@FIR31415 I think so. I was astonished to see how people really do not care about a car with no driver inside! My key takeaway is they have been seeing waymos for so long already (3-4 yrs) that self driving <-> waymo in their heads. Only recently has competition started to pop up...
@FIR31415 Fwiw FSD is still L2 officially (yes, not even L3). And having experienced both, the self driving experience in a waymo is still considerably better than the rest.
@FIR31415 Additionally, its not a two way competition here, there are numerous other self driving companies in various form factors (cars / trucks / taxis like zoox). Its essentially waymo, the proven gold standard vs the rest that have the technology but are yet to exhibit scale.
fxck Diwali parties, all I wanna do is go back in time, help my mom clean the house, decorate the house with lights, light up diya's, attend Puja with the family members, and then light up fire crackers, come back, have mastt "Mummy ke haath ka khaana", gossip, and then go and have the best sleep of my life
@fooobar@paraschopra Pretty new to AI research but benchmarking inherently seems broken (especially in vision+robotics).
I believe all moderately serious AI teams in the industry ought to have their own internal benchmarks for use cases they target right?
Blake is 100% correct, and the commenters are wrong.
The A380 was a bad idea from the start.
Airbus spent $25B developing it forecasting 1,200 sales, needing 500 to break-even.
They sold 248.
When at BCG I worked on a project on this topic (for GE), we concluded that it was doomed from the start. It didn't make any financial sense.
They misread two big trends:
- They assumed that hub-and-spoke air travel (giant airports connecting everything) would get even more dominant (requiring larger planes), but all of the data suggested the opposite.
- They underestimated how much smaller, more fuel-efficient planes (like the Boeing 787 and later A350) would change the economics of long-haul flights.
-As smaller planes had long range, you could support Nashville to London on a 787 (as BA does) rather than a connecting flight Nashville to JFK to London... point to point is obviously better for passengers.
-Even if you have enough daily traffic between 2 large airports to support 2 A380s, customers would be better off spacing them out and putting 4 787s / a350s on the route. Newark to London for example on United has 4 flights a day, but they could only do 2 a380s, making it less convenient for customers. Further, the a380 is actually less fuel efficient than 2 787s!
People will say - but the a380 enables ultrapremium stuff like showers, full bars, and lounges on a plane... that is correct but ultrapremium is not where the money is made - the money is made on business class.
Operating costs on the A380 are so high (4 engines! more landing fees! more gate space!) that filling even a fancy first-class section doesn't offset the fact that you need to fill 400–500 other seats too.
Of the ~180 A380s in service, Emirates flies 65% of them. They really hitched their wagon to it and it makes sense for them!
State-owned Emirates wanted Dubai to become the global aviation crossroads, the "hub of the world." To make it happen, they needed to funnel massive numbers of people through Dubai International. The A380 was perfect for this: huge capacity, premium brand image, exclusivity. They talk about it as a tool for soft power...
Perhaps the worst part is that they spent most of Airbus’ resources (engineering time, budget, leadership focus) during a critical window when they should have been focusing on a mid-sized, efficient aircraft - ACTUALLY the right strategy - and Boeing ate their lunch with the 787 (compared to the too-late a350).
Some of the commenters likely fly the A380 in ultrapremium and love that, but the A380 is a classic case of brilliant engineering chasing the wrong market.
It’s an incredible aircraft solving a problem that didn’t exist and ignoring trends that were shaping the future of air travel.
Benchmarking in manipulation is broken. In a world where 25% mIoU for grasps is one of the most important metrics, this project is such a breath of fresh air. Tons of work needed on tangible benchmarking approaches like this to push robots towards real world usecases