There are companies (like Kiewit) that specifically hire and pay (a lot more) to incentivize their employees to travel. It's a really really hard job and many people burn out. The travel is brutal.
The worst is that it’s not even semi-permanent move. You go to the site, you work for some months then you need to move to the next. If by accident you have a family then guaranteed they cannot follow you and you will be seeing them only sporadically.
CAT, Deere are both doing very interesting things with older autonomy techniques. Deere has acquired several companies, and partnered with others to bring in talent from outside. CAT has worked with outside companies (notably Trimble, Topcon) for key technologies when it makes a big difference. Both are awesome companies, but not AI/ML companies at the core and it'll take a lot of work for them to get there. I think this is very much like the self driving world 10 years ago where OEMs tried very hard to become software companies, but ultimately Cruise and Waymo were the ones that executed.
Neither Cruise nor Waymo seems to be profitable yet, and the jury is still out on whether they will win the market. They may be the MySpaces (or the Fiskers) of autonomous driving.
There is red tape and certainly there are planning bottlenecks, but the GCs already deal with this. Once a project is funded and started, an established GC will move quickly to execute on projects. Also - some states (CA in particular) have an absurd amount of red tape. If you go to states like Texas, it's much easier. Robotic tools are pretty common already (machine guidance for bulldozers, remote ops for excavators, mine trucks), so there's good precedent. We view our machines as tools for people to use, and given the labor challenges, have a ton of customer interest as a result.
Someone I know worked on the rad tape for the solar power plant in Ivanpah California, and he estimated that it added ten to 15 percent to the budget, not counting the added cost of the delays it causes.
I'm the CTO and one of the founders of Bedrock. I was very pleasantly surprised to see the excitement from this crowd! Happy to answer any questions about us (and will look through the comment threads here). We're looking for really really awesome MLEs and software engineers, so if you're interested take a look at our careers page https://bedrockrobotics.com/careers
Small piece of feedback: It takes far too much scrolling to get to the list of open roles. Maybe that's deliberate, so you know your applicants are truly serious about working for you, but I could see a lot of potentially highly-qualified candidates just dropping off due to plain annoyance.
to add to this, hijacking default scrolling behavior is such a massive no-no and super freaking annoying. it's also problematic for accessibility reasons.
This is more of an open-ended question, but do you think there will be a sure in traditional (non-software) engineering demand, as well as software engineering in more of a hands-on hardware context?
Software engineering IMO is alive and well, we're just going faster. Job will shift with AI, but that happened with compilers, etc. Clever engineers will continue to build. Going faster could mean you mean fewer SWEs, but could also mean you just go faster and build more things. If we go faster, the next thing will be down the chain (enough compute, enough hardware to work on). Since AI has always been insanely software bottlnecked, if it goes faster yeah there will be more jobs.
Robotics works now! We should build more robots! Some of the best companies in the next decade will for sure be building some of the robots from Star Wars and the Jetsons. This will be heavy on hardware. Outside of robots, hardware will change too - no idea if Meta's glasses are the best thing, but certainly the iPhone format isn't AI native. This will be new hardware, and we'll need smart objects everywhere in our lives (car, home, etc). Will be very cool and definitely more hardware oriented.
I read your careers page out of curiosity. It's like all software and equipment (what I interpret from "hardware"). Shouldn't you also hire some structural engineers and similar?
We’re mostly focused on the AI and software side of things. There are many great manufacturers for the machines themselves. We have amazing hardware engineers, so if we need to do structural work for our parts, we will. At the moment it’s not core.
Ah I misread. We’re working very very closely with design partners/customers for that side right now - they do more civil work than we will ever do and are happy to help us. We also have advisors who are in the civil space.
We don't have any relationship with Trimble or Topcon, but yes much more advanced capabilities on the machine. On caving/shoring - ideally we avoid having people in at least some of the trenches... if you can dig a trench and lay a pipe down without manipulating it by (human) hand, then trench collapse is less of an issue. Obviously doesn't work for everything, but we have LIDAR on board, so we can handle surveying remotely without someone standing in a cut with a stick.
Faulty geotechnical - I'd be interested in understanding how the operator helped here. Will take time, but anything the operator is doing (watching subsidance, feeling the ground, watching a wall collapse, confirming that the ground is actually what you think it is), will be doable with the sensors we have on board. The operator we're building will have many lifetimes of experience, so will be able to learn things that even the best operators can't. We saw this at Waymo (now at 100 million miles!!!!) - it seems like magic, but if you had 1000 lifetimes of experience, you'd be incredibly good at very subtle things like this.
Are you planning on having these robots work collaboratively alongside humans? Or something more complicated like only driving near humans to start with but digging near them being a longer term goal?
Humanoids will be amazing when they get here, but I think there are a lot of challenges between now and then - hardware alone is very challenging - add a ton of sticky mud and it’s going to be complex even if the software works.
i had to look up whether stuff like mini excavators use control-by-wire/CAN bus or if the controls are actually mechanically operating valves, to determine how hard it'd be to just throw a lil computer connected to the CAN bus to autonomously operate equipment. it seems like newer and nicer models are control by wire, and older ones have joysticks and pedals that directly open and close hydraulic valves.
i looked this up in a "cost of humanoid robot" versus just doing a retrofit on a computer/actuators on older equipment. i think even in the actuator approach, adding 12 electronically controlled hydraulic valves to replace the human actuated valves is still gonna be cheaper than a humanoid robot
Probably - if cost goes down to operate a machine, then the base machines can be smaller. Will eventually be interesting to go for much bigger changes... ditch the cabs, lower profiles, place your cameras in smart spots, move the pivot points so they are best for the job instead of best for a person. Vehicle distribution might change, too - heavy machines are tools and the tool you choose is a combo of best quality, right cost, speed, etc. That landscape will shift.
The variety of machines and their specificity is super fascinating and very specific. Definitely will change.
I am very much looking forward to seeing how things evolve. Do you think the bitter lesson applies here? I am assuming you are gathering reams of data from the controls and any cameras of manually operated machines. I have been thinking about the bitter lesson in many contexts recently and I think for something like this where there really isn't anything proprietary about the method of operation, doing some kind of data consignment agreement or just up front paying for data could be very valuable in and of itself. Even unlabeled data can later be autolabeled.
