adinisom
More generally, Comma's safety is handled by a microcontroller. The microcontroller limits steering torque commanded to a safe range for level 2 driver assistance and acts as a watchdog for the rest of the stack. Failure of openpilot causes it to disengage and sound an audible alarm.
From an openness perspective this is excellent. Technically it seems challenging with a DSP designed to detect a single thing using as little power as possible. Currently this balances doing as little work as possible to detect plausible utterances of the wake-word on the DSP while minimizing the costs of spurious wake-ups on the CPU. At the very least multiple wake-words seems to need the DSP to do more work and wake up the CPU more often.
Despite improving production (more tools! more assistants!) and bonus tractors (work harder! heroic labor brigade!) it's not enough to make up for phase 2's punishing 1000 tractors in 60 days, or roughly 16 tractors a day. Even if we give ourselves a full 120 days, that's 8 tractors a day and we seemingly cannot accumulate tools and assistants early enough to do that.
Hacker News faces the same risk.
Respectfully I prefer a world where Hacker News exists.
Why stop with image boards? After eliminating all photo uploads, awful people still produced and shared illegal content. So then we eliminated unlicensed camera use. 'Course the real problem still happened.
Hacker News is made possible by Section 230... should it not exist?
But the problem with deputizing intermediates is that it's too effective. It creates incentives to over-police and we have less rights against corporate policing than we do government policing. We would not have the internet we have today without the user generated content and moderation that section 230 enabled.
- secure a string to the old cable
- pull the cable out the other end, pulling the string through
- secure the string to the new cable
- pull the string out the other end, pulling the new cable into position
They contrasted this with their prior technique which is more of an image classifier that can identify general location from image features. This approach does not require their database to contain a picture of the exact location.
- Start with a list of basis functions and your signal.
- Go through the list and find the basis function that best correlates with the signal. This gives you a basis function and a coefficient.
- Subtract out the basis function (scaled by the coefficient) from your signal, and then repeat with this new residual signal.
The Fourier transform is similar using sine wave basis functions.
The key that makes this work in situations where the Nyquist theorem says we don't have a high enough sampling rate is ensuring our sampling (possibly random) is un-correlated with the basis functions and our basis functions are good approximations for the signal. That lowers the likelihood that our basis functions correlating well with our samples is by chance and raises likelihood it correlates well with the actual signal.
Often people who lose their hearing want to be able to hear in social situations such as restaurants and family gatherings. In this context, the signal and noise have similar properties and are coming from the same direction. Directionality helps but can only do so much. Noise reduction can make hearing aids more comfortable to wear but don't necessarily improve comprehension in challenging situations. Progress here is fantastic -- at the same time it helps to have realistic expectations.
Putting the mic on the person speaking sidesteps the problem -- it's like the rest of the room isn't there.
The traditional solution is an FM system where you give the person speaking a microphone linked to your hearing aids. There are dedicated ones like Phonak Roger. You could probably also use your phone as a microphone if it's bluetooth connected to your headphones or hearing aids.
Dan is a moderator on a forum and his goal is to maintain a level of civil discourse rather than an aggressive style of communication. It's a very specific definition of "violence" for a specific context and perhaps there's room for clearer terminology.
For parsing specifically, there's literature on error recovery to try to make progress past the error.
1) Embedded systems typically do not allow data to grow without bound. If they were going to keep debugging data, they'd have to limit it to the last N instances or so. In this case N=0. It seems like the goal here was to send troubleshooting data, not keep it around.
2) Persisting the data may expose the driver to additional risks. Beyond the immediate risks, someone could grab the module from the junkyard and extract the data. I can appreciate devices that take steps to prevent sensitive data from falling into the hand of third parties.
Agreed re: making the state a static variable inside the function. Great for simple coroutines. I made it a pointer in the example for two reasons:
- Demonstrates access to the state variables with very little visual noise... "s->"
- For sub-coroutines that can be called from multiple places such as "delay" you make the state variable the first argument. The caller's state contains the sub-coroutine's state and the caller passes it to the sub-coroutine. The top level coroutine's state ends up becoming "the stack" allocated at compile-time.
typedef struct blinky_state {
size_t pc;
uint64_t timer;
... variables that need to live across YIELDs ...
} blinky_state_t;
blinky_state_t blinky_state;
#define YIELD() s->pc = __LINE__; return; case __LINE__:;
void blinky(void) {
blinky_state_t *s = &blinky_state;
uint64_t now = get_ticks();
switch(s->pc) {
while(true) {
turn_on_LED();
s->timer = now;
while( now - s->timer < 1000 ) { YIELD(); }
turn_off_LED();
s->timer = now;
while( now - s->timer < 1000 ) { YIELD(); }
}
}
}
#undef YIELD
Your company is probably using hardware containing code I've written like this.
What's especially nice that I miss in other languages with async/await is ability to mix declarative and procedural code. Code you write before the switch(s->pc) statement gets run on every call to the function. Can put code you want to be declarative, like updating "now" in the code above, or if I have streaming code it's a great place to copy data.
It's interesting how polarizing the comparison of human and machine learning can be.
At the same time shaping perception is key. Authorities may push a narrative that affords them greater control. If that's not the world you want, the first step is not to legitimize it.
There is also some noise from the Pi Pico power supply which could be a good thing if you're willing to average or filter over a large number of samples.
More details here: https://pico-adc.markomo.me/
- Sometimes the creator bases their captions on the script and misses changes in edit
- Sometimes the creator's captions are perfect transcriptions but broken up and timed awkwardly
Auto-generated captions aren't always perfect but unlike human captions provide word-by-word timing.
Certain vintage hardware had a "turbo" button to unleash the full speed of the newer CPUs. The designers blind to the horrors of induced demand.