Hacker News Reader

Comment by Berniek | Hacker News Reader

Trust Crisis for AI? What after we see the board/CEO of one company fired/changed apparently for allegations of lies or manipulation that nobody is clear on? If Dropbox derives data from user data by scanning said data, then that "derived" data is no longer "users data" it is Dropbox data and can be shared. It may only be statistical in nature and not related directly to individual users but isn't that exactly what training data is? Isn't that how it works? That can be shared to train AI models can't it? Its not lies its hair splitting.

No its called unethical behavior and has become the norm for big tech.

View on HN

In fairness to Sam Altman and OpenAI, all credible reporting I've heard (and signifcantly Kara Swisher's work) has shown no issues of AI safety or lies in CEO communications to the Board, but rather broader concerns over the direction the CEO and board felt appropriate for OpenAI.

I've no dog in this hunt. I'm not partial to either Altman or OpenAI. And I have considerable reservations over where this Brave New World may be taking us. Or whether there is any credible option to stop riding this merry-go-round, no matter how unattractive the destination(s).

The DropBox behaviour described is only one in a long, long, long line of trust-violating practices by tech firms.

Too often the claims of unethical behavior are knee-jerk, baseless, and speculative. In the example the author cites, Dropbox only sends data to OpenAI when the user explicitly tells the app to engage an AI-related feature, for example to summarize a document. Yet the backlash seems to assume that they're scanning and uploading people's documents en masse, even though there is no evidence of this.

Unethical behavior definitely exists in AI companies. Personally, I'm agnostic about whether it's higher or lower than the base rate of unethical behavior in the general population. Anyway, if we're going to talk about bad behavior, we should use specific examples with cited evidence, not fear-mongering.

> Personally, I'm agnostic about whether it's higher or lower than the base rate of unethical behavior in the general population.

I don't think that AI companies are less trustworthy than our industry in general, but I absolutely think that our industry is far less trustworthy than the general population.

AI companies engaging in the wholesale harvesting of web data to train their models, though, was a particularly egregious violation of trust.

Reply on news.ycombinator.com