Hacker Newsnew | past | comments | ask | show | jobs | submit | more throwawayjava's commentslogin

Well, Google already piggybacked off of DARPA ;-)

Also, combat zones != California suburbs.


> Also, combat zones != California suburbs.

We're just going to leave that one tee'd up, aren't we?


I have to agree with this line from the report you linked: "However, it is a reasonable assumption that adversaries will eventually devise targeting techniques, especially in the littoral areas adjacent to their homelands"

If China wanted to, they could develop the capability to track aircraft carriers -- especially within striking distance of the mainland.

But there's a much larger implicit assumption threaded throughout the report -- the assumption that our aircraft carriers will be cruising at full speed and on high alert in prepared and competent battle groups at time of attack. The assumptions of a maximally defensive posture and extreme competence are present in virtually every argument advanced by the report.

The most significant US naval defeat -- in which damn near the entire Pacific fleet was wiped out -- suggests that a full threat assessment should relax most or all of those assumptions.

Particularly given the article under discussion.


Nah it's the nukes and the missiles.

We fought a big ass war with carriers. They're not the deterrent. Nuclear Armageddon keeps the relative peace


I wonder about the cost of licensing patents for a project like this one.


The exact same thing that happens now: someone dies, usually selected arbitrarily based upon the driver's selfish desire to minimize harm to themselves. Because that situation doesn't allow enough time to make anything except instinctual choices.

The driver is found at fault for that death if they were not driving legally. Otherwise, and very often times regardless, the whole thing is written off as a tragic accident (millions happen every year, just in the USA), and the insurance people do their thing.

Ethical dilemmas are a non-problem for self-driving cars.


Millions???

You are off by two orders of magnitude. It was 35,000 two years ago.

Still a lot of people, but not millions.

Put in context, it's around one death per 100 million miles driven.


Sorry, I mixed up the global and national numbers.

Thanks.


And no autopilot would intentionally ram a plane into the side of a mountain.

These edge-cases are uninteresting IMO.


"These edge-cases are uninteresting IMO."

On the contrary, these edge cases are by far the most interesting thing. Not just tabloid-interesting, but in terms of generating usable data.


On the contrary. The edge cases are really all that count.


We shouldn't demand that our systems work in 100% of cases. Nine Nines -- or even less, in the case of autos -- is enough.

You can argue that autonomous flight doesn't yet have Nine Nines. I would probably agree. But "Miracle on the Hudson" was just that, and extremely rare situations should not be a standard-bearer for allowing autonomous flight. In the same way that suicidal/homicidal humans shouldn't be the standard-bearer for allowing human flight.


And still, the pilot portion of the cost of operation for planes is in the single digits. You barely scrape into the double digits (maybe 20% MAX, closer to 10 I think) if you count all labor.


http://www.cnn.com/travel/article/how-airlines-make-less-tha...

> "With a net profit margin of just 2.4%, airlines only retain $5.42 per passenger carried," said Tony Tyler CEO of International Air Transport Association (IATA) at the group's 70th AGM in Doha, Qatar.

I suspect they're interested in single-digit cost reductions.


Airlines will kill to get another 10% in margin.


So are flight attendants impossible to automate/do without? Or are there perhaps other forces -- beyond technical ones -- keeping humans in the plane?


AFAIK, the FAA requires a certain number of flight attendants for safety reasons.


Right, that's my point.


The economics couldn't be more different.

Driver salary is 26% of the cost of trucking.

Driver salary is 52% of the cost of a taxi.

Those numbers are in the single digits for aircraft and trains.

We could replace pilots and train operators, technologically speaking. But the incentives are not (yet) large enough to overcome the regulatory and cultural barriers. We'd rather have that extra order of magnitude of safety, because it costs so little compared to other expenses.

The cost/benefit calculation for drivers is fundamentally different.


>Driver salary is 52% of the cost of a taxi.

If you drop the utilization, the cost of the driver as a percentage probably goes up a bit. But here's the thing about those numbers.

Yes, driver salary is a big chunk of taxi cost but flip things around. It's only about half the cost. In other words, there's this widespread assumption that autonomy totally changes the nature of automobiles and automobile ownership even though it only cuts the cost in half.

Ask yourself. If you own a car today, would getting a 50% discount on taxi rides completely change your behavior. Maybe it would for some at the margins but only at the margins.


> Ask yourself. If you own a car today, would getting a 50% discount on taxi rides completely change your behavior. Maybe it would for some at the margins but only at the margins.

Look longer term.

Parking spaces in cities cost ~30k. (And in some cities, a lot more than that). Included with a house, they are still part of the price (no free lunch).

Over a 30 year mortgage, that is 1k a year.

In California, car insurance is almost 2k a year. (Oddly enough, 1.1k in New York State!)

So 3k a year just to have the privilege of driving a car.

Apparently an average new car now days cost 31k. (Which seems insane to me, but that is what people are buying).

Being generous and rounding down a bit, that is another 5k a year.

So now 8k a year, before gas, before maintenance, before any warranty problems, before screws in tires, etc.

If mass transit can replace commuting to work, then weekend outings with automated taxis become very doable. If you give $5 per ride (on par with UberPool, though I realize that isn't profitable), it is 1600 rides a year to break even with owning a car. (Though ownership goes down after the car is paid off, 600 rides a year to break even!)


1) Now the taxi can operate day and night without substantial breaks. That reduces per-passenger-mile amortization costs by at least a factor of 2.

2) Especially since amortization cost is also reduced, you might as well electrify the vehicle (this requires some sort of robot arm for fast charging, but I suppose you could pay someone like $0.50 per cycle to plug and unplug a bunch of vehicles in the interim). That reduces fuel costs by a factor of 2 and maintenance costs even more.

3) No driver means room for more people or ability to shrink the car, reducing all the above costs.

So you've now reduced all the other costs by a factor of 2. That means your overall costs are now reduced by 75% or more. Yeah, that's starting to look pretty attractive.


1) It's already very common for taxis to operate two shifts per day with two drivers. Also, there's less demand in the middle of the night than there is at peak hours.


Sure. But even if we all own our own cars, level 5 autonomy is still a game-changer.

IMO the killer app is in freight anyways.


Getting a 50% discount on taxi rides would make me never consider any other option when I need a taxi.


It depends on the circumstances. On the other hand, for me, if taxi rides were half the price I doubt it would affect my ridership at all over the course of a year. MAYBE for one or two marginal car rentals.


> Driver salary is 52% of the cost of a taxi.

On top of that, the space and interfaces for the driver is a substantial part of the cost and weight of the vehicle itself.


1. Most westerners born outside the USA. And in the USA anyone willing to take out a loan.

2. The core requirements for a cs major at a typical university (read: not CMU,MIT,Stanford,Berkeley) and especially at a non flagship is basically 4 or 5 courses, which you can for sure cram into 6 months if that's all you're studying.

Personally, I'd rather work with people who learn some part of that on their own (or even not at all) but have a stellar education in reading, writing, and math. The former are often provided by good high schools. The latter not so much. At least in the states.

But if all you want is just the core fragment of just the cs, six intense months are enough.


Having a 5 year degree in Informatics Engineering, I doubt very much you could cram that much in 6 months.

Not at my university, with the amount of stuff we had to do, unless in the 20+ years that have gone by, they got the quality level down.


> I doubt very much you could cram that much in 6 months.

No, you cannot cram an entire degree into six months. Hence the second paragraph under my second point.

But the core CS sequence? 4-5 courses is a rough semester, but it is a semester!

And if you already have some mathematics or especially some programming background, the sequencing that causes those courses to stretch into 1.5 years is already resolved.


Well, 4-5 courses was the minimum we had per semester, with enough deliveries to keep us rather busy, starting with 3rd year.

Some people managed to jungle a few more courses.

I don't see such a program being much more than simple introduction to CS.


Yes, that's exactly my point. You can get a simple introduction to CS in six months, especially if you already know how to program.

No, you're not going to be publishing papers or talking through OS implementation details with CMU grads or designing computer vision systems with Berkeley grads. But you'll have the baseline knowledge that agumonkey is referring to in their comment.


You still might accidentally invent the command pattern on your own.


> The core requirements for a cs major at a typical university (read: not CMU,MIT,Stanford,Berkeley) and especially at a non flagship is basically 4 or 5 courses

I went to a school that is very far away from CMU, MIT, Stanford, Berkeley, etc and my core requirements were A LOT more than 4 or 5 courses. I'd have to check by I can name at least 8 core requirements off the top of my head.

> which you can for sure cram into 6 months if that's all you're studying

So your suggestion is for people to look at a CS syllabus and find a way to teach themselves all of the core courses?

I have my bachelors in CS. I never really programmed until college. 6 months into my BS I had barely wet my feet with Scheme, learned a little bit of discrete math, and had a VERY basic introduction to formal logic. I knew exactly one programming language.


> I'd have to check by I can name at least 8 core requirements off the top of my head.

I'm familiar with a vast array of CS curricula at many types of institutions. I have no doubt that this is true, but it's also not universal.

To be clear, I'm referring to the courses that everyone from that institution will have taken before graduating. So almost all CS majors require much more than 4 courses, but only 4 courses worth of certain, common, shared knowledge.

If you only count courses offered by the CS department, and that everyone must take to receive a major, I think 4-5 is fairly typical. Maybe 6-7. Especially at places that aren't ABET.

Most programs will also require 3-4 additional courses for the major (that the college/university does not require for graduation). E.g. at least one or two mathematics courses, sometimes a Physics course.

And of course all those programs require some electives -- sometimes selected from subject area "buckets".

But my point is that it's certainly possible for two students to get a CS degree from the same (decent or even good) institution and only share 4 "core" CS courses. Typically an intro course or two, a lower-division algorithms-style course, something architecture-y, and something systems-y.

> So your suggestion is for people to look at a CS syllabus and find a way to teach themselves all of the core courses?

Maybe.

This approach would certainly be inappropriate for a student with no background in Mathematics or programming.

But a strong programmer can breeze through or even skip those intro courses.

And a student with a strong mathematics background -- e.g., an accountant, other science major, or finance person learning to code -- can breeze through that otherwise time-intensive algorithms course.


> The core requirements for a cs major at a typical university (read: not CMU,MIT,Stanford,Berkeley) and especially at a non flagship is basically 4 or 5 courses, which you can for sure cram into 6 months if that's all you're studying.

Huh? That's an incredibly terrible CS curriculum. 4-5 courses is at most 20 credit hours. A typical undergrad degree requires about 120 credit hours of courses before you can graduate. I went to a second tier state school and our CS curriculum had 12 common required courses (excluding seminar courses) that everyone took, plus three upper level electives of your choice. Basically half of the four year degree was CS.


Please see my other comments.

I'm talking about the set of courses that are 1) offered by the CS department and that 2) every student must take.

Not even the entire set of cs courses. Just the ones every student must take.

> A typical undergrad degree requires about 120 credit hours of courses before you can graduate*

Obviously you cannot get an entire degree in 6 months. You can't even finish all of the required coursework.

That's why I said "core CS sequence", not "a degree".


> I'm talking about the set of courses that are 1) offered by the CS department and that 2) every student must take.

As I said, that was a dozen courses at my school.


12 CS courses, offered by the CS dept, and that everyone takes means that either A) your department was internalizing a LOT of service courses; B) left very little room for differentiation within your major; or C) provided very little time for exploration outside of your major.

Either way, that level of straight-jacketing and/or in-sourcing is atypical.

Even at top schools that number is usually below 10. E.g. CMU 6 (well 7, but really 6); Berkeley 4-6; MIT 8; Standford 6.

These are all off the high end of my 4-5 estimate, but only because they ALL in-source a course that at most places would be farmed out the math dept and therefore not counted by my criteria. And MIT also a comm course. Plus, there's the other extreme of places that only require the intro sequence and an algorithms course, leaving all else to the students.

In order to get up to 12 at any of these places you'd need to only offer one course for a lot of the "breadth" requirements, or else in-source the calculus sequence (or similar silly things).

Again, I'm counting courses offered by the CS dept, AND that everyone takes. The minimum viable "common knowledge" we can reliably assume is shared among all CS majors, as offered by the CS dept. IME 4-5 is pretty typical. Maybe 6.

Yes, we also assume some background in other fields of study from CS graduates (some calculus, reading/writing skills, etc.) But those courses are not the ones I was referring to with my 4-5 count. They were also addressed in my original comment.


> is atypical

I don't think it's very atypical, for example take a look at this course requirement for City College of New York under CUNY :

https://www.ccny.cuny.edu/sites/default/files/csc-prereqs-04...

Under the CS parts, only the elective in category A, B, and C is where student can pick and choose. Everything else under the grey curve across the pdf are:

> courses offered by the CS dept, AND that everyone takes.


> Requiring OCaml in the base operating system

I'm not sure I understand this concern in the majority use cases. Why would you need an OCaml compiler installed?


The base system is usually capable of compiling itself. If the source for your base system is stored on a remote git repository, and your implementation of git in your base system is written in OCaml, you're going to have to use an OCaml compiler to recompile it.


But you don't need source control to compile the base. You only need it to retrieve the base sources - and even then only when you want to get the latest, since the installer comes with a tarball of /usr/src.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: