I have a hard time believing that a human driver would be as slow as this Waymo, or even slower. I drive my kid to school where it's posted 20mph and there are cameras (with plenty of warnings about the presence of said cameras) and witness a constant string of flashes from the camera nailing people for speeding through there.
Hardcoded limits are problematic because they completely lack context.
On that very same road with a 20mph limit, 40mph might be completely safe or 3mph might be extremely negligently dangerous. It all depends on what is going on in the area.
It's possible to separate out these tasks such that no single person or group has every needed piece of the puzzle.
The Carthusian monks who produce Chartreuse (a collection of herbal liqueurs popular for use in cocktails) have been producing it and protecting the secret 130 ingredient recipe for over 400 years successfully. At any given time no more than three of the monks hold the entire recipe, and yet they have a company they have formed to execute most of the production without the secret being leaked.
The designated monks coordinate production and are involved in QC, as well as developing new blends for special releases, but much production is done by paid employees who do not know the complete recipe.
I suspect though that a lot of the secret behind Chartreuse isn't just the recipe, but the actual sourcing of the ingredients.
Presumably the recipe relies on very unique and location-specific herbs to the alps. Part of the justification for limiting supply is concern for the environment and sustainability of their production. The order also had to cease production while they were evicted.
I wouldn't be surprised if some of the key ingredients weren't wild foraged or at least very unique species.
I assure you that if you rubber duck at another engineer that doesn't understand what you're doing, you will also be pummeled with information that may or may not be relevant. ;)
I don't think that's right. When you explain a technical problem to someone who isn't intimately familiar with it you're forced to think through the individual steps in quite a bit of detail. Of course that itself is an acquired skill but never mind that.
The point or rubber duck debugging then is to realize the benefit of verbally describing the problem without needing to interrupt your colleague and waste his time in order to do so. It's born of the recognition that often, midway through wasting your colleague's time, you'll trail off with an "oh ..." and exit the conversation. You've ended up figuring out the problem before ever actually receiving any feedback.
To that end an LLM works perfectly well as long as you still need to walk through a full explanation of the problem (ie minimal relevant context). An added bonus being that the LLM offers at least some of the benefits of a live person who can point out errors or alert you to new information as you go.
Basically my quibble is that to me the entire point of rubber duck debugging is "doesn't waste a real person's time" but it comes with the noticeable drawback of "plastic duck is incapable of contributing any useful insights".
> When you explain a technical problem to someone who isn't intimately familiar with it you're forced to think through the individual steps in quite a bit of detail.
The point of Rubber Ducking (or talking/praying to the Wooden Indian, to use an older phrase that is steeped in somewhat racist undertones so no longer generally used) is that it is an inanimate object that doesn't talk back. You still talk to it as if you were explaining to another person, so are forcing yourself to get your thoughts in order in a way that would make that possible, but actually talking to another person who is actively listening and actually asking questions is the next level.
I guess I can see where others are coming from (the LLM is different than a literal rubber duck) but I feel like the "can't reply" part was never more than an incidental consequence. To me the "why" of it was always that I need to solve my problem and I don't want to disturb my colleagues (or am unable to contact anyone in the first place for some reason).
So where others see "rubber ducking" as explaining to an object that is incapable of response, I've always seen it as explaining something without turning to others who are steeped in the problem. For example I would consider explaining something to a nontechnical friend to qualify as rubber ducking. The "WTF" interjections definitely make it more effective (the rubber duck consistently fails to notify me if I leave out key details).
We'll be okay eventually, when society adapts to this and becomes fully aware of the capabilities and the use cases for abuse. But, that may take some time. The parent is right to be concerned about the interim, at the very least.
That said, I am likewise looking forward to the cool things to come out of this.
>It seems delusional to discount the possibility of such blackouts in the US domestically.
I hear that, but we are so dependent on network connectivity for commerce (and entertainment) here that there would be riots from a different subset of the population if they turned that off.
You can harass brown people and murder activists here, but if you turn off the TikTok spigot, disable access to finance, or frankly fuckin' DoorDash or Uber, people are going to have a meltdown. Modern life here just grinds to a halt without data services.
I hear that, but we are so dependent on network connectivity for commerce (and entertainment) here that there would be riots from a different subset of the population if they turned that off.
You're thinking nationally. Think smaller.
It's not tremendously hard to imagine the internet being selectively shut down in a state or city.
Look at the events of the past week. Now imagine the Insurrection Act being invoked in Minnesota, and the state's internet is cut off as Governor Walz's helicopter flees to Canada to avoid being arrested.
If you can't imagine that, remember that nobody could imagine COVID lockdown, either. We've shut down the national air system twice in the last 25 years. Unimaginable in 1999. Yet, here we are.
While I believe this may be true, there are also just people that get more reward from building than from the act of writing code. That doesn't mean they hate writing code, but that the building comes first. I count myself in that camp.
If I can build better/faster with reasonably equal quality, I'll trade off the joy of programming for the joy of more building, of more high level problem solving and thinking, etc.
I've also seen the opposite: those that derive more joy from the programming and the cool engineering than from the product. And you see the opposite behavior from them, of course--such as selecting a solution that's cool and novel to build, rather than the simple, boring, but better alternative.
I often find this type of engineer rather frustrating to work with, and coincidentally, they seem to be the most anti-AI type I've encountered.
Yeah, that's pretty much what I think too. I'm much more of the latter type you mention, but I think I have the enough acumen to be practical most times.
Its always been the case that engineers come in many flavors, some more and some less business-inclined. The difference with AI imo is that it will (or already is) putting its trillion-dollar finger on the scale, such that there is less patience and space for people like me, and more for people like you.
>I'm trying to square this guy's experience with all of the homeless people who don't seem nearly so lucky. Or perhaps they are being helped and supported and I don't see it?
I get that, but I do witness a lot of compassion and help directed to homeless folks. However, even if they're regularly gifted by strangers, it's likely not enough to materially change their situation.
I would suggest that the staggering efficacy of panhandling does demonstrate how remarkably willing strangers are willing to help a rough looking homeless person on a street. And beyond that, there are a lot of invisible homeless (the ones not struggling with mental health or drug issues) that remain off the streets because people in their community will give them a few days on a couch here or there, or help fix their car, give them a place to park a trailer, etc.
In my neighborhood, there's a homeless man that lives in a camper trailer in the back yard of some neighbors. They just met him one day and offered him a stable piece of land to be and help him out as they can. He comes around asking us neighbors for lawn care work and such to earn some money, which is how I learned about the situation.
> I get that, but I do witness a lot of compassion and help directed to homeless folks. However, even if they're regularly gifted by strangers, it's likely not enough to materially change their situation.
When I've looked at the data, the majority of homeless people have been homeless less than 12 months. This means that the majority of homeless people who benefit from support will use it to get out of that situation quickly. And for the most part, if you give help it will be immediately and materially useful.
The parent said: "Of course, by then we'll have much more capable models. So if you want SOTA, you might see the jump to $10-12. But that's a different value proposition entirely: you're getting significantly more for your money, not just paying more for the same thing."
SOTA improvements have been coming from additional inference due to reasoning tokens and not just increasing model size. Their comment makes plenty of sense.
reply