beisner

I think there’s a reasonable middle ground, which is: employers can discriminate based on record only if that crime has a direct relationship to the nature of the employment and the position provides a clear opportunity for abuse in the same fashion as the original crime. Ie people who commit fraud can’t work in financial services, people who hurt children can’t work in education/childcare, people convicted of corruption can’t become cops, etc. But there should be a process, over time, to cleanse even these records (although perhaps in practice there would be carve-outs for particularly heinous offenses...). This doesn’t seem particularly limiting, from my perspective, but protects against the extremes you outlined.

I agree with this generally. One thing to consider is how, specifically in the case of homeless with addiction issues, how the distribution of larger amounts of money affects violent and non-violent crime within the homeless population. Especially in relation to profoundly addictive/dangerous illegal drugs. Seems like there’s pretty large opportunity for exploitation by drug dealers and opportunists, if you know the most vulnerable in your community who are already excluded from the legal system have received an influx of cash. Perhaps this only works with other changes to policy in tandem.

I don’t know enough about the issue to predict if this potential risk is significant, or outweighs the substantial benefits.

FWIW when driving, tolls along that route are roughly $10 each way... and that's low for the midwest.

The United States is structurally incapable of large-scale public works projects. There are a lot of reasons for this and they aren’t particularly partisan, rather reflect how decisions are made. My guess is places like China and Singapore will have the first true deployments of self-driving, because they’re actually capable of building road infrastructure to support it when it isn’t fully level 5.

I use similar language in all my salary negotiations. Never had pushback, and every time I’ve gotten massive comp bumps as a result (up to ~50% increase, which happened twice).

Eh, from a noise level there’s no difference with having an in-person meeting at a cafe, which happens all the time and is not bothersome (depending on the cafe).

Well that doesn’t mean much, the White House could have signaled to ByteDance on the DL that they’d torpedo any non-Oracle deal... who knows, banning a single company from operating domestically without any immediate proven national security threat is unprecedented, the president just does whatever he feels like.

When it becomes Julia.

This is true enough in the US, but is prohibitively expensive in a ton of markets worldwide (maybe half the world population). Then again, the requirement to have a Mac to develop iOS apps is even more financially onerous...

“All of Statistics” by Larry Wasserman is a very good - if concise - introduction to statistics. Doesn’t require too much background, although the problems can get pretty hairy. Unfortunately, many of the proofs in the field are based on “tricks” that accumulate, which makes the learning curve steeper for people who don’t have as much background in finding/using these tricks.

Tech alone CAN be a solution to large social problems, but typically not directly. Instead, it has to make some socially-“good” effect the inevitable consequence of making something else cheaper/more efficient.

Renewable tech (ie solar, wind) is an excellent example of this phenomenon, where we are finally getting to the point where you don’t need to appeal to people’s morals/values to get their energy companies to use renewables - they’re now simply cheaper in a lot of regions! Because tech improved in the right direction, greed and economic efficiency have become much more aligned with reduced environmental impact.

So the way I like to think about problems is this: what technology, if it existed, would solve this problem even when people/societies act entirely in their own self-interest (given how many world economies typically function)?

Morality is a luxury few feel they can afford, so make the moral choice cheaper.

Er, maybe "the rich are getting richer and the poor are getting slightly less poor" is more accurate.

While I agree that designing and training a model the size of a human brain from scratch may well be an undertaking too great for any one human or even group of humans, that’s not how humans are created!

I’d say the fact that we start life as a single cell with - at maximum - only a couple of gigabytes of information packed in our DNA and yet turn out to have profound intelligence is strong evidence that we might be able to design a system that - through some combination of obscenely efficient simple-ish learning algorithms - might be able to bootstrap its own intelligence in a similar fashion to how humans do.

This isn’t necessarily a call for research to study how human embryos turn into thinking, feeling people, but more a loose upper bound on the complexity of the initial size of a model that could become AGI.

There are 100 million Instagram users in Illinois’ jurisdiction?

Past performance is no guarantee of future results.

Any evidence that grit isn’t also a heritable trait?

There’s a famous saying in the sciences: “All models are wrong, some are useful.” Many of our physics models have been “wrong” historically, but yet they describe the world well-enough to be useful in making very good predictions. Newtonian physics is a perfect example of this - it tells an incomplete story of reality, but it’s still extremely useful in making predictions about how the world works.

The reproducibility crisis across the social sciences (and sometimes even in biology and medicine and - hot take - machine learning) is evidence that many of these models are both wrong AND useless for prediction, as opposed to just being incomplete.

Unfortunately, when it comes to public policy (at the national level at least), the way our system is structured doesn’t leave room for anything except for two distinct positions.

We know how to make fast robots, we know how to make safe robots, and we know how to make useful robots. But we do not know how to make fast, safe, useful robots. A faster robot would need to be heavier to do the same work.

Any robot that can move as fast as a human and perform similar work to a human will need to be close to as strong/heavy as a human. And humans can kill each other with their bare hands; we just know not to. A Kuka LBR iiwa can only lift 7-14kg, but could easily accidentally kill/maim someone with common payloads of that mass.

Obviously there's also some slowness w.r.t. the actual control here (i.e. safety pauses, planning latency, etc.), but it's a tall order to build any kind of home robot - especially one that has such a large workspace - that is intelligent enough to be both fast and safe.

Very cool. Good touch sensors are one of the biggest bottlenecks in manipulation today - nearly everything commercially available is terrible and extremely expensive.

I’ve only watched the video here, but I wonder how separable cable manipulation is from the actual insertion task. At some point, there’s a transition between the connector and the flexible cable that, for many kinds of connectors, requires a transition in manipulation strategy. Think about how you have to adjust your grip from the wire to the jack when you plug in a Cat6 cable.

I bring this up because the headphone jack example there isn’t really demonstrative of this difficult modality transition dynamic.

Great work nevertheless!

Git is a version control system, whereas Gitlab/Github are source-code hosting applications. You use git to interact with Gitlab/Github.

Without going too in-depth, they allow teams to set up accounts on a central website, where team members can contribute to that central git repository. There are also other project-management features baked into the tool (code review, issue tracking, etc).

Of course, this can all be done using a bare git server running on some publicly-available computer, but they offer nice UIs and very friendly onboarding so that makes them attractive for teams.

Excellent for deploying small auxiliary apps that are not necessarily your company’s core competency, or a side project.

Most FAANG companies. Definitely Google and Facebook.

I’m not sure that’s true. The people who have agency to move don’t need to; Highly-skilled knowledge workers are already being showered in benefits and compensation in the USA. This is definitely true in tech, and increasingly true in conventionally “high-status” professions (ie finance, consulting, strategy, etc).

A lot of people talk about the 1%, but there’s a separate, massive quality of life divide (especially at work) between the top 10% and bottom 90% of workers. Working conditions have never been better for the top 10%, and are getting worse for the rest.

So while there are definitely other immediate negative consequences to the US not having these benefits unilaterally, the fact that they already exist for the richest/most-educated makes me skeptical we’ll see an exodus of that demographic.

This rings true in my experience. When I worked at a FAANG company that needed non-trivial collaboration between offices that were 3 time zones away, I would fly to the other office once every 2-3 months and just spend a week with the other team.

Part of it this was to develop a healthy rapport, to build mutual trust while working on large-scale projects.

But another aspect was productivity; both my team and the other teams had a lot on their plate, and it could be difficult to align incentives such that our shared projects would be completed in a reasonable amount of time. Flying out, finding a desk near the other team, making yourself visible/available, and working directly on the project would get more done in a week than would have happened in 2-3 months of remote collaboration.

Of course, it would be lovely if all companies could get work done remotely (yay Gitlab, Basecamp, etc. etc.), but most companies aren't set up for remote collaboration success, especially across management lines.

I wonder what fraction of consumption people who’ve been lost their income during the crisis is. If 20% of the workforce is laid off, does that mean consumption will go down 20%? Is the salary/spending power of the group who were laid off representative of the entire workforce? Or does it skew in a particular direction.

I ask because one explanation of the disconnect is that consumer spending in general hasn’t actually dropped that much, it’s just moved away from in-person small businesses.

Buzzfeed News has some very high quality journalism. Separate division from their entertainment division, I wouldn’t dismiss this out of hand.

I think we’ve known for a while that the stock market is not the economy. My two cents:

1) Uncertainty is anathema to markets. At the beginning of the crisis, there was a ton of uncertainty about how the disease would progress, what it’s impact on supply chains would be globally, and how governments would respond. In the last few weeks, the US has gotten a lot more information on how things progress, “flatten the curve strategies” are appearing to work (which reduces uncertainty), supply chains are still very strong and, for staple goods/services, only minority affected, and the central banks of the world are issuing a MASSIVE amount of short term liquidity to prevent credit crunches.

2) (this is a bit more cynical). I suspect the productivity of a huge number of Americans, particularly in service roles, has disproportionately less impact on the economic productivity of publicly traded companies. In other words, many jobs are disposable without impacting the financial health of companies in the stock market. Small businesses are being gutted, and so is a lot of big retail, but it’s been too short of a time for those effects to make a dent in how many of the publicly traded businesses are operating.

I don’t know, is it that impractical for a group of conspirators to meet in person a single time, exchange a bunch of 1TB hard drives with identical one-time pads, and then conduct their business accordingly? Surely that’s enough for a substantial amount of communication. Impractical for large-scale communities of illicit behavior but definitely practical for small groups of people