Comment by aurareturn
1 year ago
Just opened Claude app on Mac and saw a popup asking me if it's ok to train on my chats. It's on by default. Unchecked it.
I think Claude saw that OpenAI was reaping too much benefit from this so they decided to do it too.
Also your chats will now be stored for 5 years.
I used to not care about this stuff but with the way this administration is going about things, I suddenly care very much about it.
Trusting companies more than the government always feels strange. It's something I can't grasp.
28 replies →
It’s more that five years worth of peoples most personal conversations is an absolute treasure trove and makes their systems much more inviting for hackers and yes governments.
The part that irks me is that this includes people who are literally paying for the service.
And there's no way to opt-out of the training, without agreeing to the 5 year retention. Anthropic has slipped so far and fast from its objective of being the ethical AI company.
> If you do not choose to provide your data for model training, you’ll continue with our existing 30-day data retention period.
https://www.anthropic.com/news/updates-to-our-consumer-terms
My work just signed to an enterprise agreement with anthropic. I just checked, and "Your data will not be trained on or used to improve the product. Code is stored to personalize your experience. Applies to all team members."
Given how competitive Claude has been with ChatGPT models without training on users I'm curious how useful OpenAI could have found it.
I hope they didn't vibe code the popup, that could be bad if it didn't actually work.
We should be able to train on foundation model outputs.
These bastard companies pirated the world's data, then they train on our personal data. But they have the gall to say we can't save their model's inputs and outputs and distill their models.
I am pretty sure they try to do it all the time between themselves. Most of the real sauce in AI coding comes from reinforcement learning, usually done by armies of third world outsourced developers tediously doing all kinds of tasks with instructions to detail their reasoning behind each chance. Things like: "to run this python test in a docker container with the python image we need to install the python package xyz, but then, as it has some native code, we also need to install build-essential..."
While those developers are not well paid (usually around 30/40 USD hour, no benefits), you need a lot of then, so, it is a big temptation to create also as much synthetic data sets from your more capable competitor.
Given the fact that AI companies have this Jihad zeal to achieve their goals no matter what (like, fuck copyright, fuck the environment, etc, etc), it would be naive to believe they don't at least try to do it.
And even if they don't do it directly, their outsourced developers will do it indirectly by using AI to help with their tasks.
> those developers are not well paid (usually around 30/40 USD hour, no benefits)
$40/hour for a full time would put you just over the median household income for the US.
I suspect this provides quite a good living for their family and the devs doing the work feel like they’re well-paid.
1 reply →
You can, they might not like it but there's no legal basis saying you can't.
violating terms and conditions can be sufficient to be at least charged with computer abuse and fraud.
2 replies →