I suspect a part of the problem is the they underestimated the amount of compute required.
The vehicles they’ve already shipped have about the equivalent tensor compute as an NVIDIA V100, and much less than that for older model cars.
The Bitter Lesson is that self driving AIs will probably need large scale more than anything. If we look at the capabilities of something like GPT-4V, it’s a good sign of what’s needed as the baseline minimum.
By extrapolating industry trends, I suspect that a future GPT 6 or 7 level AI with true video input might be able to be specialised to perform real time driving with human-level capability to parse complex situations such as construction sites or accident scenes.
That level of AI inference would require at least an order of magnitude more compute than what’s in any shipping Tesla, maybe more.
Elon promised something that will be possible, but not with the hardware of they day.
When Elon made his comments about “every Tesla shopping now has FSD hardware” he may have just been honestly optimistic. This was before GPT2, let alone the current state of the art. I can forgiven him for that, but not for doubling down on statements that are clearly no longer valid.
That is a remarkable, some might say delusional, level of generosity of spirit to demonstrate towards a pathological liar.
I suspect the actual problem is that Elon is a sociopath who simply does not care about honesty in any fashion and has become powerful and wealthy enough that he no longer needs to mask.
I make a conscious effort to avoid attributing the behaviour of others to malice, unless they're clearly a psychopath in the sense of literal mental illness.
People's attitude to Elon especially cracks me up. Before he acquired Twitter he could do no wrong. He was Tony Stark, and the Internet loved him. Then he fired thousands of overpaid Silicon Valley developers that were failing to properly engineer a chat platform that sends 1KB text messages, and suddenly he become the Antichrist and could do no good.
Elon has several strong personality traits that make him effective. One of these is being an technological optimist and setting aggressive timelines for his employees instead of letting things twist in the wind for decades.
E.g.: if Elon put his mind to it, he could probably get a working fusion power plant in a decade, instead of five decades like ITER. Oh sure, he might claim it'll take only three years, but that's just puffery or pep-talk.
Seriously though: if Elon went up on a stage tomorrow and claimed to have clean fusion energy solved in three years, delivered it in ten, then people like you would be on the Internet bemoaning that he's some sort of liar and a cheat because he didn't deliver as promised.
> People's attitude to Elon especially cracks me up. Before he acquired Twitter he could do no wrong. He was Tony Stark, and the Internet loved him.
In my circle of engineer friends, he was viewed as a bullshit artist long before Twitter. His fans seemed annoying, but were easy to ignore.
I feel like the perception changed when he started marketing FSD in a way that would kill people. The cave diver incident and his attitude toward Covid were big “what?” moments, and it’s accelerated downhill from there.
I think that after his bumbling of Twitter people started looking closely at his past and realized he put a ton of energy into sculpting an image detached from reality. This four part podcast does a great job looking behind the curtain: https://techwontsave.us/episode/189_elon_musk_unmasked_origi...
I'm not saying the Thailand cave diver tweets weren't... a bit too colourful for a CEO.
It's just that I wouldn't attribute that necessarily to an inherent malicious personality type.
You've got to keep in mind that at the time Elon was under enormous pressure because the Model 3 production numbers weren't up to expectations. He was working round-the-clock, sleeping in the office, and taking sleeping pills. Ambien especially is known to cause people to behave erratically.
There's a fundamental difference between leaders that are outright evil, harming others directly for their own gain, and leaders that just fucked up late at night while zonked out on medication.
Take the Sacklers for example. They knowingly, deliberately, purposefully marketed addictive painkillers to make money. They addicted hundreds of thousands of Americans to drugs, and directly killed thousands.
Elon tried to make self-driving as safe as possible. He's limited by technology, not morals.
> There's a fundamental difference between leaders that are outright evil, harming others directly for their own gain, and leaders that just fucked up late at night while zonked out on medication.
Generous to assume, given recent revelations, that prescription medication is the drug he was zonked out on while behaving erratically.
> Elon tried to make self-driving as safe as possible. He's limited by technology, not morals.
Arguably the end goal was the make billions of dollars for himself, and self-driving is the lever he found. If he was doing all this for a researcher's salary I'd have an easier time believing he was doing it for humanity's sake.
> Before he acquired Twitter he could do no wrong. He was Tony Stark, and the Internet loved him.
I used to think the same actually. But the more I looked into it, the more it turned out that it was not just thanks to his achievement. He very actively supressed every negative article or mention of him or Tesla. He apparently figured out way before the other that you can manipulate public opinion at scale through social medias. There is an army of Tesla influencers and there are numerous cases where Tesla / Elon went after critics.
Look at the story of the guy called 'Montana Skeptic' for example - a vocal and appreciated Tesla critic - in 2018 Elon reached out directly to his employer and threatened a lawsuit (without any basis for it of course).
Everyone has their limits. Elon is great at space, software and car engineering. I doubt he knows everything about batteries, ai or fusion - it wasn't much of a thing when he did his studies.
He's effective because he's used to the pain from his terrible childhood. He is almost bathing in it. Some of that pain he eventually unlashes into public. Thats when the hate towards him start.
> People's attitude to Elon especially cracks me up. Before he acquired Twitter he could do no wrong. He was Tony Stark, and the Internet loved him. Then he fired thousands of overpaid Silicon Valley developers that were failing to properly engineer a chat platform that sends 1KB text messages, and suddenly he become the Antichrist and could do no good.
Musk has been cratering his reputation with the garbage that comes out of his mouth or from his keyboard for YEARS. From "FSD. This year. For reals this time." to "It would be financially irresponsible to not own a Tesla by the end of 2019 and have a vehicle making you money as a robotaxi" to "Funding secured" to "Pedo diver".
Your attempt to frame his reputation challenges as what, a bunch of butt-hurt geeks annoyed that they or some of their peers were fired from Twitter, doesn't pass even the most trivial of sanity checks.
The vehicles they’ve already shipped have about the equivalent tensor compute as an NVIDIA V100, and much less than that for older model cars.
The Bitter Lesson is that self driving AIs will probably need large scale more than anything. If we look at the capabilities of something like GPT-4V, it’s a good sign of what’s needed as the baseline minimum.
By extrapolating industry trends, I suspect that a future GPT 6 or 7 level AI with true video input might be able to be specialised to perform real time driving with human-level capability to parse complex situations such as construction sites or accident scenes.
That level of AI inference would require at least an order of magnitude more compute than what’s in any shipping Tesla, maybe more.
Elon promised something that will be possible, but not with the hardware of they day.
When Elon made his comments about “every Tesla shopping now has FSD hardware” he may have just been honestly optimistic. This was before GPT2, let alone the current state of the art. I can forgiven him for that, but not for doubling down on statements that are clearly no longer valid.