‘Artificial Intelligence’ is trained on wrong things

THE FREELANCE is grateful to Puerto Rican artist Karla Ortiz for alerting us to a change in the "privacy policy" at "cloud" file hosting provider DropBox. This means that the machine-learning system is Hoovering up things that we know to be false, and rearranging them into new falsehoods.

To quote the statement dated 24 October:

At this time, we're partnered with one third-party AI partner, OpenAI. Open AI is an artificial intelligence research organization that develops cutting-edge language models and advanced AI technologies. Your data is never used to train their internal models, and is deleted from OpenAI's servers within 30 days. Read OpenAI's full privacy policy.

What exactly this means depends on the interpretation of "internal models". But it does raise one interesting point about so-called "artificial intelligence".

The Freelance is aware of one publishing group that bases its entire operation on DropBox. Terabytes of data include all the layouts, all the edited copy that went into them - and all the raw articles, complete with mis-spelled names, mis-read numbers and statements that turned out on examination not to be true.

So OpenAI is training something on a catalogue of errors (as well as on the corrected versions. Not to speak of the potentially defamatory statements that got cut...

We'll return to this if we work out what that something is.

Caveat lector - reader, beware!

Also, DropBox seems to have assumed permission to do this, unless users opt out. So there will be many cases in which is is, on the face of it, in breach of copyright.

© Mike Holderness

'Artificial Intelligence' our coverage to date