AND, much less visibly, has been bleeding top engineers that choose to leave like someone cut it's heart open. There's even an internal joke about it (a specific way to refer to one's salary)
> But for things that are relevant to civilians, like the reactors, engines, etc.., I am sure that what's in subs is relatively simple, and probably dated.
It seems like submarine propeller designs are all classified past 1960, even though quiet and efficient propellers pretty relevant to civilian ship design:
The thing about military stuff is that generally the budget is large and the goal is to design something better than what the enemy has. The civilian world for a long time wasn't willing to blow hundreds of thousands of dollars on ASICs to control phased-array radars; the military was. Now as a result of lots of military investment, the technology is so well-understood that Google put a phased array on a chip inside the front of the Pixel 5.
> In addition, military technology is supposed to work on the battlefield, you don't want delicate stuff there, you want rugged, repairable, proven, reliable.
What you want is stuff that wins fights, and it only needs to be repairable and reliable insofar as it wins fights. The US has the F-22, which is an ultra expensive jet that only has ~60% uptime. In war games, it achieves kill ratios of 100:1, so the military is more than happy to keep it around. When the US raided Osama bin Laden's compound they sent brand new stealth helicopters even though they knew the platform was less reliable.
I asked the ASML rep at KubeCon who they were there for as nobody I knew had a quarter billion to spare. He told me they were just interested customers, but if I didn't want the newest machine I could get one for 70 million euros.
Right but it still gets put into boxes and flown over by a 747. Of course it‘s more complicated than that, but most contenders for complex machines are much more built in-place, and not a complete 'product' being assembled.
When it comes to controlling the wheels to prevent sliding and slipping, the AV control system is unbeatable. The ABS and traction control on a regular car has to cope with whatever control inputs the driver has made; on an AV, the computer models the grip limits of the wheel and plans a trajectory to not exceed them. It's not just for snow but also for changing pavement surfaces and the rain.
The main limitation is still sensors in the snow, but it seems to not be that big of a deal to build sensor packages that are better at seeing in the snow than a human is.
Being able to plot a series of inputs that can more efficiently use available traction than a human doesn't prevent you from blundering your way into a dumb situation where the laws of physics dictate that the only possible outcomes are various flavors of bad ones.
It's not clear how often the software will chose poorly and need to brute force its way out with traction/handling. The fact that they seem to be hedging against this by putting the hardware on particularly performant cars indicates it must happen enough to matter or be rare but bad enough to matter when it does happen.
Waymo will probably also rack up a ton of technically not at fault accidents by being obtuse in traffic since there's when there's snow there's a lot less margin for the "two people trying to pass each other in a hallway" type missteps that behavior tends to create.
They put the hardware on performant cars because it would be stupid to choose gas for stop-and-go city driving. Electric cars are fast; a new Honda Civic does 0-60 in 6 seconds!
No, this is not a "works in a textbook" take. The path planner is aware of surface conditions on the road. This is already a big deal because otherwise the AV would not be safe to operate in the rain.
Waymo will not be touched in most of those accidents. I've been driving in snow long enough to know that there are always going to be idiots who drive too fast for conditions and lose control of their car; I'm sure some of them will blame Waymo because it's nearby. I once watched a guy with a TX license plate spin out on a perfectly empty, perfectly straight freeway. Waymo doesn't really need to do anything for human drivers to crash.
> The chat interface is all that there is and their behavior in chat is not deterministic or bounded enough to be useful in most applications.
Their behavior in chat is not deterministic, it's stochastic. That is the point - the usefulness of LLMs comes from their ability to deal with the vagaries of language.
> But because the LLM cannot be trusted to behave properly around the topic, they have to filter anything which touches it.
IMO this is because giving a random person a frontier LLM is like giving them a Ferrari. Most people would manage to not crash it. A few would experiment with it and learn how to drive it very well. A few more would immediately assume that a fast car means they can drive it fast and end up wrapped around a telephone pole.
We get lots of mileage out of other stochastic systems. I've worked on a lot of projects that did, and the defining trait that made them successful doesn't seem to have a name but the closest I can come up with is "boosting." In ML (esp. in classical ML), boosting is when you train one classifier to predict the residual error of another. The first classifier minimizes some entropy loss, and then the second contributes additional bits.
In a system with a human-in-the-loop, it often takes a lot of engineering to allow the human to boost the output of a system. I once worked for a company where we had to very precisely label maps based on real-world data. We had a model that could produce a sometimes-accurate polygon, but obviously just asking a person to adjust the polygon after the model generated it was terrible because that was a vague ask that took a lot of time and effort to do. Instead, we gave users a brush tool and trained a new model to fix the polygon based on that. A simpler example was a system for reviewing user reports: we tuned our system to approve them with high precision and used a human review queue for the rest. Reducing the number of bits of entropy a human being had to contribute to a decision in the average case allowed us to smoothly iterate on the model while staying flexible.
The AI companies that actually going to deliver useful products will be the ones that engineer interfaces that quickly allow human beings to refine LLM outputs. It's going to be a long time before any of these models can reliably one-shot a complex task with ambiguous parameters. Chat is only one possible way to do this, and frankly it's not a very good one. I think that this is the point the article was trying to make, minus the corpspeak and hype.
I don't sit down and decide "here's how much I'm going to pay the plumber." I call a plumber and get a quote. I can keep going until it feels worth it.
My enlightenment comes from talking to a bunch of plumbers and hearing what they're gonna charge. That's the market rate.
If I hire an employee and they get good at their job and someone else offers them a 40% raise, they'll probably leave. That's a lot of money.
Some companies care about retention, most don't really.
> and the employer can hire a new senior engineer at below market rates to accommodate the specific learning they have to do
It sounds like what you're saying is actually that the last engineer was being paid above-market, because the price that employers are paying new employees is literally the market rate, seeing as it's the rate in the literal market.
> the price that employers are paying new employees is literally the market rate, seeing as it's the rate in the literal market.
I want to drill this into anyone that throws the word "below market" or "above market" around.
If a company pays below-market, it won't be able to hire anyone. Either the role will remain unfilled, or the employer will have to compromise on experience.
If someone is claiming to be paid below-market but the company can hire their replacement, then they're not being truthful.
Hiring the replacement ≠ finding the same or better replacement.
Will you be able to find someone in this economy? Sure
Could you fill the shoes? Much harder. Especially in the age of bootcamps, AI to complete exercises, and what have you where people went into IT for the money and not for the love of the craft, tinkering and learning.
So this is not an oxymoron: you can have 200 applicants and not a single good one to replace someone with them.
Only in a market with perfect information. With imperfect information, the market rate is an estimate of the expected or typical rate for a similar good. Because everyone has access to a different subset of the information, everyone's estimate is different, and companies often end up paying above or below the consensus rate.
That depends on the country. In the US where everything is tied to stock options and employees will jump ship if they get 50 cents an hour more next door, that may be the case. In Aus/NZ you don't have that kill-yourself-for-your-stock-options-and-then-leave culture, if you're not on a barely-subsistence wage where you don't have any choice then people will look at job satisfaction, ease of travel to/from work, and so on alongside what they get paid. I know several guys who have turned down jobs that paid five-figure amounts (this may be a two-figure amount in the US) more than they were currently earning because they were quite happy with their current work environment. Good work environment, laid-back management, the company looked after you (rec room, pool table, beer fridge, after-work gaming, being able to tell the mgt that something wouldn't work and they'd listen to you, etc), not because they saw it as a leash but because they believed in looking after their employees. Some of those guys, and their coworkers, have been at the same company for 20-30 years.
In AUS/NZ you're not going to earn even close to the same amount as you would in the US. Every time I've changed jobs, it was for an extra 100k+ per year. The rec room/pool table can't compete with that.
My current manager has been with the company for almost 20 years. He doesn't feel the need to leave because his stock options have made him a multimillionaire. The company doesn't care about him or me, but we're both happy to stay because we don't need them to. We just need them to pay us.
No, we are not talking about a commodity market with a clear exchange.
If an employee is bad at a negotiation or doesn't look around then they may accept a lower salary than another market participant would have offered and if neither side looks around, and is willing to pay some costs of a change then they are making a salary that is different than the current market rate.
I'm not an economist but that implies the market maintains some kind of optimal equilibrium price. The reality probably is very noisy like with everything else. Plus there's asymmetric information on both sides meaning people don't get what they think they do.
> If a company pays below-market, it won't be able to hire anyone. Either the role will remain unfilled, or the employer will have to compromise on experience.
This is pretty much what's occurring in New Zealand right now, yes. 2020–2023 had pretty much zero international movement due to closed borders with COVID-19, with a low official cash rate which caused business to be in desperate need of development resource; so salaries were high.
Market rate for developers has either stagnated generally or depending on the role dropped as hundreds of applicants are willing to undercut each other on what constitutes an acceptable pay check.
But most employers don't go around reducing previously-hired people's salaries for a variety of reasons.
The main reason is that employees will quit. It's not illegal in the US, but it still almost never happens. Companies would rather lay off than risk alienating employees like that.
Being around a Waymo makes me feel WAY safer than being around a human driver. If more cars were replaced, I would probably bike even more.
Seriously, Waymos follow at a respectful distance and overtake me safely. They stop at stop signs. Sometimes they even stop and wait for me to make a decision about which way I'm heading.
No, it's the human driver vs the Waymo. I'm not going to entertain fantasies where all the cars magically disappear from the road; there's no political will for that and no politician is dumb enough to try.
To be perfectly clear, the difference between an empty road and a Waymo is mostly academic if you're on a bike. The Waymo is just that good at respecting space.
I’m not talking about cars magically disappear from the road, I’m specifically talking about taxis and rideshares versus using another way to get around, which even includes personal car ownership.
Studies have shown repeatedly that rideshares lower transit usage and don’t reduce personal car ownership.
Improving transit and allowing Google to blow $100B on making cars drive themselves are not mutually exclusive.
In the US, in particular, last mile transportation is mostly done by car. Transit cannot economically serve low-density suburbs.
In a city? It depends on price versus convenience. As more and more cars drive themselves, the city can get away with taking back more and more space for transit and cyclists. People should choose mass transit because it's convenient, not because there's no way to call a car.
But the studies say that Google blowing $100B on self driving cars changes human behavior to take less transit which then costs taxpayers money because transit ridership and revenues decrease.
This idea that companies have to be allowed to do business however they want is not something we just have to do.
We can as citizens and governments say “no, you’re not allowed to run a robot taxi company because it’s overall bad for our city.”
Laissez Faire capitalism is an ideological cancer to American society.
...I don't know why you think that a majority of drivers respect bike lanes. They don't. Nearly every driver has parked a car in a bike lane at some point. At least in the US it's uncontroversial - the bike lanes tend to be so poorly designed and thought out that it's much easier for cars to use them than cyclists.
I think majority respect them, because I drive bike in a city that is not exactly friendly to that. The situation is, some people break into them regularly, most dont.
reply