Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Short-Term TSLA Price Movements - 2016

This site may earn commission on affiliate links.
Status
Not open for further replies.
Update from German stock market: Tesla up 1%, market indicators also about 1% up. So is oil - but wheat is down, what do we make of that?!

- - - Updated - - -

Well, a bit off topic, but really priceless: BMW CEO Harald Kruger makes somewhat harried and less than gracious exit from Tesla Model X that he was checking out at the Geneva Auto Show after being questioned by the German journalist - all caught on tape:

Watch BMW [Video] | Electrek

:biggrin: Looks like he totally didn´t get the situation, trying to show the Tesla manager how to open the door. Also love the quote from reddit in the electrek article:

“He literally hit himself on the head with denial…” – CadarF

 
This is off-topic, but it's the only place where people will read it, so here it goes: DeepMind's AlphaGo software defeated the best human Go player alive, Lee Sedol, in a historic game. Go is an ancient Chinese game (2500 years old) vastly more complex than chess. For perspective, the number of possible Go positions is larger than the number of atoms in the Universe.

Just a few years ago it was generally believed that it would take decades for AI to defeat the best human player at Go, with many believing it was never going to happen. It happened this morning.

This is relevant for self-driving cars. I will go on record and say that Elon is mistaken in his belief that machine learning cannot go the whole way in the quest for 99.999% reliable autonomous driving. It can, and it will, and it will happen sooner than people think.

For the Go players out there, this is the YouTube stream with commentary provided by Michael Redmond, the only Western player with the 9-dan professional rank (the highest rank in Go).

For historical reference, TSLA stands at $206.87 per share right now. I can't wait to see what the price will be when cars capable of completely autonomous driving are on the road.
 
Last edited:
This is off-topic, but it's the only place where people will read it, so here it goes: DeepMind's AlphaGo software defeated the best human Go player alive, Lee Sedol, in a historic game. Go is an ancient Chinese game (2500 years old) vastly more complex than chess. For perspective, the number of possible Go positions is larger than the number of atoms in the Universe.

Just a few years ago it was generally believed that it would take decades for AI to defeat the best human player at Go, with many believing it was never going to happen. It happened this morning.

This is relevant for self-driving cars. I will go on record and say that Elon is mistaken in his belief that machine learning cannot go the whole way in the quest of 99.999% reliable autonomous driving. It can, and it will, and it will happen sooner than people think.

You may be right that Elon's mistaken, but I'd say the Go analogy is apples<->oranges. Go has a few simple rules and (given enough memory and CPU) you can see every possible "thing" that can happen. Driving on streets around the world has a billion (or zero, depending on your perspective) rules once you factor in peoples' idiosyncrasies (stupidity), weather, road condition, large game animals, and...

Totally different than solving go (though that is a milestone achievement in itself).
 
You may be right that Elon's mistaken, but I'd say the Go analogy is apples<->oranges. Go has a few simple rules and (given enough memory and CPU) you can see every possible "thing" that can happen. Driving on streets around the world has a billion (or zero, depending on your perspective) rules once you factor in peoples' idiosyncrasies (stupidity) and weather, road conditions, large game animals, and...

Totally different than solving go (though that is a milestone achievement in itself).

The computer can totally NOT see "every possible thing" that can happen in a Go game. It is physically impossible to do due to the very large state space. As I said, there are more positions in Go than there are atoms in the Universe. Good luck with that search.

Of course, you're right that Go and driving are different problems, but the significance of this win rests on the fact that the methods AlphaGo used to get to this level are not specific to Go playing. This is a proof-of-concept only, and DeepMind (owned by Google, for those who don't know who they are) will make a big push into domains that have nothing to do with strategy games.

Driving is also not anywhere as complicated as Go. The main differentiating aspect is its real-time nature, but let's remember that most people can drive, whereas only a handful play at Lee Sedol's level. I don't want to push this too far, since again, you're right that the problems are different, but the state-of-the-art in AI today is moving towards general methods, and today's AI win is proof that this is going much better than most people expected.
 
what is going on with TSLA calls today? stock up 5.80, options barely moving across the board. Strange to see it so consistent. My Jan '17 180's (volume of 1...) have a lousy bid and ask.
It could be that shorts are building synthetic short positions if they are having a hard time finding shares. They have to sell calls to do this and that increases supply and drives the price down. The good news is when they start covering that will go the other direction for you.
 
Status
Not open for further replies.