jdewerd

I was hoping for at least 190% earnest.

My parents infected me with the same disease -- an intense conviction in the nobility and virtue of individual hard work -- and it led me to a terrible place, the physical sciences, where I received a harsh lesson in the nature of capitalism. It's not about creating value, it's about capturing value. In a low-leverage situation, when everyone works hard you are simply paid less. Surplus is captured through ownership and ownership is concentrated at the top. In low-leverage jobs, this system transforms hard work from a noble individual value to a collective curse.

The explosive growth of the software industry over the last few decades has led to an abnormal situation where regular employees have enough leverage to capture a substantial fraction of their surplus. Most people are not so lucky. For most people, the system looks more like:

    You load sixteen tons, what do you get?
    Another day older and deeper in debt
    Saint Peter don't you call me, 'cause I can't go
    I owe my soul to the company store

and having experienced work environments that were low leverage through no fault of the employees, I understand this, I respect this, and I did not forget it when I left for software, where we have it much easier not because we work harder but because we are in a growth industry.

I too found it easy to climb. I too enjoy the "free" money from real estate investments and a stock portfolio. But I know where it comes from. I do not pound my throne and launch spittle as I demand for those beneath me to work harder for my benefit.

You shouldn't either.

> AMD diverged their GPU architecture development into separate CDNA and RDNA lines specialized for compute and graphics respectively.

Ooooh, is that why the consumer cards don't do compute? Yikes, I was hoping that was just a bit of misguided segmentation but this sounds like a high level architecture problem, like an interstate without an on ramp. Oof.

He ran for office.

Are you saying he also went around denying climate change and screaming about vaccine microchips? I don't see that part. Am I missing something? Or did you make that up?

> The 3-step plan you just outlined

Speak for yourself.

> (1)...

Holy shit your fevered imagination came up with a bad plan. But they are your words, not mine. Your plan, not mine. You should reflect on that. There's nothing insurmountable about the fights I'm picking and that's the entire point.

> Picking fights is actually a political tactic for preserving the status quo

> If you don't like the status quo you should avoid taking adversarial stances

Oh, tell me about your plan for winning without picking a fight. I'll wait.

Fights are inherent to good political strategy, it's quite silly to suggest otherwise. Politics when people agree isn't politics, it's administration.

Strategy is about picking the correct fights or inducing your opponent to pick the wrong fights. Georgism and socialism and other causes that strike at the heart of the ownership issue are the wrong fights because the opposition is too entrenched. You will lose. If you win it will be by destroying the country, but realistically speaking, you will lose. If I wanted to be politically ineffective, I might join you in your cause, because in principle we agree, but I don't want to be ineffective.

Instead, I will continue to go after the tax and zoning laws. There is a lot of good to be done on that front. It's messy, dirty, non-utopian, and has no use for someone afraid of a fight -- but these are fights that can be won and that makes them infinitely better fights to pick.

"I spent 1 word giving lip service to side A and the rest of my post scolding side B. I am very fair and balanced."

I admit, I could be wrong. Maybe if I go through your history I'll see attention paid to the prior rounds of class warfare in proportion to their severity rather than in proportion to your opinions. But I doubt it and it's getting late.

To pick favorites, selective attention is all you need.

40 years of pressure-cooker zoning: not class warfare.

Blatant tax advantages for capital: not class warfare.

Reporting tax cheats: class warfare. Oh no! (Clutches heart dramatically.) How detramental to social cohesion! We must do something!

The system is run by the rich, for the rich. The unfair advantage is the entire point. You will run face-first into a titanic wall of money if you try to deprive capital of the various rent-seeking advantages it grants itself.

Sometimes naïvety is a superpower, so godspeed on the Georgism but I will not hold my breath and I will enjoy my wins where I can find them, thank you very much.

One pays deadweight loss, the other receives deadweight loss. If they are literal neighbors, they are not metaphorical neighbors.

I'm not keen to report my neighbor, but my landlord is not my neighbor. They could hardly be more not my neighbor if they tried.

When my rent skyrockets amid lots of vacant new construction (I'm not in CA, we're actually building around here), "gee, I'd like to help these fine fellows cheat on their taxes" is not the first thought to cross my mind.

Tesla has simulations too, of course. Thing is, the world has fractal complexity and synthetic data only captures the variation that you tell it about. Simulations for synthetic data generation are literally my job -- this is where my intuition is at its strongest.

> if you think driving in SF is easy

I'm not here to trivialize SF driving, I'm here to say that it's a small corner of a big weird world that Tesla has done a much more thorough job of exploring. Google and Waymo are used to having a scale advantage, but Tesla's fleet utterly dwarfs them.

> Waymo is expanding both scope and reliability faster than Tesla. They just announced airport terminal rides in PHX today.

That's nice, but FSD already drives almost everywhere in the USA. It can usually even figure out poorly mapped rural roads in flyover country. What's Waymo's timetable for those?

> So far, evidence suggests Waymo is expanding both scope and reliability faster than Tesla.

FSD has gone from private to "mostly works everywhere in the USA" in the last 3 years. It has taken Waymo 10 years to start escaping the hubs one city and route at a time. They don't even use freeways, last I checked. Slow and steady is a fine strategy, I wish them luck in the marathon, but it's a bit early to call this one for Waymo.

I'm used to holding back the enthusiasm in Tesla circles where we are naturally predisposed to call the race early for Tesla, so this is a fun change of pace.

Maybe, but I think that feedback and training data from supervised driving will continue to propel Tesla along the march of 9s until FSD has genuine superhuman performance. Of course, there is only one way to find out. Time will tell.

Safety-driver miles are more useful for training. Driverless miles are more useful for proving. Like I said, different paths: remove the human then expand scope vs expand scope and then remove the human. Waymo is undeniably further along path 1, Tesla is undeniably further along path 2. Which path is shorter? I'm pretty sure it's path 2, but time will tell.

> All that data collection and they can't handle the easiest of situations in San Francisco

Unlike Waymo, they haven't limited to and exclusively trained on the easiest situations in SF for the last decade. Just like Tesla hasn't even attempted true driverless miles, Waymo hasn't even attempted unmapped back-alley navigation, let alone unmapped back-alleys in the middle of nowhere. You need scope and reliability to win this race, and it seems to me that the Tesla formula for improving reliability scales better than the Waymo formula for expanding scope. Time will tell.

Of course. I feel the same way. Nobody is asking us to behave differently.

Tesla's data isn't just more volumnious than Waymo's, it's also higher quality. Macrodiversity is important. LiDAR isn't. Scene reconstruction got good in the last few years, and that's what LiDAR was competing with. Meanwhile, there's no substitute for macrodiversity.

Anyway, you asked for citations. Here they are.

Waymo:

> Waymo was the clear leader in 2022 with 2.9 million miles with safety drivers

https://www.eetimes.com/waymo-cruise-dominate-av-testing/#:~....

Tesla:

> FSD Now Drives 1 Million Miles Per Day

https://www.notateslaapp.com/news/1327/tesla-fsd-milestone-1....

> Graph showing 2M miles per day

https://digitalassets.tesla.com/tesla-contents/image/upload/...

> Tesla has been "collecting data" for years without much to show for.

Except, you know, millions of miles per day of self-driving in situations that Waymo isn't even planning to attempt for years.

    DRIVE PILOT is ready to chauffeur you under conditions that help ensure a secure ride.  Conditions include:
    Clear lane markings on approved freeways
    Moderate to heavy traffic with speeds under 40 MPH
    Daytime lighting and clear weather
    Driver visible by camera located above driver's display
    There is no construction zone present.

and from the manual:

    The distance [to the vehicle in front] is neither too small nor too large

How is this remotely comparable to FSD? Drive Pilot is less technologically ambitious than the 1995 CMU No Hands Across America vehicle, which happened 28 years ago. Polishing it enough to accept liability is cool, but there is no technology path from follow-the-leader to general purpose self-driving.

Do you think it isn't?

I'm not here to argue that the overly enthusiastic timetables weren't cringe. They definitely were. I'm just here because there seems to be consensus that FSD is doomed and it doesn't remotely seem that way to me.

Should you remove the human and then expand scope (Waymo)? Or expand the scope and then remove the human (Tesla)?

I dunno. I personally think the Tesla approach is fundamentally better because it obtains more data, faster. Millions of miles per day rather than millions of miles per year, with better macrodiversity to boot. However, that's just a heuristic: we've got a Tortoise and Hare scenario and as the parable teaches us, it's not so difficult to imagine a scenario where slow and steady wins the race. Even though I favor the Tesla approach, I could see this going either way.

That said, I do think it's uncharitable to read failure into Tesla's high level decisions. They could have chosen the Waymo path and dumped all their effort into maximizing reliability on a small scenario subset. They chose not to. There are Andrej Karpathy interviews about it. They wanted to data-max. They data-maxed. Now they have 300x the data collection rate of Waymo. That's not failure, and it may even be a killer advantage. We'll see.

ChatGPT intentionally plows through low-confidence scenarios. It just tries to do its best because "I dunno" is reliably low-utility while a hallucination doesn't gamble lives. It errs on the side of hallucination by design: if you alter the RLHF feedback, you can alter the confidence level.

In FSD, as you point out, the stakes are higher. It makes a different design decision: it slows down in low-confidence situations. Just like a person. If you watch a FSD video, you'll see that most of the human interventions happen because FSD lost confidence in a tricky scenario and halted the car even though the pathing ultimately wound up being correct. This won't solve every edge case, but neither do humans -- the bar is (or should be) "better than human," not "perfect," otherwise we are sacrificing lives on principle.

I have no idea if training will stall out half way through the march of 9s. It might, but if it does, it won't be because there is no high level strategy to control hallucinations. That already exists.

Their existing system is pretty strong evidence that it was in fact a model training problem -- do you mean to say that you think training will stall out before it gets to the point of being viable?

Why would they want to abandon FSD? I look up videos whenever these threads come out and over the last few years it seems to have made astonishing progress.

2 million miles per day of FSD driving is impressive. The resulting training data from every corner of the USA is a bonkers advantage.

Sure, a few brief points.

1: Remember that 3BP had to break the speed of light (among other things) to make the story work. A first strike is a huge gamble, you have to make sure your shot in the dark kills the monster rather than makes it angry. If you put information lag back into the equation (not to mention other intelligence gathering and military imperfections), the confidence that you need to make the game theory work starts to look incredibly steep.

2: Nearby grabby aliens are highly unlikely in any event because otherwise they would have grabbed already. The time to colonize the galaxy (millions of years) is tiny compared to likely variation in the timescale on which interstellar-capable life emerged and evolved (billions of years). If you turn the colonization of the galaxy into a 2 hour movie and separate it into periods of "no aliens," "sparse aliens," and "dense aliens," the "sparse aliens" section is one small scene lasting seconds. It would take an enormous feat of synchronization for us to be "born" into those few seconds rather than before or after (and "after" probably doesn't work because colonization stops development). "Sparse aliens" makes for great sci-fi, so we are predisposed to imagine that it is more likely than it actually is. It's like Jurassic Park: the dinosaurs don't always escape because the mathematics of chaos theory dictate that it would be impossible to put resurrected dinosaurs in a zoo, they escape because otherwise I wouldn't buy a movie ticket (and neither would you).

3: Suppose that the situation has defied the odds and there are spooky aliens out there just waiting to kill us. What's our move? Hide? Or grow aggressively? The "hide" option is already a bust, so we should grow aggressively. Everyone else will do the same. Just as the desirability of a first-strike hinges on unlikely confidence in the ability to effectively carry out complete retaliation-eliminating destruction on a rapidly advancing adversary far in the future, the desirability of hiding hinges on unlikely confidence in the ability to hide from adversaries with truly staggering surveillance options. Even if the forest is full of monsters, we do not expect them to hide.

I love the 3BP trilogy and the atmosphere of horror that it builds, I just don't think its thesis works very well IRL. Which is a good thing :)

Right, and while I'm no expert in this field I vaguely recall some proposals that consciousness / self-awareness fell out of planning and the need for a representation of your future / hypothetical self in those plans. If so, that would make it very important or a harmless side effect of something very important. I was mildly upset to see characters in Blindsight repeatedly bring up the "consciousness is pointless" thesis without any allusion to the common counterarguments, if only to dismiss them in the context of the story. But maybe that makes for better horror?

3 Body Problem did the same thing. It repeatedly presented the strengths of the Dark Forest Fermi paradox solution, but I don't think it ever presented the weaknesses, not even to lampshade. I usually try appreciate what's in a story rather than whine about what it "should have done" (which is a recipe for never enjoying anything) but in this case I've seen it negatively affect people who didn't already have the "antidote" background knowledge. I've seen this in online communities (you'd think there would always be at least one smartass in each sci-fi thread but I guess not) and I've seen it in person with a coworker, so I wonder if it isn't a bit of an ethics issue.

Ah well, on the scale of ethics issues this one is pretty minor. Most people are stable enough to handle spooky alien stories, and maybe the double take is part of the fun.

Right, but the engineering output exists... in the google graveyard.

Yeah. Back when machine architectures were churning so fast that compiler flexibility allowed portability that otherwise wouldn't have happened, UB freedom was a net good. Now that those arbitrary architecture choices have stabilized, the good is no more, and the evil of optimization rules lawyering has replaced it. UB needs to go.

If Big Tobacco is full of spooky masterminds, why were we able to virtually eliminate smoking in the USA?

Yeah, it's a shame they've been audience captured. At the beginning they leaned a bit to the rosy side, clearly glossing over visible negatives. Somewhere around when they left and were able to speak about good and bad but hadn't committed to a narrative was probably the point of peak value. Now they are almost comically anti-china. Ah well.

It was never worth it to retool for decades?

Or was there a pathological lock-in between suppliers and manufacturers that prevented even the most obvious innovations from happening if they required any amount of coordination?

12 Volts is 12 Watts per Amp.

48 Volts is 48 Watts per Amp.