a data scientist hat on to try to get myself to think about life expectancy as a distribution, rather than an exact number.