Musk should stick to building cars that aren't constantly recalled and like the Cyber Truck not have faulty issues before they are released. His ongoing "I'm smarter than the rest of you" will eventually kill his Tesla success.
Getting basic material science correct would be start. Rusty Cyber Truck
I’m a massive Tesla / Musk fan, but as the article points out, he clearly and stupidly jumps to conclusions which he has to roll back on. Something he really should be smart enough to recognise and take a moment to properly digest. He does even admit he’s pretty dumb at times, so he should learn from it!
Musk and Trump are both examples of men who buy businesses started by smart people, or buy smart people to run their businesses. Then promote the myth that they themselves are brilliant. Mediocre brains with a nose for business + affluent parents to prime the pump.
Musk and Trump are just separate cheeks of the same ass - Kardashian size.
Musk should understand that LLMs (Large Language Models) do not require personal data for training. They are trained on gigabytes of curated data, guided by human engineers who establish acceptable and unacceptable outputs before release. Using personal data for training would actually harm LLM.
LLMs process queries using tokens, vectors, and embeddings to generate responses without storing user data, which is a privacy-conscious approach. Apple's method does not raise privacy concerns.
Regarding ChatGPT, no data is transmitted without explicit user consent, eliminating privacy issues.
On-device processing, such as that used by "Siri" with basic LLM updates, is straightforward and secure. Since the data remains on the device, processing vectors, tokens, and embeddings should not be a concern. These computations are meaningless to other without the user data from which they were derived, even if they leave the device.
He’s POed b/c Apple found a way to kneecap X.AI’s monetization model.
Bet MS/OpenAI pay Apple like Google does for search b/c data sent to ChatGPT 4 will help train their models.
Data from Apple devices will not be used to train OpenAI’s models. If it did, then Musk would have a point. You don’t want queries and supporting information that pertains to you’d business secrets being added to a 3rd-party commercial LLM model. And so Apple’s deal with OpenAI explicitly excludes that capability.
He’s POed b/c Apple found a way to kneecap X.AI’s monetization model.
Bet MS/OpenAI pay Apple like Google does for search b/c data sent to ChatGPT 4 will help train their models.
Data from Apple devices will not be used to train OpenAI’s models. If it did, then Musk would have a point. You don’t want queries and supporting information that pertains to you’d business secrets being added to a 3rd-party commercial LLM model. And so Apple’s deal with OpenAI explicitly excludes that capability.
Does it include anonymized and aggregated user data collections, which Apple does not consider to be "user data" anymore and is not covered by Apple privacy policies, or just personally identifiable data? I don't remember that question being answered, but perhaps it has been. To be clear I would have no issue with the former being shared if it improves the overall results.
He’s POed b/c Apple found a way to kneecap X.AI’s monetization model.
Bet MS/OpenAI pay Apple like Google does for search b/c data sent to ChatGPT 4 will help train their models.
Data from Apple devices will not be used to train OpenAI’s models. If it did, then Musk would have a point. You don’t want queries and supporting information that pertains to you’d business secrets being added to a 3rd-party commercial LLM model. And so Apple’s deal with OpenAI explicitly excludes that capability.
Does it include anonymized and aggregated user data collections, which Apple does not consider to be "user data" anymore and is not covered by Apple privacy policies, or just personally identifiable data? I don't remember that question being answered, but perhaps it has been. To be clear I would have no issue with the former being shared if it improves the overall results.
I'm not sure radarthekat understands what omasou is saying. As soon as you say you want to use ChatGPT ten OpenAI will have access to those queries and requests to use to train their system, regardless of how well Apple anonymizes the originator. OpenAI could be paying Apple to be the default just as Alphabet pays Google for internet search. I think we're on the same page here.
He’s POed b/c Apple found a way to kneecap X.AI’s monetization model.
Bet MS/OpenAI pay Apple like Google does for search b/c data sent to ChatGPT 4 will help train their models.
Data from Apple devices will not be used to train OpenAI’s models. If it did, then Musk would have a point. You don’t want queries and supporting information that pertains to you’d business secrets being added to a 3rd-party commercial LLM model. And so Apple’s deal with OpenAI explicitly excludes that capability.
Does it include anonymized and aggregated user data collections, which Apple does not consider to be "user data" anymore and is not covered by Apple privacy policies, or just personally identifiable data? I don't remember that question being answered, but perhaps it has been. To be clear I would have no issue with the former being shared if it improves the overall results.
I'm not sure radarthekat understands what omasou is saying. As soon as you say you want to use ChatGPT ten OpenAI will have access to those queries and requests to use to train their system, regardless of how well Apple anonymizes the originator. OpenAI could be paying Apple to be the default just as Alphabet pays Google for internet search. I think we're on the same page here.
Actually no, Apple has locked OpenAI into this agreement from a legal standpoint. *By Law* OpenAI can't keep any data sent to it by Apple devices. No files, no queries, no IP. Nothing. It is all tossed.
Musk and Trump are both examples of men who buy businesses started by smart people, or buy smart people to run their businesses. Then promote the myth that they themselves are brilliant. Mediocre brains with a nose for business + affluent parents to prime the pump.
Except Trump’s “nose for business” is itself a myth. Trump is like a living Ponzi scheme, always needing cash flow from new scams to stay ahead of the crash-and-burn of the previous “businesses.” This is a man who bankrupted casinos, businesses where the odds are always in favor of the house, where the customers literally pay for the entertainment of giving the business money. It’s been calculated that Trump would be much wealthier now if he’d done nothing other than put the money his father gave him into an index fund.
Musk had a nose for business, but his ego has overtaken it. Musk never intended to buy Twitter, but his smack talk ran into finance law that wouldn’t let him back out of a threat, and Jack Dorsey called his bluff and took his money.
Rather than acknowledge the humiliation, he doubled down (now that is a Trump move) and insisted only he could fix twitter. In the process, he has totally destroyed Twitter while embracing the conspiratorial Trump fan base, which is an idiotic move for his other business, Tesla, as Trump & Co are committed to the destruction of the EV market. The two may be more alike now, but they weren’t always so. Musk’s brand of narcissism used to actually make money.
Comments
Tell us you don't understand anything that was presented by not actually saying it.
And you 'trust Enron Musk' ouch!
Musk should understand that LLMs (Large Language Models) do not require personal data for training. They are trained on gigabytes of curated data, guided by human engineers who establish acceptable and unacceptable outputs before release. Using personal data for training would actually harm LLM.
LLMs process queries using tokens, vectors, and embeddings to generate responses without storing user data, which is a privacy-conscious approach. Apple's method does not raise privacy concerns.
Regarding ChatGPT, no data is transmitted without explicit user consent, eliminating privacy issues.
On-device processing, such as that used by "Siri" with basic LLM updates, is straightforward and secure. Since the data remains on the device, processing vectors, tokens, and embeddings should not be a concern. These computations are meaningless to other without the user data from which they were derived, even if they leave the device.
..."without prejudice", so he can reopen the lawsuit later, it he choose, though it never made any sense to begin with.