Another day, another company using my data without asking….
So it is now LinkedIn’s turn to do what they like with user data, though I guess it is their data and I allowed the company to use my data – I would have appreciated a “Heads up” PRIOR to collecting what they did, that to me seems like a “oh crap” moment when someone has been informed that they should tell their user base, not a genuine care about the customer.
As technology and our business evolves, and the world of work changes, we remain committed to providing clarity about our practices and keeping you in control of the information you entrust with us.
LinkedIn
Though the LinkedIn blog post is nethier genuine and or user centric this is not exactly surprising and probably also not too much of an issue for me at least, my main concern being that the majority of information on LI is NOT informative or educational, it tends to be self serving or vendor xyzzy marketecture rubbish – so my question here is how accurate would LI training data be expected to be?
This is a significant issue as Gen AI is NOT smart, not smart in the slightest, not is not close the Artificial General Intelligence (AGI) the “holy grail” and it is actually not that useful when trained across too much and too varied (opinions and inaccurate) public data.
Platforms are really are scraping the bottom of the barrel now days with information to train their LLMs from: Reddit, Facebook, LinkedIn – as much good data as bad data exists on these platforms and X well thats a whole league beyond the aforementioned platforms.
But thinking further about this: Ensuring a Gen AI model even returns the right information and not bad information is complex because how it works, is not known and filtering of output is rudimentary and subject to prompt hacking, therefore it is not even possible to know yet what personal data could actually be maliciously retrieved because the platforms are not being open about what data is being used for training.
In addition, training on random Redditor comments is a recipe for a disaster, unfortunately we are already well on the way headfirst into this disaster because of Money, because to be the Gen AI platform that succeeds is going to net the victor a vast new revenue stream, probably this will be Microsoft and one other, any guesses?
Summary
Large Language models can be very useful when they are trained on “your” data and your companies data, because that information can be trusted to be accurate, this is the use case for Microsoft Copilot for 365 and GEN AI models that use your companies information – because here the output is going to be mostly correct as the inputs were already validated beforehand.
When an public LLM model exists simply because of MONEY then there will be problems, whether moral, ethical or financial, Facebook, X, Google are all ones to watch out for here as they have no clear way to monetize and are simply building for the sake of keep their investors happy – feel free to prove me wrong 🙂
As always feel free to reach out if you have any questions or would like to write to me here.
Leave a Reply