Having worked with the Twitter API, I think he's rather underestimating how much volume this is gonna wind up being, the amount of processing it'll entail, and the difficulties of determining "this is a bot" from just the Tweet data.
Source: Former employee of one of the few companies who got the "push" firehose and not just the API.
"1 terabyte per day of data [...] - that is how much is produced by the firehose of 250 million tweets a day" [0]
i.e., ~4kb per tweet. So, 2ish Tb/day of text + metadata at current rates?
[0] https://techcrunch.com/2011/12/07/datasift-historical-tweets...
[0] https://techcrunch.com/2010/09/17/twitter-seeing-6-billion-a...
I don't see how he thinks tweet content alone will be enough to verify the 5% of monthly monetizable users
The other poor indicator was his tweet about sampling _100_ of his own replies to detect bots. That's a horribly small sample and horrible selection bias
The firehose is terabytes a day, add storage + indexes and it's pretty far from "fitting on an USB stick" in order to analyze.
Maybe from now on we should use statistics as the epitome of difficult intellectual activity instead of “rocket science”.
His modus operandi is taking quick decisions and backtracking if he is wrong.
Actually it's not only his modus operandi, SpaceX according to NASA evaluation is also working that way.
It's basically depth first search vs. breath first search.
Dr. Daniel Rasky, a NASA Senior scientist
https://youtu.be/MxIiiwD9C0E?t=1464 :
> I think there are some things and kind of going back to in some ways back to the Apollo days if you take a look at kind of the environment that NASA had in Apollo it was kind of what Elon has with his young engineers. you had you know people trying things and in what works and what doesn't work I think over the years we've gotten into a very risk-averse profile and we're I call it full matrix engineering that okay that all concerns that anyone can raise have to be chased down a hundred percent before you can make any decision. Okay, so yeah like a full matrix of all the things that may be pertinent to a certain decision and you have to fill that entire array before you can make a decision to go forward and there's times when that makes sense but it doesn't make sense for everything we do. SpaceX uses a sparse matrix okay 51% okay when do when we get to 51% okay enough we're gonna make a decision and then we're going to try it and move on.
That is the modus operandi with products/services, it's not anything new anyways. It has been named in a bunch of different ways ranging from "agile" , "move fast break things" etc. also less charmingly "Microsoft vaporware products" or "Google graveyard", "Steve Jobs' bluffs"
They all indicate the same strategy: ship something and get feedback, if feedback is ultra-negative then you can always kill it.
However when legal stuff is signed, and especially legal stuff of this magnitude then the only way in which you leave yourself room to backtrack is if there is a clause which lets you do so in the legal document. When legal stuff is signed is not in your hands anymore, you have to treat it as being in the courts hands. Because that is what happens if you backtrack.
From what I understand he can legally backtrack from the deal, but he would have to pay some billion dollars penalty. Not cheap by any means, but he can afford that. And having billion dollars at stake you can try to look for a lot of legal reasons to not pay that billion.
But my guess is that he genuinely thinks that 5% number is not accurate based on his twitter feed comments. His feed comments is mostly crypto scammers. Some independent people also claim that 50% of his followers are fake.
Twitter claim of 5% is very specific and probably accurate in a legal sense. But if you have tens of billions at stake and a deal that looks like a mistake it's entirely reasonable to dig deeper.
Isn't a bit premature to declare it a successful revolution? Cases against Tesla autopilot are mounting, there is a reason why "move fast and break things" isn't used in fields where the people's lives are at stake. It's the reason why Theranos imploded. Who could have possibly predicted what could have happened if US Govt let Theranos run, maybe they'd have really developed the product in the following iteration of the company?
> From what I understand he can legally backtrack from the deal, but he would have to pay some billion dollars penalty
That is only the case if the FTC or the US. Govt. blocks the deal because of any reason.
With SpaceX it is certainly a successful revolution. There is whole bunch of new startups that would not exists without SpaceX. Most of them will die, but some will not.
It is also quite clear that Tesla already achieved it's goal of accelerating transition to sustainable energy. Volkswagen shift is certainly linked to Tesla. And Volkswagen actually committed fraud on emissions that probably led to death of many people.
Situation with Theranos and Tesla is completely not comparable. Autopilot works fine. You are safer with autopilot activated on a highway than you are driving in a city per kilometer driven. Whether you are safer on a highway with or without autopilot is debatable and we don't have good enough data to settle it. So, Autopilot is not prefect by any means, arguably not market leading solution in the category of drive assist, but it works fine.
The bigger issue is with Full Self Driving. Tesla can drive you around in a city - so the way they spin what FSD means is almost satisfied. The issue is that it is extremely stressful experience and I'm 100% convinced that you are more likely to get in an accident when babysitting it vs. if you would drive by yourself and pay the same amount of attention.
But I think in 5 - 10 years they may be able to actually deliver proper working solution. They are making progress.
Apples and oranges. Volkswagen committed fraud on behalf of the consumer. In the endless war for quality of life as a consumer who has to spend 45k for a car I know that Volkswagen is conspiring on my behalf to sidestep environmental regulations and thus providing a vehicle which is more affordable, more durable with the tradeoff of being less green.
Same goes for oil companies, they are just in the business of extracting oil, they are the ones who take the blame for global warming whereas it's me burning it for quality of life purposes when refilling my Escalade or flying to Cabo for the weekend. I owe them one otherwise the nuts over at Extinction Rebellion or GreenPeace would attack my Escalade and cause trouble to airlines.
Tesla is the opposite of that, it's conspiring against the consumer by making a poorly refined car with a terrible design and lots of problems ranging from faulty Autopilot to sudden braking and unintended acceleration. They want to wash away all such disadvantages with a supposedly green car. But in the end if you understand anything about the industrial economy and power generation you get how Teslas are hardly green cars.
But toxic exhausts within a city can not be tolerated.
I don't want externalties being offloaded on me and the democratic rule of law is there to make sure that we balance it.
https://www.youtube.com/watch?v=MxIiiwD9C0E
Here is Elon answering questions of National Academy of Science members about Starship. He explains in detail why he pushed the team towards steel and away from carbon fibers:
https://youtu.be/da3iF2Np51A?t=951
Here you have another direct conversation with Elon:
https://youtu.be/t705r8ICkRw?t=172
Here is Elon discussing with Sundy Munro mass manufacturing and engineering.
Sandy Munro is a veteran engineer from General Motors supplier and directly at Ford. Sandy Munro is now running a consultancy company that tear downs cars and suggests improvements to wide range of clients from the car industry.
Also, Musk was doing a physics PhD in battery tech at Stanford when he dropped out to found Zip2.
There is some long running pattern here with getting cold feet on bad decisions ^^ .
It’s funny how quickly the left changed from loving him for turning the car industry on its head to hating him for taking away the left’s hate machine. Oh well.
Being accepted is not the important part of PhD work, no.
Either way though, I expect Musk, the richest man alive, can afford to pay a handful of senior engineers a couple weeks pay to store and run some analysis on the firehose. It's a lot of data, but at "fuck-you" level money it's certainly manageable.
Some accounts are a mix of human and "bot" updates--and you can get a very different impression of their nature depending on your sample.
https://twitter.com/JeffLindner1, for example, posts automatic alerts from the Harris County flood warning system (such as https://twitter.com/JeffLindner1/status/1529386618938920960). If you look at his account when it isn't pouring outside, you'll see human updates and retweets. If you look during a three-day downpour, you may see dozens of automatic tweets per human post.
NOTE: For the ones operating legitimately I mean.
Although I will say that UI hint is extremely subtle.
While I do understand your point about the difference, neither type should be measured for things like (e.g.) advertising revenue. Weatherbots don't buy goods or services.
If people aren't viewing those tweets or deriving value, they're consuming resources without adding much to the ecosystem. If people are viewing those tweets and deriving value from them, the bot is increasing ad revenue.
What a terrible cultural icon Musk has become.
We should all aspire to be fentanyl dropping, pregnant women torturing street trash.
We should all aspire to accomplish great things like AOC when she ...uh she did ...um ...drop Adderall and start a podcast about putting on makeup! Yeah! That podcast alone employs 50,000 engineers and has saved or created 100,000,000 jobs a month, at least.
He’ll likely have to sample the data anyway before he can do any meaningful analysis, so the whole exercise is pointless.
Edit: to be clear, I'm saying this about what Elon said, not you.
The hypothesis is that sampling those buffers at scale should be enough to detect significant bot activity on the platform, without needing any additional metadata.
That may very well be the case, so why the snide?