A look under our trunk: what's in our compute | Hacker News Reader

A look under our trunk: what's in our compute

waymo.com

139 pointsra7110 comments

Apologies for repeating myself, but it's hard to convey how far ahead Waymo is in every area: sensors, vehicles, training data, markets served, simulation, infrastructure, operations, regulatory.

They're growing fast, but I do wonder what they see as the bottlenecks: municipal approval, training, manufacturing? It's really unfortunate that lobbying is slowing self-driving rollout in some of the biggest and most important cities: NYC, Chicago, DC.

Is it unfortunate, or just a different priority in a different place?

Those are also cities with a great public option for transit (unlike, say, Dallas or the Bay Area where different municipalities squabble). One might laud New York driving people away from car culture when the subways/buses can serve more people, greener.

Perhaps, but I can't think of why Waymo should be prohibited but not Uber or taxis.

Taxis and barely-legal app taxis both require a human to drive. Someone who's earning a living doing this, and paying taxes to NYC. Waymo is driven by a robot, owned by a company out in California, who has a lot of lawyers doing their best to avoid as much financial and legal liability as possible. Not that Uber and its ilk isn't doing the same thing but they've still got a tax-paying citizen in the loop.

The current mayor of NYC is a DemSoc so he might have some ideas about keeping NYC's money in the pockets of middle-class NYC citizens.

I don't think all of the economic benefit stays in NYC even when you take a yellow cab. A significant share of NYC taxi medallions are owned by entities based outside the city. Marblegate Capital in Connecticut owns about 2,146 medallions—nearly 16% of all NYC taxi medallions. They make more profit off the medallion financing business than actually operating the taxis which is basically break-even AFAIK.

I wouldn't be surprised if that's somewhere on Zorhan's to-do list :)

Yeah but the plan of "ban them because the taxi drivers union said so" is ludicrous. Why not have a partial integration with re-training programs for existing taxi drivers, and work with the city council to establish laws that effectively strip Waymo of its profits for operating in the city, making them only barely break even, and siphoning the rest of the funds into the aforementioned retraining programs and transport infrastructure more broadly. Even a 3% profit margin in one of the largest markets in the world is better than none at all, Waymo is effectively forced to comply with any measures NYC imposes. This doesn't have to be the case nationally either: many areas are relatively underserved by even Uber and Lyft, and there is plenty of room for Waymo in those markets, and I imagine those "empty" parts of the US, in total, eclipse the size of NYC. There, it will be nothing but a boon. But in established cities like NYC, it is almost certainly best practice to keep as much profit in the local economy as possible, since Waymo is not, in the end, drastically multiplying economic productivity in those cities.

Then again, the areas underserved by Uber/Lyft are underserved because they're bad neighborhoods so drivers won't go there. How many Waymos need to be torched or their passengers carjacked before Waymo decides to stop going there? Fun fact, if you throw a traffic cone at the Waymo, it panics and stops moving, meaning your can now break the rear windows and assault the passengers for their valuables. If you wear a mask and leave your phone at home, there's no recourse for the passengers.

I was thinking of rural areas, not dangerous neighbourhoods, which are already served by uber/lyft

Allow Waymo and leave the choice up to residents. If they want to pay a driver they can.

Nobody is keeping money in anybody's pockets. What you are advocating for is a tax on NYC people to force them to use human drivers instead of new technology. Just like the toll booth operators of the 90s.

It is so disappointing to see the tech class advocate for laws that keep jobs that are unhealthy and lead to nowhere (cabbies) instead of moving the country forward (and those people getting other, hopefully better, jobs). But again, that type of short-sighted, zero-sum/socialist thinking is becoming more common, sadly.

Buses and subways are not necessarily "greener".

30 people in 30 waymos vs 30 people in an electric bus?

No contest.

No contest when the bus is full, but the overall efficiency (/"greeness") of the system is much less clear considering: buses are rarely full, most buses are diesel, heavy buses tear up the pavement on their routes (especially near stops), and buses don't take riders directly to/from their destinations.

Most of the city buses in my UK home town are now battery electric double deckers. These carry up to 85-90 people (with 25 standing), weigh 11 tonnes, and is 10.5m long.

A typical jaguar Waymo usually carries up to 4 people (usually just one), weighs 2.1 tonnes and is 5.2m long.

So the double decker bus should be more environmentally efficient with maybe 5 passengers making independent journeys, though operational costs for paying a driver probably make the Waymo more financially efficient in that situation.

I suspect robotaxis will become very important in suburban areas with distributed jobs where buses don’t work well. In busy cities there simply isn’t the roadspace to replace buses with robotaxis (or any other form of car) for commuting.

> A typical jaguar Waymo usually carries up to 4 people (usually just one), weighs 2.1 tonnes and is 5.2m long.

A more reasonable taxi for single riders is going to weigh around 1.5 tons (e.g. the prototype Tesla Robocab).

You also incorrectly assume that the efficiency depends linearly on weight. It doesn't, it depends more on aerodynamics, speed, and the number of stops. London buses (11 tons) average around 2.2 kWh/mi, Tesla Model 3 is at 0.21 kWh/mi while weighing 1.8 tons.

Road damage scales hypercubically with axle weight, so an 11 ton 2-axle bus does ~1400 times more damage to the road than a 1.8 ton Tesla.

https://en.wikipedia.org/wiki/4th_power_law

This is the right way to think about it. Larger buses make sense when there are a lot of people going to and from the same places at the same time, but I think that is much more rare in my city after remote work became widespread.

The real question is what's the optimal size for a bus if you don't have to pay a human to drive it? Once the technology is proven, there's no reason you can't put the Waymo or other self-driving technology in a bus. A bus designed for a dozen people instead of 30, and a fleet of them that did point to point rides instead of a fixed route, just all booked via the app so you share rides with other people. I dare say that would be the best of both worlds.

> The real question is what's the optimal size for a bus if you don't have to pay a human to drive it?

It will depend on the area. Where I am in San Francisco we have articulated buses designed for 100+ people running every few minutes above BART. (We probably need more BART stations.) There’s simply not enough road space for smaller buses (let alone cars) to carry commuter traffic downtown.

(I’ve never worked downtown though so always drove to work in desolate Silicon Valley office parks.)

Even in my smaller UK home city (250k people) you need double decker 70+ capacity buses running every 15 minutes on two routes from city centre to university to handle demand because the jobs are not so distributed.

Different 30 people in the US. I’d drop my car if Waymo was around and financially made sense. You’ll never catch me wasting my time on a bus though. We’ll never even be close to Europe or elsewhere wrt where public transit was even decades ago in those places.

People here only ride the bus when there’s no other option for them.

Waymo could make a bus too anyway, especially after they’ve been in an area a while and have data about where they’d have the most impact. Internet at large would probably decry it as enshittification.

Bus was just fine in Portland, Seattle, and San Francisco.

> Bus was just fine in Portland

Yes, if you live within about 1 mile of downtown and also work there. If you live downtown and your job is in Beaverton, you're likely in for a 90 minute commute. Or if you live in Beaverton and your office is downtown. Portland's mass transit is adequate for a subset of people and completely impractical for most.

Source: I've lived in and around Portland for 50 years. Worked downtown much of it, and have lots of friends and former colleagues who used public transit. You have to really want to not have a car because the time commitment is real.

Okay, but clearly there’s a difference between, “the bus makes sense if you live or work near downtown,” and, “people only ride the bus when there is no other option,” as the GP said.

Yeah theres a couple spots that it’s fine for people that have never experienced what’s actually possible.

That 90 minute commute by bus isn't one where you're driving though. A strict time comparison ignore the fact that the bus or train ride merely requires you to get out when you reach your destination. You don't have to sit in stop and go traffic driving. Sitting in traffic driving sucks unless you have comma.ai or equivalent. Mentally it's just exhausting by the time you get home/to work. If you have to drive and can shitpost HN for those 90 minutes, vs trying to do that while driving (don't do that), then who cares if it takes a bit longer? There's obviously too much longer, but still.

Bus is fine in select areas at current fares, but would be less popular if people realize each drive costs more like $16 in total instead of the $2 fare they pay.

> Waymo could make a bus too anyway

Oh hell no. Cars are bad enough from the damage they can cause by kinetic energy, but a bus? Especially a bus where people don't wear seatbelts or, worse, are standing?

It would probably be safer than most bus drivers currently, so I don't see the issue. The discomfort around damage being caused by statistical anomalies vs individual error is not a rational one if the former is, in the aggregate, smaller than the latter. Individuals have stories, but societies operate more closely in line with the laws of physics: probabilistic at the small scale, deterministic more broadly. From that view, we are just replacing one statistical possibility with another, one which will result in less people dying; or, more importantly, less likelihood you or your loved ones will die because of a bus accident.

> don't wear seatbelts

I just want to point out that the safety record of public transit drivers is typically impeccable. There is no safer vehicle out on the roads except buses and trains. I have been a rider of public transit for 50 years, minus about a decade when I owned a car, and in 3 different major cities, and literally I've never been aboard a bus that hit another vehicle, hit any obstacle, or had a reportable accident. Literally never happened. (I was aboard a light rail train that collided with an SUV, but that's a whole other ballgame.)

I've never felt unsafe without a seat belt on board a bus. Seat belts in cars may be a bit of a panacea, but I challenge you to find any metropolitan area where the buses have accidents, much less injury accidents, on a weekly basis.

Well, you'd need to crash for that too be an issue. We have to wait for the data to prove it out, but I much prefer a Waymo driver than a human one, especially if I'm walking home right after the bars close.

An average bus load in the US is around 18 people. So if you just compute the raw bus emissions, then an electric bus is maybe about 2x more efficient than personal EVs.

But wait, there's so much more!

A single bus/subway ride costs about $20, which is a surprisingly consistent figure for various US cities. In NYC it's $17, for example. I'm talking about the _true_ cost, as in: "yearly operational budget divided by the number of trips". And most of this cost comes from 3-4 drivers that every bus needs, not from the hardware itself.

Price is a pretty reliable proxy for CO2 emissions for general market goods and services. And with these kinds of numbers, personal cars are competitive. Even if we factor in subsidized road maintenance cost ( https://taxfoundation.org/data/all/state/state-infrastructur... ).

If we now add mild carpooling, so that a single robotaxi drives 2-3 people, then classic transit just dies. It's like comparing horse-drawn carriages with steam trains.

Most of the cost comes from paying the drivers, but also cost is a proxy for emmissions?

Yes? Why is this strange?

Drivers emit a lot of CO2, just like every person in the US. A driverless car allows these emissions to be reallocated to other more productive areas.

You're counting emissions from a person's breathing as part of the vehicle's emissions?

People take their salaries and buy food, clothes, heating and air conditioning, cars and fuel, and plane tickets, and furniture. All of those things generate emissions in their production and/or transportation. This is why dollars is such a good proxy for co2 emissions.

Surely a person driving 18 people will emit less than each person driving themselves? I can see how dollars can be a proxy for macroeconomic stuff, for example by comparing countries, but it doesn't make much sense for comparing individual units of transport.

I don't understand why we have to do napkin maths to guess what's more polluting. There is countless research showing buses emit less than individual cars per passenger per km, self-driving or not.

And there's no reason buses can't be self-driving as well, if anything, them driving slower, on dedicated lanes and on fixed routes would make it much easier.

> Surely a person driving 18 people will emit less than each person driving themselves?

No. A person driving themselves does not create any _incremental_ CO2 emissions. They don't need a dedicated full-time position to do the driving.

> There is countless research showing buses emit less than individual cars per passenger per km, self-driving or not.

Care to share it? I actually read through most of highly cited articles in this area, and it's not at all so clear-cut. For example: https://ourworldindata.org/travel-carbon-footprint

A single bus journey is $20? That's insanely expensive.

Maybe it's different where you live, but 95% of the buses I see driving around here are completely empty.

Are you near the city center, or out near the suburbs? It's common for a bus to be crowded on one part of the route and nearly empty at the end.

And this is a fundamental issue with transit. It _has_ to travel as mostly empty on avereage to be useful.

...no? In the heart of downtown, with people riding in both directions, I don't see how it follows that it has to mostly empty to be useful. If it's full you can't get on so that's not useful, but anything less than totally full should still be useful.

Its not a meaningful issue. The passenger density is higher on even a mostly empty bus than cars, and the pollution is lower as well.

I just took a Waymo in San Jose, having taken a few from San Francisco - it dropped me off in a place that would have made sense in SF (across the street from my destination at a light, avoiding a U turn) but was terrible in the suburban streetscape where the road was literally 10 lanes 4 / 4 / 2 turning and no one would ever drop someone off like that instead of U turning so they didn't have to cross scary giant traffic.

So weird little things for scaling, but overall it was a good ride till that point. Also, the dropoff was at a hospital/clinic and Lyft for the return trip had no issue with going into the property to the actual destination.

That seems like something not hard to fix, I am surprised they haven't already addressed it. Uber, as a comparison, absolutely narrows down the drop-off & pickup to the side of the road you want. I would demand the same of Waymo.

Waymo also lets you pick, but only from a predetermined list of pickup/dropoff locations. Sometimes there's not a spot on the side of the street you'd want, other times it will default to a spot on the other side of the street but you can override.

Your market may vary, but I can assure you that in Phoenix, you can drop the pin practically anywhere that's safe and legal.

The app has changed the method a few times. It used to show green or blue segments for valid spots, but today as it is, I just pick up the pin, drop it, hope for the best. It also demonstrates which way the car will be facing.

Dropping a precise pin on the map is no guarantee that it will actually stop right there, depending on the situation on the ground when it arrives. But, you can still adjust the pin while the car waits, or you can hit the "Pull Forward" button to try and improve it.

The app has also been updated to disallow adjustment of multiple dropoffs beyond the initial stop. So if you want a precise pin-drop in a multi-stop trip, you need to intervene along the way, instead of during your initial setup.

There was a recent case reported in the news, where people were trying to Waymo to the San Bruno airport, but their pin got dropped off in a residential neighborhood. This was apparently because Waymo doesn't directly serve the airport's lanes yet, so the pin dropped as close as possible, but the residents of that street were nonplussed about the results. So were the airline passengers who were lugging luggage!

https://sfstandard.com/2025/09/16/waymo-receives-permit-oper...

Waymo does have a permit too go to SFO now, so it's only a matter of time.

Fwiw the accessibility options have a setting to avoid across the street dropoffs

I'd settle for "on the same block" pick up/drop off. For some reason, Waymo has decided my street is a busy street in the middle of the night when it isn't, so even if I override it to stop down the street where there's plenty of space for traffic to go around the Waymo, it only gives me a short 2 minutes or else it'll drive off. Which would be less of a problem if it didn't decide to also change the location of the pickup last minute as well. It's enough of a problem that I've switched back to Uber/Lyft if I'm not feeling like dealing with it.

FWIW they started actively testing in Chicago a few months ago. I frequently see them around town now

Yes! I've seen cars as far north as Evanston and as far west as Oak Park.

I’m glad you said it because as a 40 something nerd roboticist taking a Waymo is nothing short of a miracle to me.

I mean drafting code on my phone with chatgpt in a Waymo is legitimately some Jestons shit and I’m not hearing otherwise.

Amazing

And that's not even pointing out that also makes Waymo the most advanced robotics company.

So are they? I know that tesla is behind, but they have so many simulation miles, several generations of their own hardware, and lots of people behind the wheel taking over.

I've also seen the autonomous waymo vans, which seem nice too.

Is there a "theoetical" ahead and a "practical" ahead?

How do the finances look and does financial solvency count? will someone run out of money first?

There have been two waymo articles (about sensors, and now about hardware), is this a sign they are worried?

I am actually wondering.

> Is there a "theoetical" ahead and a "practical" ahead?

Yes. Waymo is doing something like 400K fully autonomous rides per week and Tesla is doing hundreds or low thousands of supervised rides.

Tesla has 1.5m subscribers to FSD, maybe that's millions of drives per day (or two)

> it's hard to convey how far ahead Waymo is in every area: sensors, vehicles, training data, markets served, simulation, infrastructure, operations, regulatory.

In the last two years there have been dozens of incidents of Waymos obstructing first responders in life-threatening situations. Every major metropolitan area Waymo has deployed in, the police departments have reported extensive issues with Waymos obstructing officers and emergency vehicles, and worse, their remote support staff often being unable to do anything about it.

If that's "so far ahead" of everyone else, then we are pretty fucked.

> it's hard to convey how far ahead Waymo is

What are you comparing it to though? Tesla Robotaxi service is fine, there's a safety driver that isn't really supposed to talk, and I'd say that Zoox is better than Waymo, except getting rides on that is very limited so I have less experience with it. I like the Zooclx vehicles way more than even the new Waymo Ohai vehicles. I think there's a fourth competitor in this space currently active but I can't remember their name off the top of my head. Cruise is defunct though they were mostly fine as well.

The real quandary is for the cars to freak out if something happens and just stop and sit there, while a human driver is going to drive off if you're getting eg carjacked.

They really are the impressive tech, it's really a shame that they aren't the one capturing all the media spotlight.

I don't know if we'll get there eventually, but if we do, I won't be surprised if it's because of Waymo.

Riding in one is pretty mind-blowing at first. But what's more shocking is how fast you get used to it. After a while it's just a boring car ride.

I wish more people would experience it and see that the future is now.

It's beyond boring after a bit.

It reminds me of putting on noise-cancelling headphones in an already quiet room. You suddenly gain an appreciation for just how much quieter things could've been thanks to the contrast.

Similarly, having a vehicle accelerate/decelerate consistently, always taking turns with the same sharpness, etc.. all of it builds to this subtle feeling of the car ride fading into the background of the experience. You didn't realize those were features of a drive you could appreciate until experienced.

>it's really a shame that they aren't the one capturing all the media spotlight

This does seem at least a bit intentional on Waymo’s part. They appear to have adopted an under promise, over deliver mindset (unlike some other robotaxi alternatives that come to mind). They’ve seen a lot of their competitors (Uber, Cruise, etc) capture a lot of ultimately negative attention and folded as a result; I would probably try to avoid the limelight if I were them too. The tech speaks for itself.

Every Waymo on the street is already an advertisement, especially since they're so distinctive with their LIDAR units and now they're driving Chinese cars. Every time a Waymo is remotely associated with anything bad, the media jumps on them anyway, like the times they've "hit" a pedestrian but literally after a human driver hit that same pedestrian.

Tesla Robotaxis need to hog the media spotlight because they're trying to prop up the stock price, not because they work.

How many times have you ridden a Tesla Robotaxi? They work fine, in my limited but actual real world experience, the problem is they don't like you giving rides to other people, vs Waymo you can call them for other people and they're fine with that. Tesla Robotaxis hog the media spotlight because it's Tesla and every tiny thing they do gets blown up because in the attention economy, a Tesla mention is a giant shiny distraction to go off and everyone piles on their opinion of Elon because blah blah blah lidar Elon stupid no he's not and media sites get all that rage engagement and that juicy ad revenue. Who outside of the industry can name Waymo's co-CEOs off the top of their head? I can't.

It's been 16 years since Google announced they were working on driverless cars. Waymo's been offering service to the public for six years. There's been plenty of news coverage and people already know about them. Being in the spotlight all the time just wears people out.

The more I think about it the more autonomous cars look like extremely demanding edge computers. You have limited power and cooling, unreliable connectivity and very strict latency requirements. It's probably one of the few places where designing the hardware around the workload makes more sense than trying to make a general purpose platform do everything

A self-driving car is an "AI robotics" platform.

Both self-driving car and general purpose AI robots are emerging technologies right now. My bet is that eventually, the HW will converge accordingly.

>high-fidelity data from 13 high-resolution cameras

is it me or their example photo doesn't look even like 1 high-resolution camera (and very noisy at that, looks like a pretty high ISO setting). Their example looks similar and i'd say a bit worse than what my Fuji camera did 25 years ago in the dark after i'd adjust dynamic range of the image in GIMP. (and the cameras on the Waymo cars i see around have about the same lens size as my Fuji)

If it is really a showcase of their visual pipeline than it makes sense that they are still so reliant on lidar - which is pretty expensive because of the number of rays you have to support to maintain good situation-awareness and which is still hard and expensive to scale to get good resolution. That also means that their tech can't really be put onto other platforms like drones for example.

Absolutely not an expert on this, but I would imagine that they would train the pipeline on a raw-ish image data to have better consistency? I would think that some kind of algorithm "bettering" the image suddenly might cause a discontinuation in the pipeline that is harder to account for versus the ML machinery being trained to do this job as well.

Also, what looks good to us may not be "good" to a machine.

But I'm really just guessing in the dark here.

>they would train the pipeline on a raw-ish image data to have better consistency? I would think that some kind of algorithm "bettering" the image

of course, whatever "bettering" algorithm makes sense, the neural network would probably emerge it inside itself as a result of training. My comment was more about hardware in the sense that "garbage in - garbage out", and so you'd want to have a decent hardware which is available and pretty cheap today while Waymo seems to be using, i'm stating it politely :), very-very-commodity level.

>Also, what looks good to us may not be "good" to a machine.

the kernels of the first layers of visual CNNs (at least the ones i looked at) converge to the Gabor-like kernels what are in the first layers of the human visual cortex. Bunch of years ago, before deep-learning CNN, etc. i saw an article (in neurophysiology field) proving an optimality of such kernels for those visual tasks. Of course things may and probably diverge a lot in the deeper layers, yet deep dreaming, etc. suggests to me that the nature of the divergence may be less severe than expected.

That night image looks very heavily cropped based on the field of view.

Very cool. To be honest, I'd assumed they were using commodity hardware, but no real reason for that. Cool chip. It is quite interesting that so many large companies have their own chips for these things. I almost managed to get my hands on a couple of accidentally liquidated MTIA chips but I was too slow. I wonder what the chip is like and what the dev experience is writing for it.

It would just be cool to see what these guys have cooked up dev tools wise since Google has always been famed for good developer experience (often to such a degree that Xooglers are useless outside the exosuit).

Downside of the MTIA chips is no open backend - so quite a lift to get them working if possible at all :(

REing Meta's custom AI HW to get it running with something would be one hell of a project, and one hell of a flex.

It would rely on them being nice enough to leave at least some binaries on the thing though. Because I'm not sure how would one even get an in on how to boot the things without.

It really doesn't say much. Tesla had more detailed talks on their HW and AI stack. This reads like a glorified marketing blurb.

That's exactly what this is, a marketing blurb. Was this unclear?

Recruiting

I've been wanting to start the hashtag #gohomewaymoURdrunk for when it gives me stupid routes like circling the block pointlessly or when dropoff is stupidly far from my destination. (firstworldproblems except for when I sprained my ankle). Anyone else on board?

Odd. I didn’t see any photos in the trunk or the compute bays, did anyone else?

Would love to get them in nyc

I thought that was stalled because the mayor didn't like it?[0] Any new developments?

> New York abandoned its proposal to allow self-driving taxis earlier this year. New York City’s new mayor, Zohran Mamdani, has said he would heavily weigh the interests of taxi drivers in deciding on rules for the technology.

0: https://www.nytimes.com/2026/06/17/technology/waymo-driverle...

Contact you mayor

I'm very open to being wrong about this, but will they actually work in NYC? Waymo requires some grace from other drivers and pedestrians in order to function and I just can't imagine New Yorkers will offer it.

Do you think they will?

I've been cut off by a Waymo before, so the real question is if they Waymo backend is designed is such a way that the NYC Waymo driver can run with different aggression settings from the one in Phoenix Arizona and Tokyo.

> Redundant: Because there is no human to take over, proactive safety and redundancy are natively built in. Our compute is designed like two independent engines. While they normally operate as one unit running full parallel workloads, if one experiences a fault, the other seamlessly takes over.

uhhh i've been in a situation where a remote human took over before

it's rare but it can happen (or at least it could 1.5 years ago-ish)

the specific situation was a tight alleyway with cars parked on either side and people doing things like opening car doors and crossing randomly. the waymo obviously wanted a much wider safety margin than what was available and stopped to wait for things to clear up. the remote human driver instead drove more aggressively towards the pedestrians so they would clear the road.

They mean no human at the wheel to take over if the hardware dies. If you are driving full speed down the highway and your gpu packs up, you’d better have a backup to bring the vehicle to a safe stop.

Waymo has zero "remote human drivers". It is literally impossible to "drive" a Waymo by remote commands.

The Waymo agents are constantly monitoring the fleet for exceptions. When an exception is raised, a human reviews the footage and the situation. They can send commands to the car and suggestions about which way to branch or a course of action to choose. So ultimately, it is the Waymo Driver doing the actual driving.

If a Waymo gets really jammed up, they can indeed send in a human driver, who will appear in person, disengage the Waymo Driver automation, and drive the car normally. This has happened to me only once in 3,000 miles. The Waymo Driver, in my experience, is pretty good about keeping passengers informed when it's a bit delayed or confused by a complex situation. And the CSR will call in if they need to.

But no, it is a misconception that people have: Waymos cannot be "driven remotely" by any means.

Interesting that the article uses "ML" everywhere instead of "AI". I'm sure some marketing person chose that term deliberately, and I wonder what went into that decision.

All of the industry leaders of AI can’t stop bragging about how their AI is committing what would be felonies if a human did it, warning that it’s about as likely as not that AI will literally lead to human extinction, and threatening that it will erase all white collar work (or not depending on the day) and investing resources in a manner that will only be worthwhile if it actually does.

I don’t think there’s any mystery as to why AI is not a good marketing term.

I find the juxtaposition of "can't stop bragging" and "is not a good marketing term" to be fascinating. Have you considered that they know these recent reports aren't a good look for them but that they feel compelled (either because they know it would be far worse when it inevitably got out on it's own or maybe even from some moral sense) to report somewhat truthfully?

Two different audiences. Unfortunately for them, messages cross streams. I think it is a good message if you’re trying to scare people into giving you money (or else they lose out on unlimited profits down the road).

Who is this audience that thinks unasked-for and un-controllable felonies is a good message, or who wants to fund that?

Everyone who is buying into RSI/AGI hype. That hype is the only thing sustaining this investment cycle. The bubble will pop as soon as people realize “AI” is a normal technology.

I had a whole long response typed out and I realized that it probably wasn't worth it. I think that I just have to admit that I find your view and understanding of the world completely incomprehensible. Thanks for humoring me, and have a nice day.

I don't think there's anyone who wants to fund "committing felonies", for obvious reasons.

Apple did exactly the same thing a few years back:

Apple avoids “AI” hype at WWDC keynote by baking ML into products (2023)

https://arstechnica.com/information-technology/2023/06/at-ap...

First of all, machine learning is a distinct category of A.I., so they are being more specific about their methods. If we get right down to it, "A.I." is, and has always been, a nearly meaningless buzzword that means whatever you want it to mean. In 2026 it is highly stigmatized by haters, anyway. So I appreciate that a techblog endeavors to explain exactly what sort of A.I. they're employing in these systems.

they now are doing a 5nm asic, what did they have before this?

Waymo has never disclosed this information.

like it.

Compute is under the trunk ??? Noted. At the price of RAM nowadays it's like the new copper !

I hope such persons will be met with a spray of indelible paint.

Custom silicon while it still looks like they've got multiple gaming PC's worth of compute gear mass isn't a flex when you remember what Tesla has with their FSD computer comparably just sipping power, while processing signals that are less straightforward as unscalable LiDAR

Reply on news.ycombinator.com