[GTO]

An exceptional article on Bloomberg. We've had many debates on this topic on the forum, and some have always insisted that self-driving cars are a pipedream for India (I'm part of that camp). Heck, forget India, they cannot even function properly in USA where the driving, road & traffic conditions are 50X better. The article says that there has been no real progress in the last couple of years and the self-driving tech of today is just a "glorified demo".



Even Levandowski - who started this whole autonomous driving movement - accepts that public roads are no place for the technology. I kind of agree. For the foreseeable future, self-driving vehicles are best suited to controlled environments only (factories, shipyards, godowns, big hotel properties etc.). But in India? Labour is cheaper than self-driving tech and can multi-task (e.g. a driver who also acts as a security guard, luggage handler or assistant).

And they tend to be confined to a handful of places in the Sun Belt, because they still can’t handle weather patterns trickier than Partly Cloudy. State-of-the-art robot cars also struggle with construction, animals, traffic cones, crossing guards, and what the industry calls “unprotected left turns,” which most of us would call “left turns.”

“It’s a scam,” says George Hotz, whose company Comma.ai Inc. makes a driver-assistance system similar to Tesla Inc.’s Autopilot. “These companies have squandered tens of billions of dollars.” In 2018 analysts put the market value of Waymo LLC, then a subsidiary of Alphabet Inc., at $175 billion. Its most recent funding round gave the company an estimated valuation of $30 billion, roughly the same as Cruise. Aurora Innovation Inc., a startup co-founded by Chris Urmson, Google’s former autonomous-vehicle chief, has lost more than 85% since last year and is now worth less than $3 billion. This September a leaked memo from Urmson summed up Aurora’s cash-flow struggles and suggested it might have to sell out to a larger company. Many of the industry’s most promising efforts have met the same fate in recent years, including Drive.ai, Voyage, Zoox, and Uber’s self-driving division. “Long term, I think we will have autonomous vehicles that you and I can buy,” says Mike Ramsey, an analyst at market researcher Gartner Inc. “But we’re going to be old.”

“You’d be hard-pressed to find another industry that’s invested so many dollars in R&D and that has delivered so little,” Levandowski says in an interview. “Forget about profits—what’s the combined revenue of all the robo-taxi, robo-truck, robo-whatever companies? Is it a million dollars? Maybe. I think it’s more like zero.”

One of the industry’s favorite maxims is that humans are terrible drivers. This may seem intuitive to anyone who’s taken the Cross Bronx Expressway home during rush hour, but it’s not even close to true. Throw a top-of-the-line robot at any difficult driving task, and you’ll be lucky if the robot lasts a few seconds before crapping out.

“Humans are really, really good drivers—absurdly good,” Hotz says. Traffic deaths are rare, amounting to one person for every 100 million miles or so driven in the US, according to the National Highway Traffic Safety Administration. Even that number makes people seem less capable than they actually are. Fatal accidents are largely caused by reckless behavior—speeding, drunks, texters, and people who fall asleep at the wheel. As a group, school bus drivers are involved in one fatal crash roughly every 500 million miles. Although most of the accidents reported by self-driving cars have been minor, the data suggest that autonomous cars have been involved in accidents more frequently than human-driven ones, with rear-end collisions being especially common. “The problem is that there isn’t any test to know if a driverless car is safe to operate,” says Ramsey, the Gartner analyst. “It’s mostly just anecdotal.”

For now, here’s what we know: Computers can run calculations a lot faster than we can, but they still have no idea how to process many common roadway variables. People driving down a city street with a few pigeons pecking away near the median know (a) that the pigeons will fly away as the car approaches and (b) that drivers behind them also know the pigeons will scatter. Drivers know, without having to think about it, that slamming the brakes wouldn’t just be unnecessary—it would be dangerous. So they maintain their speed.

[srini1785]

List of projects on which big bucks were/are being spent on:

1. Artificial Intelligence : Includes all manner of pursuits including the subject matter.
2. Alchemy (Turning everything into gold) : Start a business and learn yourself.
3. SETI (Search for extra terrestrial intelligence) : Blue men/women are already among us.
4. Holy Grail : No comments.
5. Social Equality : Forum rules prohibit political discussions.
6. Bug free software : I am no Software guy but others can comment.
7. Immortality : Last checked it was achieved by cryogenic freezing.
8. The perfect car : Enough Said.
9. Traffic free Bengaluru : Go to https://www.team-bhp.com/forum/stree...ml#post5418005 (Rants on Bangalore's traffic situation)
.
.
10. Wisdom : Its quite within reach. Just get past the first 9.

[dragracer567]

Never say never but the point is that self-driving cars can't work with the current road infrastructure (even developed ones) because these roads are specifically designed for human drivers, not robots. This is the equivalent of trying to drive a train on the road.

For self-driving cars, road infrastructure will have to be specifically built around it but the problem is, on such roads, human drivers may not be able to drive and the overwhelming number of cars on the road only have basic cruise control by the way of driving assistance. It's a chicken or egg situation.

[pacman2881]

And it is no coincidence that Amazon has scaled back on it's delivery robot - Scout

https://nypost.com/2022/10/07/amazon...o-early-grave/

The company didn’t elaborate, but some observers reported that it struggled to get past obstacles on the sidewalk, such as trash or other minor debris.

[Samurai]

I used to be very scared of AI taking away jobs. In fact, back in 2015-17, I went around giving talks on how AI is going to kill IT jobs. Since then, I put plenty of efforts into getting deep understanding of AI applications, just to protect my own career from ending.

Now I work with multiple AI companies, helping them deal with system engineering challenges. I am still no AI expert, but I have a close view of what is possible and what is not possible. I am no more afraid of AI, and I have a whole new appreciation for the human brain.

Why is that? It goes to one more point mentioned back in 2017:

Originally Posted by Samurai

Think about evolution. It took humans 500+ millions of years to evolve from a single cell organism, because the evolution by natural selection is so slow. However, computers can do the similar evolution in matter of minutes. It took humans 400 years to go from thinking sun goes around the earth to landing a man on the moon. But computers are at least million times faster in thinking. Once AI is mature enough to start evolving, it can evolve so fast, human brain will be incapable of instructing such an advanced computer what to do. This is what Elon Musk fears.

AI can build general intelligence only if it starts evolving. Otherwise, they are just doing regression analysis (linear & non-linear) at a very high speed. Unless they can evolve, they can't come up with original ideas. Humans cannot be beaten unless AI can come up with original ideas.

[Kosfactor]

Self driving cars will always compete with public transportation system - you do not have to drive it anyway as a passenger. It would be a better solution than clogging up the roads with these dimwitted machines. There is always drivers for hire, cabs, train, bus etc.

I like ADAS though, takes away a lot of fatigue in stop and go traffic, boring speed limited highways etc.

[locusjag]

Self-driving tech and AI in general have hit a glass ceiling. If my understanding is right, the computing world's gains as per Moore's law had stumbled for the first time in decades a couple of years ago. I remember reading about limitations in current chip etching ability beyond the nanometer scale.

There will be some other breakthrough in computing power until when, the current AI winter (as I read about it) will continue.

For reference - the current level of self driving tech and the entire buzz around it came about once they figured out a way to use Nvidia's GPUs for Deep Learning. This has been fully milked for its worth, in my understanding.

There are many efforts going on meanwhile, to break new barriers in computing power - the human brainome project, the bacterial quantum computing project etc. Who knows what will be the next big proven thing?

Caveat: I source my knowledge on these things from e-books on Kindle and news articles. Please do educate me if I got anything wrong here. I am NOT a computer engineer or a VLSI chip guy or whatever.

[apollo_1999]

At least for now self driving cars with a human override system (Level 3 or level 4 ADAS) is the way forward. Cars like any other machines are prone to failures. In that case an input from a human driver can avoid a potential catastrophic situation.As of now, I am quite skeptical about sitting as a Passenger in a driverless car.

Also self driving cars requires a whole new set of legal formalities, which are yet to be introduced by many countries in order to drive on public roads.

[dragracer567]

Originally Posted by Samurai

I used to be very scared of AI taking away jobs. In fact, back in 2015-17, I went around giving talks on how AI is going to kill IT jobs. Since then, I put plenty of efforts into getting deep understanding of AI applications, just to protect my own career from ending.

Same here but with automation systems (context: I’m a marine biologist, not an engineer). I used to think automation will replace a lot of jobs but after working with some automation systems on a project recently, all I can say is that it just seems to change the ‘type’ of workforce needed rather than reduce the workforce requirement drastically in terms of quantity. It’s more about increasing productivity and product quality than reducing the workforce (perhaps a real engineer can correct me on this).

I know this is a cliched line on T-Bhp but this is the equivalent of expecting stable-boys to lose jobs at the advent of cars - cars actually bring many more opportunities.

[ajayc123]

That we are taking time to make progress in the field is indication that it is a hard problem to solve.

Mankind has made unimaginable progress in the last few decades.

Given the talent, money and hunger for innovation, that we have in this world, I am reasonably optimistic that lot of us should be able to withess self driving cars in our lifetime.

[Spaced Out]

Whether self-driving cars ever make the road or not is missing the point. These moonshots are what leads to new tech in other areas and entirely new industrial / economic sectors being created.

It was in 2007/8 that I saw self-driving cars researched/ tested in a tech University in USA. I used to pass a car with laptops etc insides and one day I finally asked the grad students working on this. It was a research project on self-driving/computer assisted cars with a focus on elderly.

Post-2010, from an occasional article on AI/self driving cars et al, the Wall Street Journal began reporting on it nearly daily.

Bottomline: US universities excel in these pure research projects that often translate into industry based moonshots that may advance or create entirely new industry sectors.

Contrast this with my work here with our premier tech institutes. The general mindset here is that we have already arrived.

[shree_shell]

I read this article's headline on Instagram a few days ago and chuckled. I am sorry, $100 Bn is not a lot of money when you look at other tech bets or gambles. Meta alone, for instance, has made a $10 Bn bet (and growing) on the Metaverse, and there are many more players - Apple, Bytedance(TikTok), Microsoft, HTC, Google, etc. I think we can safely say that there are more players in the self-driving car game than in VR; I am aware this may be false equivalence, but I had to compare the tech's two biggest bets in our times.

As far as what may be holding back self-driving tech to be on roads, it is important to realize that the current players are working on multi-year research projects. These cars are quite literally experimental setups and experiments in action. Again, to put the $100Bn dollars in perspective, you may want to take a look at how much money is being spent on under-funded academic research. These experiments on public roads are privatized and corporate versions of what was undertaken through Darpa by academics ~2 decades ago. The benefit of the Googles of this world running Self-driving tech research projects is that they have a lot of money to attract the right talent, more specifically Deep Learning and Control systems talent.

There are lots of moving parts on this robot-on-wheels. Apart from the usual suspects - DSP, Sensor-fusion, there's the new electricity - Machine Learning. ML can help with four main tasks - Perception, Localization, Planning, and Control as well. Also, Deep Learning, a sub-field of ML started solving seriously non-linear and challenging problems in 2012-14, so from that perspective, we have really spent only 8 years in Self-driving tech - developing the AI stack from the ground up.

Given all of this, expecting a Driverless car on US roads today (forget about India) would've required a technological-singularity miracle. It hasn't even been a Decade since this mega-research project moved on to ML-stack. I would recommend the skeptics look up major ML conferences like - NeurIPS, CVPR, ICML, etc. to understand the state of the underpinning that eventually goes into the clumsy experiment on the road.

This is an ML research scientist's POV, who works at one of the major FAANG. Of course, if I were to be an investor who dropped millions in the 2000s, I'd be pissed. And no wonder this article is on Bloomberg and not on MIT Tech review.

[mxx]

Self-driving cars will be on the streets some day. It is definitely not ready now. Even for developed countries it is still some time away.

What is absurd is the extremely optimistic time lines many companies and scientists had announced in the past. Probably to get funded easily as most investors wont invest on very long return time lines.

[Tanmay007]

Originally Posted by GTO

The article says that there has been no real progress in the last couple of years and the self-driving tech of today is just a "glorified demo".

We are in the initial phases of a new technology. A tech as complex as self driving will take several years to mature. The no real progress in past 2 years is not really true, check out some of the videos below to know where self driving is at. These would be enough to blow your minds.

https://youtu.be/6v036bBD31o

https://youtu.be/N4X4GMFmTb0

Originally Posted by locusjag

the current level of self driving tech and the entire buzz around it came about once they figured out a way to use Nvidia's GPUs for Deep Learning. This has been fully milked for its worth, in my understanding.

It's hasn't, not even close. Self driving as of this moment is a software problem, not hardware. The hardware we have today is more than capable.

Originally Posted by mxx

Self-driving cars will be on the streets some day. It is definitely not ready now. Even for developed countries it is still some time away.

What is absurd is the extremely optimistic time lines many companies and scientists had announced in the past. Probably to get funded easily as most investors wont invest on very long return time lines.

100%. Fully self driving cars without any drivers are inevitable. The absurd timeline initially promised by Elon and others has set wrong expectations among people, now that the timeline has well past people have started losing faith it will ever come.

IMO, the US is at least a good 10 years away from robo-taxies going mainstream. There will be some such as Cruise doing controlled runs in certain areas in next 5 years.

As for India, timeline will probably be 2-3x of that.

[roy_libran]

I tend to agree, and for one hope that humans continue to be at the wheel. The way I see it, AI can be trained, but that has to happen in the real world, where real humans will be subjected to the risk of the AI failing. The number permutations and combinations of situations and responses is infinite, and the implications of getting it wrong, disastrous.

Example -

https://www.youtube.com/watch?v=yRdzIs4FJJg