I saw my first self-driving car parked at Google in roughly 2008 or 2009, and watched them toodling around campus shortly afterward. Many people (like Larry) thought it was a 5 year project.

There is this famous story about research into image recognition back in 1960s - everyone involved thought it was literally a matter of few months, maybe a year at most, to have a system that has 100% accurate recognition capabilities.

Yet we're in 2022 and the best of the best image recognition systems spew out absolute nonsense no matter how much money is poured into them.

It's just the same sort of issue. It can work amazingly well 90% of the time, but the remaining 10% is basically impossible without some kind of actual general AI that can understand the context of what it's looking at. I honestly believe that current approaches to this tech will not get us to general self driving tech that is safe enough for use on our roads.

What do you mean spout absolute non sense? The progress in image recognition has been insane in the past decade so your comment might have been accurate 10 years ago. If there's one thing DL/ML are very good at it's computer vision

Sorry I thought it was clear from the rest of my comment. Image recognition is amazing in 90% of the cases, and then in 10% it's worse than a blind human. Another famous example of that is the(now solved) Google image recognition saying with 99% certainty that a sofa in a zebra print is in fact, a zebra(4 legs, zebra pattern = must be a zebra). Look at any demonstrations from Tesla at what comes out of their clasiffier as the car drives around, sure, most of it is correct but some of it is insane, the car thinking there's a ship or a cow in the middle of the road when there's nothing there. Sure the car doesn't act on it, but the fact that the recognition gets it so catastrophically wrong is indicative of the quality of the whole thing.

Another story in a similar vein. There are many, many stories about how the advent of DNA sequencing would finally solve the issues of creating the tree of life for all species on Earth, and solve the problem of defining species boundaries, this was some 20 years ago or more. We're still nowhere close thanks largely to the unimaginable quantity of biodiversity out there. We have the low-hanging fruit genomes (specimens readily available, known to be commercially important, etc), but that's about it. Even with those "genomes" (most are not curated to any real level of completion) we still haven't solved compute at scale, even small analyses (10k terminals, 2k genes) take months if not years to comprehensively run (i.e. explore parameter spaces), albeit thanks in part to limits on resources available to scientists.

The Earth is big and complex, I can't see the singularity on the horizon yet, but my vision is aging.

I solved the computational problem for genomics already, since it's a mostly-embarassingly-parallel problem (in paticular, I made an idle cycle harvester at Google that applied 1+M xeon cores to science problems, one of which was genomic assembly).

I don't know that creating super-accurate species trees is really the most important problem to solve in biology, though. Certainly DNA sequencing for health has been a really mixed bag.

Nice work with that solution! I did similar! I solved the earth's hunger issues, funny thing is we produce enough food, we just can't deploy it to the people who need it. Problem is I can't figure out the logistics of how to make it work... I think maybe you have the same problem? We should compare notes.

lol

> Certainly DNA sequencing for health has been a really mixed bag.

Ah, the old why bother since it's a "mixed bag" argument. If it wasn't a mixed bag I'd be really worried, as someone has some pretty good snake oil.

I vaguely recall something about sequencing and its role in creating a vaciine that turned out to be pretty useful, something recent maybe?

I don't have a complaint about sequencing that leads to vaccines. That's what sequencing is for.

What i'm troubled by is the projects that sequence a million people, using $1B to do so, and then just dump a bunch of genomes with some hand-wavy claims about how it's going to cure cancer. The costs here are rathre significant, but the health outcomes are not.

Another example. An airplane broke the sound barrier in 1948. Today, we are as far removed from Yeager breaking the sound barrier as Yeager was from a civil war hero being President (Grant). But we still have no supersonic airplanes in commercial service.

tbf, lack of supersonic airplanes is a commercial problem rather than a technical one.

On the other hand, most of the commercial jets we do fly are basically 1950s designs with bigger engines and some computers in the cockpit, and HN thinks it was crazily irresponsible for Boeing to have decided to use sensor-triggered autonomous systems to do something as simple as temporarily adjust the trim system...

It was irresponsible to use one sensor to adjust the trim in a not-very-visible way.

Likewise, a lot of computer scientists didn't take Engelbart's 1969 "mother of all demos" seriously because they thought general artificial intelligence was right around the corner. Why bother with a mouse and bitmapped display when you can just tell the computer what you want?

the mark i perceptron did image recognition with a single layer neural network and was surprisingly good at it. Some days I think Rosenblatt was on to something, but about 40 years ahead of the time.

I think your response is exactly why people are skeptical now. Why do I think they suddenly have a better grasp on an estimate that's a decade out?

That was wrong, clearly. But it was a prediction made with far less evidence than we have today. Considering progress so far (which we had far less of in 2012) I don't think ten years is wrong today.

And even if it is wrong, it's obviously not stupid to think we'll have driverless cars in 10 years. Waymo are literally down to solving edge cases. There are more than people might have expected, but there isn't anything fundamental to solve.

> Waymo are literally down to solving edge cases.

But: there are very large numbers of these edge cases and they are of the kind where new ones keep popping up all the time. This is a very easy problem for the first 99% or so and then the last 1% is super hard.

Sure - but they've been working on the last 1% for years now. Nuclear Power plants are almost always over budget and over time, big misses. And they've built lots of them. Every piece of software ever built suffers from this issue to some degree. Elon Musk misses every deadline he ever sets. Something taking longer than initially thought shouldn't instantly set off alarm bells suggesting it won't ever get done.

> Sure - but they've been working on the last 1% for years now.

Indeed, and we aren't any closer compared to say 5 years ago. That is more or less my point: you can solve for some of this and then you are still left with a mountain more. And then at some point there is an even harder problem to deal with: how to ensure that fixing one problem won't regress one or more others.

> Indeed, and we aren't any closer compared to say 5 years ago

That's just flat wrong.

I'm sorry, I evaluate tech for a living and I don't see us being closer in a way that moves the needle. Objectively, however long it will take to get to self driving we are 5 years closer, but that doesn't really matter. What matters is whether or not the state of technology is at a level where it will enable self driving to the point where a car will no longer have a steering wheel. Because less than that isn't going to cut it.

And that may well require another leap of capabilities on the AI front, it might even require GAI. But you are entirely welcome to your own opinion, as I am to mine, but after watching the self driving industry since the 80's my timeframe is something like 25 years before fusion will arrive.

There may be some intermediary 'artificial artificial intelligence' solution where when stuck a remote driver can take over. But that has its own set of problems and I don't want to set this up as a strawman.

There are cars on the road now, driving without drivers. Google (and Cruise) have built cars without pedals or a steering wheel. The needle has moved.

Ok, let's see one drive through Amsterdam tomorrow morning at 9 am between two random zipcodes and house numbers. I'll bet you it can't do it. "Edge cases" indeed.

I'd take that bet at even money all day long. But, maybe it can't. It doesn't change the fact that they are materially closer than they were 5 years ago.San Francisco streets aren't a snack. There are hilly, narrow streets with cable cars, constant construction, homeless people wandering onto the street.

It's real progress. Anyone around Mountain View, or SF will have seen the progress with their own eyes. It's not a theoretical debate.

> But, maybe it can't.

But that's the whole point: if this were a reality there wouldn't be a maybe there. So thanks for making my point.

"Materially closer" isn't the bar to cross the bar to cross is "It just works" which is for me what I'm getting at when I say that we are not 'closer', because it's a binary thing, arrival means 'it works' not that we have made some incremental change towards an ill defined goal.

The 'it works some of the time or even most of the time' bar was crossed years ago, but it needs to work all of the time.

The question was if they are closer. Materially closer is closer? How would materially closer not meet the bar for closer?

Let me give you an analogy. Say we are on Earth looking up and we see the moon. We decide we want to go visit. Every day we add a couple of stones to a large pile on top of a mountain (those are the edge cases). We are provably getting closer to the moon every day. But we will not get there because we need better technology, we don't need piles of stones, we need a rocket.

In your view we are within striking distance of the moon, another couple of piles of stones and we are there. In my view, I'm not sure that it will just take more stones and not a better technology, and I'm beginning to lean more and more towards the latter because the problems that we were seeing 10 years ago we are still seeing today: more and more edge cases. And edge cases can be a symptom of a problem that is almost solved or it can be a symptom of a problem that is being solved using the wrong toolset, especially when there are lots of them and new ones keep popping up all the time.

Clear now?

And by the way: none of this is to imply that what has been done isn't impressive, the MB statement that they will accept liability if their software is in control of the vehicle is a step in the right direction and shows a lot of confidence. But it's incremental progress towards a goal that might not be reachable in that way.

This whole line of reasoning sounds like a philosopher debating the number of teeth in a horse's mouth.

Just come visit San Francisco and see the evidence. Look in the horse's mouth, so to speak. The cars are driving around a complex city, without a driver, and doing an amazing job. The progress is real.

>How would materially closer not meet the bar for closer?

I don't have a horse in this race, but I think what the other commenter was getting at is that: if there were 10 problems that needed solving 5 years ago, and in the course of solving those 10 problems you discovered 100 more than needed solving that you weren't initially aware of, then sure you got closer to the goal, but the progress was so tiny that it's almost not relevant.

I'm personally not arguing one way or another, but that's how I read the other person's comment.

I guess that would be a reasonable take if Waymo announced they were "close" or had solved "x technical problems". But here progress = cars on the road with no driver in more and more difficult situations. So, even if more problems keep coming up they seem to be overcoming them at a good clip.

fwiw, I agree that progress has been astounding in the past 10 years... but on the other hand I really do wonder what the limits of the current approach is. After all humans don't drive by analysing 1 frame at a time, we have a model of how the world works, we anticipate, we have intuition etc...

I know the whole human flight is different to bird flight argument, so I guess we'll just have to wait and see how it all plays out. Exciting times :)

> but on the other hand I really do wonder what the limits of the current approach is.

If the technology stagnates, the hardware stops improving, and the software stops getting better, then we would hit the limit pretty quickly.

> After all humans don't drive by analysing 1 frame at a time, we have a model of how the world works, we anticipate, we have intuition etc...

All self driving cars also maintain those models as well, and use frame by frame input to correct those models in real time (but they have already guessed where something is going to be 100+ frames from now).

cool! I hope we go all self driving in my lifetime, i'd be so stoked to not have to worry about my young boys driving when they become teenagers :)

If your young boys are like mine, then they are sad that they might not actually get to drive cars in the future. However, my 5 year old already talks a lot about self driving Teslas (too much YouTube?) so what do I know?

Isn't the kind of driving that waymo and others have been doing so far the edge case? Normal/ideal driving conditions is not what I spend the majority of my time driving in even if the different kinds of edge cases I drive in are individually less of my time.

They are driving around the streets of San Francisco. It may not be Mumbai, but it's not the highway either.

No they didn't. Cruise didn't even exist 10 years ago.

Why are you bringing up Cruise specifically?

And we have various versions and levels of drivereless cars today...

Do we? Where? The best version currently in existance could be (generously) described as an automated bus that only drives on a pre-mapped grid and still needs human supervision because you can't quite trust it yet. The most well known commercialy available "autopilot" (from Tesla) is so hilariously bad that I honestly don't understand how it's even legal to use, much less to sell for real money.

Neither waymo nor cruise currently require human supervision. That's the whole point of this announcement: waymo is expanding their driverless program to SF.

Whether the grid is pre-mapped or not isn't particularly relevant. It's just how the technology works. The vehicles aren't driving on rails, they just have more semantic information about the world around them.

Waymo arguably reached level 4 with commercial taxis with no human at the wheel a while back: https://www.youtube.com/watch?v=__EoOvVkEMo

Even wide distribution of long-haul trucking exclusively on the interstate, or an automated bus-like commuter service (but actually good, unlike the regular bus), would be a pretty meaningful innovation, wouldn't it?

Of course, and it would be awesome, and I believe that sort of thing is actually possible within a relatively short timescale. General self driving like 90% of people do every day? Nope, don't see that.

We have various automated versions of lane assist, auto-braking, and parallel parking that mostly work.

Fortunately, we also have a developing world where we can pay humans $2/hour to mechanical-turk tele-operate our 'mostly-working' AI.

Past performance does not indicate future returns. However, I too don't believe that self-driving cars will be here in 5 years either.

They also told us we were supposed to have mass produced electric cards in 5 years 30 years ago.

First time I heard of this. Who promised this 30 years ago?

That would have been right around the time of the gm ev1.

A lot of people were expecting electric cars to kick off right after that, so OP's claim is not unreasonable.

We’re talking about promise, not expectation. I don’t recall GM promised electric car mass production within 5 years of EV1.

But with that you also had other fossil fuel companies working hard to make sure that didn't happen

Please explain how fossil fuel companies can keep auto manufacturers from selling electric cars? The reality is that the biggest obstacle continues to be the high cost of battery raw materials.

Who is "they" ?

Google, the tech journalist/blogger consensus, etc.

Sergey might have said it, I don't think the people actually working on it ever thought that or said it?