2017 Investor Roundtable:General Discussion

[dakh]

This is just fundamentally wrong -- Sterling Anderson doesn't know what he's talking about and is exposing his own ignorance. He's missing two major characteristics of auto traffic:
(1) Rural areas. The value of your own car is that it is ready to go immediately. A shared autonomous car has to come to you first, and in a rural area, that's a *long* delay.
(2) Rush hour. The number of cars is sized for *peak* demand, not average demand. There will always be cars sitting around doing nothing at midnight, being used during rush hour.

Now, if you happen to live in a big city where there is likely to be a car nearby at any moment, (1) doesn't apply as much... unless you're in rush hour and all the autonomous cars have been taken already. And if you happen to live in a city which is very active 24/7 and has widely staggered shift starts and stops, then (2) applies less than it does to, for example, the Gigafactory with the entire shift arriving and leaving at the same time.

But these two factors are huge, and dominant. How many cars are really sitting idle *during rush hour*? That is the MAXIMUM number of cars that autonomous vehicles can permanently remove from the roads. Then subtract from that all the cars which are at isolated ranch houses or farms where it would take ten minutes to get to the next house. That even lower number is the MAXIMUM number of cars you'll see displaced. You'll actually see fewer than that displaced because many people will choose to have their own car for other reasons (not wanting smokers to use the car, etc.)

My point is that autonomous cars aren't really going to reduce the number of cars much. If you have data which shows that there are a lot of cars, in dense urban/suburban areas, sitting idle *during rush hour*, I would love to know how many there are. Maybe it's more than I thought.

The effect of autonomous cars on *cost* is another matter. I've often thought that big cities ought to have a lot higher usage of taxis (or the taxi-like Uber and Lyft and so forth) than they actually do, and the reason is probably the cost. Autonomous cars, once you've got enough corner cases working, are strictly superior to taxis in every way while simultaneously being much cheaper to operate. They won't just replace the taxi market, they'll substantially enlarge it.

But there are no taxis or Ubers in rural areas and it isn't for lack of drivers, it's for lack of practicality.

Ugh. Especially because you so confidently state that Anderson doesn't know what he's talking about I gotta get back at ya on this. autonomous cars aren't really going to reduce the number of cars much is a bold claim but your supporting arguments are weak.

1. Urban population in US is 83%. Rural doesn't need to be in the picture at all for a substantial impact.
2. Current personal vehicle utilization is ~5% so replacing 5 cars means the 1 replacement fleet car will be sitting idle 75% of the time. That's quite generous.
3. Anderson says it CAN displace, he doesn't make any claims about how many WILL. It's not about the rush hour, it's about how many people will find cost/benefit of not owning a vehicle preferable, and how many will like the vehicle to pay for itself when they don't need it. We know millennials don't fancy car ownership so the norm of car ownership might change to using fleet cars faster than you think. Sure some people will cling to vehicle ownership but overall the rational thing to do is to share so unless you can prove that people will be especially irrational in this area, it'll happen.
4. Rush hour isn't an hour, it's many hours and usually reverse traffic is light. One fleet car can do multiple trips on the same day in the same direction, going back empty if needed.
5. Real-time carpool arrangements with fleet vehicles is be a killer app. Sure you have to deal with the vehicle picking up other people but you get where you're going much faster due to carpool lane utilization.

I can keep typing but I think that's enough already.

 

[dakh]

I've been through AI projections by "experts" and companies before. They're always wrong; it always takes longer than they think. And this isn't just analogy; I actually do understand the technology.

Look, their concept of the specifications they're trying to meet isn't even right yet. Once they understand what the actual hard problems are (they're only just starting to get to them, in the last year) they can start working on them.

It's harder to specify the problem than it is to solve a well-specified problem. So-called "deep learning" does not get around this issue; you still have to be able to tell good driving behavior from bad, including in many many many many corner cases. They don't even know what the corner cases *are* yet, and they've made that VERY clear with many monumentally ignorant and arrogant public announcements.

Oh, they'll get it done. But not by 2020.

Elon Musk, the eternal optimist, is not even claiming that they will have full all-cases autonomy by then, at least not in any interview I've seen. He's been talking about "highway entrance to highway exit" autonomy in the next few months -- *that*, I believe they can do.

I don't have time to type up longer responses now but this also is an opinion that doesn't hold water. The corner cases might be many but solving them isn't rocket surgery and is a simple matter of throwing more resources at the problem. Moreover, they can be solved in parallel so the return on increased resources is near linear. What they claim is that with the current hardware those problems are solvable, and they'll get there pretty soon.

Computer doesn't have to mimic humans. It just needs to figure out a way to get from A to B without breaking the rules, getting into accidents or creating road hazards for others. And it's vastly superior to humans in the ability to react to a dynamic road situation.

And no the goal isn't highway exit to highway exit. It's SF to NYC including city driving, without touching controls. And they'll get there soon, not in 10 years.

 

[The Blue Owl]

Another article in the Financial Times about the aggressive investments in renewables being unleashed by China (paywalled, unfortunately -- you could Google the title and try to get in that way).

Wave of spending tightens China’s grip on renewable energy:

Chinese companies made 11 outbound investments in excess of $1bn in 2016, adding up to a combined $32bn, compared with eight deals for a combined $20bn in 2015, according to research by IEEFA.
...
Four of the five biggest renewable energy deals worldwide in 2016 were made by Chinese companies, according to Mr Buckley, and he predicted this trend would continue irrespective of any change in approach by the US.
...
By far the biggest of the deals struck in 2016 was the acquisition of a controlling stake in CPFL Energia, one of Brazil’s biggest generators and distributors of renewable power, by State Grid Corporation of China as part of a proposed takeover expected to total $13bn once completed.
...
China invested $103bn in domestic renewable energy in 2015, according to Bloomberg New Energy Finance, compared with $44bn by the US. China’s National Energy Administration this week committed to further increases in years ahead with a plan for Rmb2.5tn ($363bn) of domestic investment in clean energy by 2020.​

It's not a wave. It's a tsunami.

 

[ggies07]

On that train of thought: what's the next big business idea that will be able to use all those obsoleted gas stations and the little chunk of real estate they're on? Suspect someone will come up with an idea and be able to pick up lots of disused gas station properties for a song, because they're too small for many purposes

My aunt used to work for corporate BOA and one of her clients was QuikTrip. She told me two years ago that she still talks to the CEO from time to time and that they are working on this transition we are heading into and which options are best for them. Should be interesting to see what happens with the company.....

 

[JGInnovation]

Pizza delivery? That's so last century. More like a monthly bulk delivery of base components for your 3D printer/baker.

your right, not the most exciting example. The interesting thing with Musk is he designs a product to create the demand for the network he wants to create e.g. Superchargers, inexpensive batteries, autonomous car sharing. He needs alot of rocket launches to bring the cost down to make the Mars mission make sense, so he creates a product to make it make sense and thats why he is talking about a satellite communication network. Tesla cars create demand for a communication network, the satellite communication network creates demand for rocket launches etc... Tesla creates demand for electricity which creates demand for solar panels etc... He may have pushed Hyperloop out to the world as a concept because he wants it to exist but it doesn't fit into a symbiotic relationship with these other initiatives.

 

[SBenson]

<snip>
They don't even know what the corner cases *are* yet, and they've made that VERY clear with many monumentally ignorant and arrogant public announcements.
<snip>

Can you give a few examples? Thanks

 

[Zaxxon]

Well, gosh. I'm not sure how stealthy that big-assed building they're building at breakneck speed is (ahead of the schedule they told the state of Nevada), but okay if you were looking the other way you might have missed it. I suppose one might have been fooled into thinking they were filling it with runaway Pokémon...because yeah, what else would you put in it? Certainly not battery cell manufacturing equipment for less than 20B.

That'd be one hell of a Pokémon gym.

 

[racer26]

Once cars attain full autonomy, the space can be used for parking the cars in each neighborhood on stand-by

That's... actually a pretty good idea. Bravo.

 

[jhm]

You don't understand the business plan.

The former gas stations (now convenience stores) simply install FREE TO USE chargers.

Why would they charge money for using them? The gasoline was a loss leader for them anyway. The charger is going to be a loss leader too, but it'll burn less money than the gasoline tanks and pumps did.

Look, this is already happening even without electric car chargers! I can name three former gas stations in my area which ripped out the pumps and tanks but are still convenience stores.

These "gas stations" were located at spots along the road where people like to stop to get a bite to eat or a drink or go to the bathroom; they're still good convenience store locations.

Perhaps you don't understand the existing economic situation. The 'upstream' oil companies make money selling the oil they pump out of the ground; the refineries buy it, turn it into gasoline and sell that at a profit; the delivery companies charge a markup and truck it to the gas stations; but then the gas stations sell it at cost, or sometimes at a loss. In order to get people to come into the convenience stores. They hate it when you buy gas and don't come into the store to buy soda or candy.

The gas stations are not owned by the oil companies any more, with rare exceptions. They actually pay franchise fees to the oil companies to put the Exxon/Shell/BP/Sunoco name up, which makes the gas even less profitable.

If you offer the owners of the gas stations a different way to bring people into their convenience stores, with no franchise fee, no environmental liability, no tank inspections... they'll go for it. (As soon as they believe that there are enough electric car drivers that it will work.) The amount they pay for electricity for the free charging? Probably smaller than what they were losing on maintaining the gas pumps to EPA standards. As soon as the number of gas car drivers drops too low, so that the gas pumps are not bringing in enough business, they'll rip the gas pumps out.

I do believe that there is a sustainable market for nice travel stores, and gas retailers are definitely moving in this direction. As an exclusive Tesla driver, I use these stores about 10 to 20 time per year, mostly for a restroom and drink. This use is only when I am traveling long distance, not daily driving. So as more families go all electric there will still be demand for these store even from drivers who do not consume gas. But the worry for the industry is the frequency of use. Some gas stations will survive this transition, but many will not. Consumers will ultimately decide. Personally, I'd rather do a little charging as a cafe then a typical travel store. So travel stores are competing with Starbucks, McDonald's, Dunkin Donuts, and on and on. And they know that. But the key thing is that gasoline will lose its ability to generate foot traffic. It will cease to work as a loss leader. That's the disruption.

 

[NOLA_Mike]

At the risk "poking the bear" but waaaaay too funny:

 

[RobStark]

Yes. I definitely agree that the "four gas stations at one corner, one on each side" thing is going to go away fast.

That is already gone in my section of Los Angeles.

In the last 10 years we have had 9 gas stations closed in residential areas and 3 open hugging the 405 Freeway (Highway to non-Californians).

The new ones all have a minimart and at least one fastfood franchise co-located inside it. Taco Bell, Subway or Jack-in-the-Box.

 

[RobStark]

Costco chooses their markup on items

Costco must adhere to market markups.

To low they make now money.

To high they lose customers.

Grocery stores have a 2% margin.

Gasoline stations 1% on gas.

Kirkland clothes are a commodity. No Dolce and Gabbana markups much less Ralph Lauren.

 

[jhm]

Sure, its a subset concept of disruption theory (but not limited to low cost entry). Basically you look at a future network change that is coming e.g. Charging vs. gas stations, packet switched data on mobile networks vs. circuit switched (circa 2002), High speed mobile data (3G) vs low speed mobile data (2G), High speed internet vs. low speed internet, cheap satellite launches vs. expensive satellite launches, autonomous cars, local power grid, cloud vs. local etc... Instead of waiting for the new network to be ready you build a business concept that can make some sense today but really makes sense when the new network becomes established. Thus giving you first mover advantage with all the right business values where the incumbents are stuck in the switch of networks with old values. BlackBerry did this in the early days (1999) by launching data only devices on outdated data networks which allowed it to compete later when the mobile networks added data capability, Nokia couldn't compete in this new space because they viewed the world as a phone, BlackBerry viewed it as data device. Then Apple did it to BlackBerry by launching a device that viewed the world as a high speed connected computer, the original iPhone was 2G and the apps like Browser and Maps only worked effectively when in WiFi. When mobile networks (3G) were like WiFi is when BlackBerry lost. Netflix did it with some luck, they leveraged the postal service and a subscription model and a recommendation engine which fit perfectly when they went to streaming. Tesla is doing it with Charging but they are generating their own demand as a reason to build a network.

Autonomous cars will disrupt alot of stuff, but one of the networks will be vehicles not designed for people at all - e.g. A pizza place could be located in a bad location and it is 100% delivery based with the pizza cooking while the autonomous car drives. Or a retail store on wheels without a physical location, it comes to you. The trick is starting the business now for this autonomous future in a way that can still work. Uber is doing a bit of this with things like UberEats.

Super! Perhaps the oldest example would be Noah building the Ark. Thanks.

 

[JGInnovation]

Costco must adhere to market markups.

To low they make now money.

To high they lose customers.

Grocery stores have a 2% margin.

Gasoline stations 1% on gas.

Kirkland clothes are a commodity. No Dolce and Gabbana markups much less Ralph Lauren.

Costco's markups are half of Walmart. It's designed to break even. Profit comes from memberships. They are different business strategies, that's the point I'm making. Tesla can employ a break even operating strategy for superchargers like Costco because of an upfront access fee embedded in the cost of the car with no need to sell high margin items at a Supercharger to make up the difference.

 

[Off Shore]

may even get subsidies from the government.

May even KEEP subsidies from the government. FTFY.

 

[neroden]

I don't have time to type up longer responses now but this also is an opinion that doesn't hold water. The corner cases might be many but solving them isn't rocket surgery and is a simple matter of throwing more resources at the problem. Moreover, they can be solved in parallel so the return on increased resources is near linear. What they claim is that with the current hardware those problems are solvable, and they'll get there pretty soon.

The software doesn't even know that it's supposed to honk the horn when going around blind corners.

Bluntly, they haven't started implementing the hard parts of autonomous driving. If you think they can do it in two years, you haven't been paying attention to how *amateurish* they are.

Maybe if they hired me they could do it in two years, I dunno, but with the sort of clowns who like to work on this stuff, it'll definitely take longer than that. I mean, it's taking them years just to get the cars to maintain lane position.

If they hired expert driving instructors to tell them what corner cases they needed data from and then started collecting that data specifically, they could perhaps do it in two years. They haven't.

They have a billion miles of data and how much of it is from one-lane bridges with traffic? Yeah, probably about 4 miles of the data? Maybe 10?

Water sluicing across roads in torrential rain conditions? (They'll collect this data sooner.)

 

[3Victoria]

The software doesn't even know that it's supposed to honk the horn when going around blind corners. ...
Water sluicing across roads in torrential rain conditions? (They'll collect this data sooner.)

Yes very hard to get 100% complete ... but it doesn't need to be 100% complete to pass muster. You cases are a miniscule part of driving. If they do well on the commonest 90% and better that wetware, then they will be ahead. I don't ever expect them to navigate goat paths, drive across lakes, or even in blinding snow storms ... discretion is the better part of valor.

In other words, lets see how they get on ... they don't need to, and will never, make AP perfect ... not thd goal.

 

[Alketi]

Maybe if they hired me they could do it in two years,

Wow.

 

[Krugerrand]

The software doesn't even know that it's supposed to honk the horn when going around blind corners.

They have a billion miles of data and how much of it is from one-lane bridges with traffic? Yeah, probably about 4 miles of the data? Maybe 10?

Water sluicing across roads in torrential rain conditions? (They'll collect this data sooner.)

Realistically, an untrained computer program isn't going to behave any worse than people do in these situations. IE., I've lost count the number of times I've seen news reports of people needing to be rescued from their vehicles because they attempted to drive on a flood road. Just sayin'.

 

[dakh]

The software doesn't even know that it's supposed to honk the horn when going around blind corners.

Bluntly, they haven't started implementing the hard parts of autonomous driving. If you think they can do it in two years, you haven't been paying attention to how *amateurish* they are.

Maybe if they hired me they could do it in two years, I dunno, but with the sort of clowns who like to work on this stuff, it'll definitely take longer than that. I mean, it's taking them years just to get the cars to maintain lane position.

If they hired expert driving instructors to tell them what corner cases they needed data from and then started collecting that data specifically, they could perhaps do it in two years. They haven't.

They have a billion miles of data and how much of it is from one-lane bridges with traffic? Yeah, probably about 4 miles of the data? Maybe 10?

Water sluicing across roads in torrential rain conditions? (They'll collect this data sooner.)

Not much to respond to. Why do you think they haven't talk to driving instructors or whatever else you're suggesting they do? Of course you know they didn't talk to you but given your responses that maybe is a good thing.

Not a single one-lane bridge traffic situation going from SF to NYC. Sucks for you that they might not address your corner case for a while (and anyone who lives in Hana or wants to go there). If they got all of the 4 miles of such data, that simply means they would be well served not to care about it for a while.