Ask HN: Does ChatGPT gather your stylometry or typing cadence? | Hacker News Reader

Ask HN: Does ChatGPT gather your stylometry or typing cadence?

13 pointsvmoore10 comments

I was thinking the other day, ChatGPT asked for my phone number when signing up. I understand this, they want to reduce the amount of willy-nilly signups. But then my account is tied to my meat-space identity now, which means I am now probably training it with things like my stylometry[0] and perhaps even the way I type (typing cadence, or typing fingerprint). Does it gather things like that behind the scenes? Because if it is, I'm going to stop using it. That's a privacy violation. Anyone know if it attempts to fingerprint you using those methods?

[0] https://en.wikipedia.org/wiki/Stylometry

View on HN

Just so I understand: You're totally cool with ChatGPT being trained on everything you've ever written and also logging everything you ask, but stylometric analysis of that log is your bridge too far?

IMO - the problem is the inability to opt out of anything AI going forward.

We thought the singularity was when AI is smarter/developing faster than humans - but the singularity has already been passed ; you cannot opt out of anything AI-mining your every-thing.

You have to trust OpenAI follows their privacy policy. Otherwise, you shouldn't use their service.

'When you use our non-API consumer services ChatGPT or DALL-E, we may use the data you provide us to improve our models. You can switch off training in ChatGPT settings (under Data Controls) to turn off training for any conversations created while training is disabled or you can submit this form. Once you opt out, new conversations will not be used to train our models.'[0]

'When you use our Services, we collect Personal Information that is included in the input, file uploads, or feedback that you provide to our Services (“Content”).

...

We may automatically collect information about your use of the Services, such as the types of content that you view or engage with, the features you use and the actions you take, as well as your time zone, country, the dates and times of access, user agent and version, type of computer or mobile device, and your computer connection.'[1]

[0] https://help.openai.com/en/articles/5722486-how-your-data-is... [1] https://openai.com/policies/privacy-policy

What if there was a highly paid and trained governmental entity that did surprise on-premise inspections for claimed and/or legally required data privacy standards.

Wait until the NSA has a blackbox 'CARNIVORE' style AI black-box that they require to install at all major ISPs, just like they did in the 90s, 2000s, 2010s, etc...

Cisco was required to include backdoor accesss to NSA in all routers. ISPs, were required to install CARNIVORE boxes, and now I am convinced that all AI companies are forced to include NSA doors...

https://en.wikipedia.org/wiki/Room_641A

Same building as twitter was founded in, I believe...

https://en.wikipedia.org/wiki/Utah_Data_Center

Recall when some White-Hat hacked a bunch of home ISP cable-modems and patched them for a security hole...

it would be awesome if a blackhat/whitehat/greyhat would create black-list FW rules blocking all address space (ipv4/6) for addresses known to utah data center.

Stylometry is going to be baked into all LLM training without even intending to capture it. It is unlikely that it would bother to associate you in particular as a marker of style.

>>*It is unlikely that it would bother to associate you in particular as a marker of style.*

But it WILL be used to create a signature in searching for people / entities / systems / things/ in the future.... AI is the pattern recognition system as mirrored in human conscious - that pattern recog is what makes our species so intelligent, but we cannot accelerate our ability in the same way we can augment and accellerate that of an artificial, engineered pattern recognition engines we control, design, build and deploy.

prompt harvesting is going to be the golden goose of AI scraping of data to human identification.

What meta-data is captured in each prompt????

Username, UID, IP, GEOLoc, browser ID, blah blah blah - where is that data scraped to?

The killer AI is going to be in a supply-chain logistics efficiencies model.

Who is buying X from where and is delivered by/manufactured by whom?

The benefit in AI is going to be in supply chains - the negative impact is going to be survelleince.

AI will be able to figure out inefficiencies - and if you can have / own the layer of shipping data - you have an incredible ability to build other services on top of that insight.

The biggest company that I can see (as a startup) is ImportYeti.com

But if youre a MAERSK, MATTISON, CHINA - you'd be the perfect market for AI pattern recog which was already being done, but here is the next level ;

if you have the ability to analize the sentiments of purchasers, from the perspective of Amazon, and then inform the supply chains upstream, you then become a driver of what that vendor can charge... so AI supply chain management puts the power into the middle-man.

We can already do this at scale to about the same extent that an AI model is likely to be able to do imo. The data scraping is more likely the limiting factor. Looking for content similarities, not style, is probably more of the AI superpower

ChatGPT doesn't gather any keystroke analysis data, fwiw.

Assume they are doing whatever they can with the data they're gathering.

Reply on news.ycombinator.com