seiji

But, why? Nothing about "AI safety" is remotely useful. It would be like researching "alcohol safety" (something everybody has access to and nobody can restrict) or researching how ants can influence "human safety" (tiny things trying to influence massive things). It's nonsense no matter which way you think about it unless you just want to write SF novels.

Open-source infrastructure.

Sadly, we don't need "research" to study how it works, we need direct payments to contributors. As a "startup culture" we need to stop the trend of billion dollar companies using free open source labor to reduce corporate costs and boost corporate profits while paying highly skilled professional volunteers nothing.

Yudkowsky Ambition

A key part of _research_ is you need to have a plan of attack. Saying "We're going to fund research into antigravity, time travel, and zero point energy" sure sounds like fun, but it's an absolute waste of time.

One difference: C->x86 is a static translation layer. Other network/system things dynamically change out from under your "designed" system and alter threat/security/disaster/reliability/consistency models in a potentially unpredictable combinational fashion.

Saying "cloud abstraction" or "I trust this API and don't care how it works" is basically committing every https://en.wikipedia.org/wiki/Fallacies_of_distributed_compu... and just saying "X can't break because we use provider Y who guarantees they can violate the laws of physics for us!"

I've done the whole norovirus dance once too. It really does knock you out for a week or two. (I started feeling sick when walking around outside one morning, decided to take the subway back home, the first sickness struck while waiting for the train—I just had to throw up on the subway tracks. Pretty awful. Some nice bystanders gave me an unopened bottle of water they had with them which was nice.)

Most developers I've seen reject even learning about networks or DNS or operating systems or databases. Such willful ignorance boggles the mind, but they are praised because their goals are shipping half-broken things as rapidly as possible to flip upwards for those oh-so-tasty acquihire payouts.

We even saw this week how overconsumption of convenience APIs can put entire companies in danger when those privately controlled convenience APIs just decide to shut down one day. Convenience of immediacy always seems to trump connivence of long term stability.

But really, why?

Is it just institutional laziness on the part of all developers? We had reliable rsync CPAN mirrors in 1995. In the early days of the Internet, companies would mutually host secondary DNS for each other to be more reliable. For some reason, we've forgotten all about reliability and disaster recovery and geographical distribution. Now the collective programmer mindset with regards to global infrastructure seems to be "lol, we're too dumb to make things work, let's just outsource everything to closed source, for-profit companies and hope for the best."

Very few companies build their "data centers" (apple, google, amazon, NSA, actual 'data center' companies, etc). Most companies rent cage space in a larger data center and call that their "private data center." Smaller companies will rent a few dedicated servers or colo half racks from other resellers.

If a company can't maintain their own internal tools and self-hosting servers, why does the same company think it can run reliable services for users?

Not putting the core of your business on a remote platform is disaster mitigation 101.

I'm pretty sure elder care isn't indefinite. Maybe we end it after this current generation of old people (our parents/grandparents) are no longer around.

One of the biggest social problems we'll face in the next 10 years is millions of people with dementia who can't take care of themselves just... existing. What do we do?

One fun story I've heard is a doctor gave an old woman with full blown dementia an operation to keep her alive, then she lived another 15 years just completely out of touch with the world. The family had to sell all her belongings (eating through their inheritance too), then pay even more to keep her in full time care. It makes no sense, but that's the standard set of rules. Little known fact: before medicare will pay out for extend services, they force your relatives to sell all your belongings (property, land, investments, etc) to pay for medical services before any other payments kick in.

I'd love to stop paying $3500/month in rent, but any property to buy around here has a median price of $1.2 million. Unless we all move to rural Indiana, housing + healthcare is designed to eat 100% of your after-tax income with no room for further growth.

(of course, that's assuming you're just a putz serial employee like myself and not a wise investor who can share the success of 1,000 companies at once without having to do any real work yourself.)

It's the same reason why on a Jet Blue flight, my phone shows me geolocated to Long Beach, CA for the entire flight (if I don't hold the phone up to the windows).

Phones prefer wifi positioning, and that's based on the base station mac addresses. If you are in a highly wifi dense area, moving one AP can auto-propogate its new location by phones reporting in bulk groups of mac addresses they scan at once back to your location provider. ("30 mac addresses detected are know to be in Dublin, but one mac address detected is known to be in New York. Let's update the New York mac address to be Dublin now.")

The auto-location-updating is one reason SkyHook wireless died. Once they did the initial scanning, the entire wifi location database can be made reasonably self-updating (and other correlative updates can happen with your drive-every-street-in-the-world mapping vans much less often.

Wifi position is more accurate than GPS positioning (plus, faster to locate and less power intensive), so it's always preferred first. GPS positioning is worse (accuracy/power requirements), so GPS tends to be a fallback if you're in a "broad location" category like cell tower only positioning or zero-location positioning (airplanes, mountain tops, etc).

The cycle will continue for the foreseeable future: people want to communicate, someone creates a new communication app for modern platforms (or a cheaper way of communicating than had come before), old communication apps fall out of fashion, new communication app stagnates, someone creates new communication app for more modern times, rinse, repeat.

One day we'll have a majority marketshare 3d telepresence app, a majority marketshare direct mindsharing app, a majority marketshare sensorium sharing app, etc. It's just things people want that we don't have the technical capability (or wider vision) to instantiate into reality yet.

As far as "value add" or "marketplaces" built around chat apps, just ask yourself how many AOL Channels are still profitable today. We're stuck living in the times we inhabit, so sometimes there's short term value to capture, but in the longer term, everything fades.

Things like Docker and systemd are exactly EEE land grab lock-in plays to redirect "open source" value back towards private entities and unilateral control structures (instead of cross-vendor standards bodies, because standards bodies can't make individuals into billionaires).

How do we as lowly developers with effectively $0 net worth fight companies with billion dollar war chests? Over time, better does win (see: the collapse and accelerating irrelevance of Microsoft), but in the short term it's likely we'll get stuck in another 10 years of platform "dark ages" until we see the light again.

If I tell someone I made a deep learning system, it conveys no details other than it's something besides a single linear discriminator.

deep learning is in the "Peak of Inflated Expectations" phase of our never ending tech hype cycle. Like most buzzwords, it holds meaning for practitioners, but when marketing dweebs with no knowledge are hired to write endless clickbait about technical terms with no underlying understanding, it dilutes the entire vocabulary ecosystem until even the practitioners can't describe what they do using their own terminology anymore.

Basically, if you can write "while (true) { doStuff(); }" (or an equivalent TCO version), your system is turing complete.

So, all the "deep learning" fads aren't even turing complete (if we don't consider RNNs "deep learning") because they are feedforward/backprop networks of explicitly limited depth.

Humans are turing complete, but only for limited time spans. Humans get tied up in a version of the halting problem called "this is dumb let's do something else." It's difficult for humans to execute "find the highest prime number" without other needs derailing the process over time (boredom, bathroom breaks, biological death).

Can also modify the idea a bit to create an "unknown person" warning system if it detects unauthorized device IDs broadcasting within range.

I have the same problem. It's a numbers game. You just have to interview at 8-12 places until either a.) you find people who understand you or b.) you find a company with lax hiring processes.

Here's a review of my 2013 job search (all rejections): https://matt.sh/searching-2013

It's so early in the evening to have reached Peak HN already.

It's almost like everybody forgot about https://matt.sh/howto-c#_formatting already. :(

Exactly. GitHub offers no support for on boarding new issues, so a lot of well meaning misplaced questions can pile up quickly. There's no GitHub issues MOTD or "have you tried X, Y, Z first?" or even a way of saying "issue tracker is for provable bugs only — not for asking why our software doesn't compile on your 16 year old version of RedHat."

We still need to figure out a reward system for open source

The universal currency for exchanging units of time for units of economic productivity is money. (and not "oh, my employer gives me $5 a month to donate anywhere I want, we are so special!" but rather, $200k to $3m per year to projects that help run your company.)

They could pool results saying "these 10 scans are potentially from these 40 people" and cross reference against every airport visit, but are they that competent?

Where's this meme that Google just hires everybody who wants to work there coming from? Google has known-broken interviewing processes for hiring capable people (he says, bitterly, after never having had passed a phone screen with them).

Earlier this year, Terminal 5 at Heathrow wouldn't let me opt out of the scanner. Kinda felt like a personal violation, but the duty guards just said I was uneducated for having any concerns outside of their training.

So, unless they are using face recognition to scan you when you get your boarding pass checked then re-compare it at the rapiscan, there's no PII gathering happening.

(Now, they could always make a policy change and say "now you must tag your boarding pass on the scanner before entering" but we're not there yet.)

You can do neat things like: create a base filesystem, create a snapshot of it (i.e. "freeze" the filesystem in time), send the snapshot to another system (or over the network for S3 storage, etc), continue modifying files, create another snapshot on top of your first one, now you can send/backup the new snapshot which is only a diff of the original snapshot.

Many other systems besides ZFS support snapshots/copy-on-write layers, but they are mostly hacks and your performance will drop considerably during the "snapshot existing" time. ZFS is different because the snapshot mechanism doesn't kill your performance (unlike the 18 layer Linux file system organization).

"Sending" a snapshot can either be directly over the network to another system to ingest the snapshot (enabling you to stream a new filesystem/snapshot into existence on a remote machine), or you can gzip the snapshot stream and send it off to backup storage.

You can layer your snapshot diffs as much as you like. As long as you have all the diffs back to the original file system, you can re-create any point-in-time snapshot layer from your backup archive.

Longer details with examples are all over, but a good overview point is: https://docs.oracle.com/cd/E18752_01/html/819-5461/gbchx.htm...

Also, it's basically the only sane way to run containers in production since all the copy-on-write and snapshot sending isn't an 8 layer deep hack cake. In Solaris, containers are aware of filesystems at the deepest levels and the networking is aware of containers too. All Linux-based container fads over the past 5 years still haven't caught up to Solaris+ZFS+Zones+Crossbow from 6 to 10 years ago.