by rootlocus on 12/17/15, 10:45 AM
Both articles were written for a wide audience, which includes non technical people. Tesla's image and technology were bashed by someone claiming to know everything there is to know in the field. Non technical people see 'a limited demo on a known stretch of road' and assume Tesla engineers are idiots who wasted two years when they could have employed this awesome dude to think outside the box (that's always cool) and magically implement machine learning to solve all the problems.
'Learning by doing' as presented by Hotz is all fine and dandy when you drive on a motorway with a very limited set of road signs, no pedestrians, no traffic lights, etc. Hotz also assumes people learn to drive by watching others, which is only partially correct. Traffic rules are not learned from experience. Low light conditions, fog, handling the car on snow and wet surfaces and numerous other factors have to be trained (if they can be taught through training). And after all that work, the algorithm must be tested extensively before it can be shipped into production. The reporter omits asking any questions regarding these aspects.
I don't see the problem with Tesla highlighting these aspects to people who may be fooled into thinking one boy genius would solve all the problems with self driving cars in a couple of years, while they (Tesla) are wasting time.
by m-i-l on 12/17/15, 9:54 AM
Seems a reasonable response to me. After reading the original article I was left with the impression that Tesla had outsourced the development of their autonomous driving technology, and this changes that impression.
And I have to agree that getting such a complex and safety critical system production ready is beyond the reach of a single individual, no matter how talented they are. Its like a single individual might be able to build a rocket which can reach the edge of space, but not extend it to take human passengers into orbit (and back again safely).
And I'm not sure Tesla should be publicly congratulating him. While I do think it is very impressive what he has done, there is an element of reckless endangerment if he is indeed testing this on busy public roads after only a few hours of development.
If I were George, I would I'd use this project to either (i) get a job at one of the companies with significant resources so I could take the work further (but his short stints at a number of technology companies suggests he doesn't work well for others), or (ii) get funding so I could build a much bigger team to take the project further.
by amatic on 12/17/15, 9:10 AM
They seem sort of defensive and needy. "Oh, the didn't
reaaally make a better car then we did, we had that two years ago, and ours is so much better and ... ". Of course it is, the guy made it in his garage.
The classy thing to do would be congratulating the guy and encouraging even more open experimentation. I mean, they already are open-sourcing a lot of their work.
This statement doesn't seem in line with that.
by tonylemesmer on 12/17/15, 9:27 AM
Would you buy a system like that from a guy? I wouldn't and I'd feel uncomfortable, just like I am around poor drivers, if I knowingly encountered one on the roads.
Its a neat project and a serious amount of effort, well done. But come on, as others have said - a setback to the entire industry would be the outcome if he had a crash - doubly so, since he's publicly stated he intends to use the same Israeli component that everyone else uses. Tesla should try and distance themselves from this type of enterprise.
The way it is portrayed in the article geohot seems to have a slightly flippant approach. Not sure if that's fair to him or not. ("Dude the first time it worked was this morning", "Don't touch that or we die.")
by jzwinck on 12/17/15, 9:08 AM
This piece is really well written in an accessible style. It's even more clear than the original article, and does a great job of sounding genuinely kind to all parties (except maybe the journalist, who will be prepared for it).
Well done, Tesla. Well done for making awesome cars, responding nearly instantly to something you barely needed to bother with, and not letting your legal team rewrite everything you say into oblivion.
by dutchbrit on 12/17/15, 8:52 AM
I applaud George Hotz, will be cool to see what the end product will be like. Tesla seems pretty sure of themselves that he will fail, but I hope he'll prove them wrong.
by caractacus on 12/17/15, 9:28 AM
I am more surprised that Tesla did not make a point of saying 'Elon Musk did not encourage George Hotz to play fast and loose on San Francisco highways to win a bet' and 'there is no formal financial relationship between Tesla and geohot'.
by carbocation on 12/17/15, 11:33 AM
In addition to the substantive points raised by other commenters, the context is notable. The Geohot story was written by Ashley Vance, author of an unauthorized Elon Musk biography and with whom Musk and Tesla have had prior disagreements. That may influence the degree to which a rebuttal may be felt necessary, and the tone chosen for the rebuttal.
by rl3 on 12/17/15, 9:09 AM
>
We should also clarify that Tesla’s autopilot system was designed and developed in-house.I thought it was quite odd the article was claiming otherwise. Remarketing an off-the-shelf solution for a feature that complex didn't seem like Tesla.
by cconcepts on 12/17/15, 12:04 PM
I just read the associated article [1] and somehow, despite the fact that I've understood that neural networks and deep learning have been coming for years, it has caused some kind of explosion in my brain - like this is all coming true quicker than I thought. What struck me in particular was the notion that physical work has been revolutionised and now thought-based work will be revolutionised - removing from large chunks of humanity that must fundamental of needs - something to frantically do all day.
Can anyone recommend online courses in deep learning, neural networks and machine vision? I have kids that I need to prepare for a world that I just realised I hardly understand.
[1] http://www.bloomberg.com/features/2015-george-hotz-self-driv...
by Angostura on 12/17/15, 9:42 AM
An unusually poor PR mis-step from Tesla with an ill-judged tone. It is equally possible to make exactly the same points by congratulating the lone hacker for what he has managed to achieve by himself and then lead the reader through the extraordinary areas of innovation that Tesla has achieved in housed and the challenges that Hotz may yet find himself facing.
by revelation on 12/17/15, 9:45 AM
I'm usually supportive of Tesla in their righteous crusade, but even they have to realize that they can't be using MobilEye forever, right?
It's a stopgap solution to get a basic system up (and that's what the Autopilot is, this kind of functionality was in cars years ago), but if they want to move into fully autonomous driving, they will need their own vision solution.
by dayaz36 on 12/17/15, 9:53 AM
It's so obvious MobileEye asked them to right this. The original article had Elon quoted saying that he would give him a multi million dollar bonus once they drop MobileEye. Unless that quote was falsely attributed to Elon I'm pretty sure the original article was correct and MobileEye was bummed.
https://i.imgur.com/4gdCzKj.pngby jsjohnst on 12/17/15, 9:54 AM
It's a terribly sourced article, so I personally don't blame them for responding in the way they did. He's an extremely gifted developer, but from the title down (he wasn't the first one to "hack" the iPhone) it's factually wrong on many levels. This reeks of a reporter who's either a fan boy or has an axe to grind with Tesla.
by untilHellbanned on 12/17/15, 1:48 PM
Why respond? You really think Google [or self-respecting company] would dignify some random person one-upping their self-driving initiative with an official blog post? Seriously? Elon Musk's hair-transplant-insecurity is a strong one.
by AKifer on 12/17/15, 10:24 AM
Anything can be said and PRed, the only thing that matters is results, and only a real world contest will prove who's who. But I'm still wondering about the objectivity of the original article, does 2k locs written by geohot really beats the systems existing today? Or is that kind of a high level code in top of something else not mentioned in the article, I'd rather think of that second possibility.
by Lordarminius on 12/17/15, 11:47 PM
What gets me puzzled about this article is an angle pointed out by a journalist (I forget whom now): how on earth do you rationalize "correcting" an article written by someone else? It is unusual and borderline impolite.
by hanlec on 12/17/15, 10:37 AM
Who signs this post?
by damniatx on 12/17/15, 9:02 AM
I support geohots
