34 comments

[ 1.8 ms ] story [ 5.4 ms ] thread
Using data from failed businesses as training data: what could go wrong?
Maybe we can help businesses fail faster
I would love to see a rider from the courts mandating a significant third party test by privacy researchers to see if the deidentified data can be reconstructed from models post training.
[delayed]
In theory, a buyer could always execute that strategy after the sale, so it's essentially "priced in". By keeping the auction exclusive, the seller also keeps the possibility that the data is worth more to someone as an exclusive asset.
It's high time we get signed warrants that upon bankruptcy all data will be destroyed so it won't be used to make good either the creditors or the shareholders of the company that went under. I can't imagine any good outcome here, and change of control should result in an automatic withdrawal of any consent, no matter what assurances of privacy are given by the buyer.
Worth a try, but just see how hard it is to get mining companies to clean up literal physical superfund sites.

How do you even prove all data was destroyed, let alone actually do it?

> upon bankruptcy all data will be destroyed

So, PBS should lose all their archives? https://news.ycombinator.com/item?id=49333344

There is always a way to interpret a comment in the worst possible way. I'm pretty sure you understood what I wrote.
There's a clear and significant difference between content that a company created itself, and the data they collected from their customers.
(comment deleted)
Anyone see Google's intended use of the collected data spelled out in the "Primary Purpose for Collection and Use of Data" in the Spirit Privacy Agreement? Asking for 1.2 billion Spirit passengers. https://content.spirit.com/Shared/en-us/Documents/Privacy_Po...
on twitter I read data will be fully anonymized and scrubbed for PII before handover by a third party.

supposedly google intended purpose is to train LLM's for enterprise/office work etc

For the 1.2B passengers: the transfer doesn’t contain passenger data, just internal operations data

Eg. Airline code, marketing materials, pricing histories, etc

Also, to answer the question:

> Affiliates and Acquisitions. We may share information with any corporate affiliates (e.g., parent company, sister companies, subsidiaries, joint ventures, or other companies under common control). If another company acquires, or plans to acquire, our company, business, or our assets, we will also share information with that company, including at the negotiation stage.

So Google is training AI to run businesses that go bankrupt now? That's not very smart.
(comment deleted)
> The data does not contain any personally identifiable information, according to a court filing by PJT Partners vice president Dylan Friesner, an investment banker for Spirit.

> The data will be "deidentified" — meaning that it won't be associated with individual people and the buyer agrees not to attempt to re-identify the users.

Sure, 100 million emails and 500 million chats totally won’t have any mention of customers. No employee has ever shared PII on email or chat. Their redaction is guaranteed to work correctly. Nothing to see here.

Doesn't matter. Same as Facebook or any other data hoarder. They use the data to manipulate society. This has always been the problem with personal data as the Red Herring.
USA: Hyperdetailed electronic information about you does not belong to you. EU: Well, it sorta does, altho we're still working out the kinks.
I was wondering if this is valuable for the mineable PII or for the AI training data? Maybe boyh.
Talking like a middle management drone from a failed corporation is perfectly on brand for Google AI.
Sad that an affordable airline had to dissolve from increased fuel costs due to America and Israel's wars. Google isn't the right company to capitalize on their data but now their competitors don't have it. This is what happens with Big Tech goes unchecked and their AI can swallow an entire airline.
The issue wasn't the fuel costs - it was their inability to adjust in time or effectively.
title can be viewed as googles entire strategic planning objectives realised.

"google Wins!, Bankruptcy Auction"