There have been so many problems with investing in Robotics. The collapse of Rethink Robotics (led by none other than Rodney Brooks) and Jibo and Anki as well as many others leaves one wondering about long-term investor commitment to robotics.
Robotics has a lot of different areas. I think it's safe to describe self driving, drones, and delivery as very actively under development. It's not obvious who will win.
Then you get more industrial applications that attack a specific vertical, like in farming and construction.
Generally when it comes to autonomy, the application needs to work on top of hardware that works. I think that is the core reason why progress is hard: two different very hard problems to solve.
There are a few controverial bits that I'd love to hear what HN thinks.
1. Boston Dynamics Spot, where you have a technical achievement without a clear application
2. What it takes to win in self driving. At one point Musk said "If you have accurate vector space representation, then you're kind of like a video game." That's just not true.
2. "you're kind of like a video game." I think this is a fair characterization. It's obviously not a video game, but there is enormous overlap, once you've successfully modeled your perception in vector space.
This is actually my entire strategy for approaching autonomy, I make as much of the problem space a game; just a serious one. And the most successful optimizations and solutions are the ones that adopt techniques from games. Many academic solutions are formalizations for old-school game "hacks".
In my line of work, training, in many occasions, is about making deep computer vision match ground truth, generated by game tech.
Perception is where it begins. Then you can start to reason. Case in point: just the other day I drove past a line of cars. Right hand side passenger door opens, so I slow down by lifting a bit and sure enough, less than a second later the driver side door gets flung open. If not for reasoning then I would not have lifted and likely would have hit either the door or the former driver of the car when they got out. Those little hints that we humans are good at processing (passenger disembarks: driver likely to follow) would be very hard to teach a computer in such a way that it would not lead to false positives all the time.
I think maybe you could, but I think it would take a lot of data to correlate the edge cases.
Think about seeing a playground ball rolling into the street. Most human drivers would anticipate a child jumping into the street after the ball even though they’ve never experienced that situation before because they can correlate the perception with non-driving scenarios (I.e. playground + ball is associated with children). A self-driving car may not have the non-driving context and I’m not sure society is willing to accept the outcome while the cars gather enough data to learn.
And that’s to say nothing of all the off-nominal maintenance related conditions to account for.
Maybe there can be enough data sharing to update models effectively. It will be interesting to see this industry evolve
Comp vis often reclassifys objects, loosing history..
You see them get in- you know they are closing the doors.
Comp vis sees a butterfly, and sees a car with doors half open.
If you do not allow for frequent doubt, aka re-classification, you get hallucinating systems, imagining a bike-rishka as a slow car with people hanging out
There’s been some cases that highlight the danger of this as well as bad examples of how mitigations were poorly implemented.
I believe a case of a pedestrian being killed by a self driving car was related to the system reclassifying enough to initiate a deliberate delay timer. In a system traveling 60mph this delay in decision was enough to cause a mishap
Cite? The Uber killing was simply that the self-driving system was not allowed to do an emergency stop. All it could do was alert the driver. Presumably because they had a high false alarm rate.
The serious accidents involving self-driving cars have all been huge failures at the basic "don't run into things" task. Google/Waymo, which seems to be past that, has subtle problems, like "projected that bus couldn't fit through space to left of car in wide lane and so turned slightly into path of bus and was scraped."
The standard Waymo accident is "advanced into intersection, detected cross traffic, stopped, was rear-ended by human driver at low speed".[1]
“The self-driving car was fully autonomous at the time at the accident” [1]
It appears the mitigation for a high false positive (I.e., nuisance alarm) was to program in a delay (I.e., “action suppression”) which inadvertently introduced a worse hazard, the timetable from it is in the link below
- perception: sensors -> vector space (a list of objects with their locations)
- planning: vector space -> path
Both are hard, but for the planning problem you can do most of the training and testing in simulation. It's fairly easy to simulate the response of the car.
The perception problem is more expensive to solve, because you have to collect so much data under every weather and lighting condition.
Spot does have that problem. Back in the 1990s, before Boston Dynamics had legged running really working, I was looking at that and running simulations. The only market that looked reasonable was toys. A $1000 toy seemed marginal. The Sony Aibo was in that space, and it was withdrawn. Sony had a follow-on humanoid, but it never shipped as a product.
I met the designer of the Furby and found out what its manufacturing cost is. Let's just say it's less than a meal at most fast food places.
Big Dog and the Legged Squad Support System were impressive, but not useful. The USMC rejected the LS3 and bought small ATVs instead. Much more practical.
(Reminds me of the U.S. Army's flying saucer, the AvroCar. That was a disk-shaped aircraft that was supposed to be a flying Jeep. It was unstable, underpowered, and couldn't get out of its own ground effect. Really cool, though. The Army had another project at the same time, the Utility Helicopter program. That resulted in the UH-1 "Huey", one of the most successful helicopters of all time, with over 16,000 made and many still in use after six decades. It's not exotic but usually will get you there and back.)
I heard through the grapevine that SoftBank sold something like 22,000 Pepper robots at $12k each. I never understood the utility -- I only ever saw them used as chintzy marketing gimmicks.
I could easily see 1000 qty sales of SpotMini for the cool factor alone -- i.e. research labs, corporate labs, startups, etc without ever finding a real "value" application. This drives me nuts as a roboticist & CEO of a service robotics company. But so it goes -- at least it's better than their past product: Youtube Videos.
If BD/Softbank could get the price of Spot down to $12K, it might sell. Softbank might even subsidize that to grow the market. That's what Softbank does, after all.
What the world needs is a good US$5000 assembly robot. Universal Robots has a good one for $23,000. Nobody is building good robots in high volume. Not even Foxconn, which tried.
Your company https://cobaltrobotics.com/ is really cool. I think it highlights that a vertical with specific needs can be met with the right hardware. No need for a face or arms.
Memes in the video. Iterate quickly. Copies in China. Horizontal platform plays are hard. Regulation sux. Public POC makes competitive landscape more dangerous. Validation vs. investment for mass manufacturing / prototyping vs. DfM. Value of robotics in food delivery.
Yet to see anyone else with a robotic startup conclude "iterate quickly" + "relaxed regulations" = move to China and own your own factory, production hardware and know-how, operations and deployment, keeping R&D and POC in-house, in the food distribution domain. This happens to be our exact strategy and play.
Mainly, I'm surprised this was even allowed. My understanding is that if you're not a Han Chinese, you cannot really do business in China, that you're basically swimming against both cultural exclusion, and government exclusion.
You can do business OK. Areas of difficulty include: highly dynamic environment, heavy paperwork, sometimes opaque bureaucracy, really unfamiliar banking system (particularly forex), many regulatory bureaus with seemingly overlapping responsibilities, etc. However, the reality is once you are up and running there is a tremendous degree of freedom, perhaps a more friendly regulatory environment than in most western countries, 100% mobile payment penetration, rapid supply chains for everything, simplified tax, single timezone, etc. It's simultaneously scary and awesome.
8:04 What do you think of Boston Dynamics Spot Mini
9:50 Vertical vs Horizontal applications in robotics
12:24 A proof of concept proves to everyone what is possible, and emboldens copy cats
13:37 What would have happened to Cruise if GM hadn't acquired them
16:15 What does it takes to launch self driving car service like Waymo or Cruise
18:04 Benchmarks for autonomous driving performance
19:48 Why it's better to have customers hungry for a solution
21:06 How to overcome the challenge of rule based behavior planning, like Tesla doing end to end ML
23:42 How early stage startups should build hardware by focusing on iteration speed and tight customer feedback loops
26:58 Vertical vs Horizontal applications and their impact on iteration speed, like building 3D printed houses
28:43 What should hardware companies get done during the YC batch. Ideally cheat with teleoperation if you can
31:17 Why Tesla's approach to getting data from real drivers is very effective path to full autonomy. You need to figure out a business model that isn't bleeding money while you're getting data.
33:47 What Elon Musk gets wrong about end to end autonomy
35:38 What robotics companies would you like to see applying to Y Combinator?
36:36 Drone delivery is working well where people are more desperate. Generally technology is developed where need is greatest, and then moves more broadly to other applications.
38:50 We get accustomed to high quality, like the speed of delivery. The standard can always improve. How many more iPhones would people buy if delivery were every 15 minutes.
40:19 Speedy delivery for auto parts is a clear application.
41:38 How telemedicine can be much more efficient
42:32 Regulations in medicine make developing products far more expensive.
43:15 Regulation is like a fog of war which dramatically lowers iteration speed
44:50 Inaction causes harm, including an example of the cost of $1 parts and the tradeoffs for human life.
46:06 What do you think of Marc Andreessen's "It's Time to Build". San Francisco doesn't feel like the future. We need more experiments.
I like working in unflashy industries. A robot that can harvest peaches (incredibly hard) or work in semiconductor industry. Lots of robots automating the shit out of a very complex process. Toilet roll mills. Anything at 3M.
There is a whole world out there and potentially more fun than trying to invent a 4 legged thing that just looks cool and doesn’t do anything useful.
We need to make it cool to automate clothes manufacturing.
I've always wanted to work on a peach harvesting robot, having grown up on a peach farm. You're right that it's an incredibly hard problem.
We actually had an automated peach harvester when I was a kid. It worked by encircling the base of the tree with a huge net and then shaking the tree trunk to make the peaches all fall into the net. It worked poorly and bruised the peaches too much to be used, so it sat and rusted away. A useful peach harvesting robot must pluck each peach individually.
YES. This is exactly on point. In a top-level comment on this page I posted the same that we need a non-hypey view on robotics and shared a podcast worth listening to on this subject. I find the rosey view of VC on robotics to be completely disconnected with what's happening in industry.
Listen to this podcast on the subject of challenging investing climate in robotics: https://www.cognilytica.com/2020/07/08/ai-today-podcast-what...
Just an alternate non-hypey, non-rosey, non-theoretical, non-SV perspective on what is actually happening in the robotics industry.