Your Data, Their Rules?

Data.

That’s primarily what good generative AI is all about, right? If you have enough good quality, accurate, structured data, you’re already well ahead of the AI game. Yet, there is a lot that we - collectively as an industry - haven’t figured out.

Take Bluesky, for instance.

To use or not to use user data for GenAI training

Earlier this month, Bluesky, a social networking platform, announced that they won’t mine user data to train their Generative AI models. However, that didn’t stop others from doing so, considering it was publicly available data.

Reddit, X (formerly Twitter) and Meta are all still evolving their policies around this, which is, really, a euphemism for already training their models with user data. So far, the underlying philosophy seems to be that it is enough to give users the choice of opting out.

If you had user data in your organization - employee information and internal forum discussions, for example - how would you treat it? I’m very curious to know!

Do customers care?

At the other end of the spectrum is the challenge around customer awareness and consent. Bain and Company has some interesting findings this week.

• 71% of online shoppers don’t notice GenAI even if their retailer uses it

• 41% are comfortable using GenAI from a brand they trust

• They are also happy to give personal data for GenAI training in exchange for personalization

Well, what then needs to be the legal and ethical boundaries of such usage? The answer seems to be: To each organization, their own.

Speaking of ethics.

Digital twins of you and me

Stanford PhD candidate, Joon Sung Park, last week published a new paper on generative agent simulations. Essentially, they brought 1000 people to interview with AI for two hours. Based on this data, they created simulation agents, which replicated “each participant’s values and preferences with stunning accuracy.”

This is unlike any previous “AI is taking over the world” predictions! This isn’t about AI doing the tasks we can do. This advancement is in AI understanding and replicating our thoughts, responses, and value systems - aspects foundational to human decision-making.

So, it might not be too far away by the time we have digital twins of you and me - changing everything about knowledge work as we know it.

When this happens, what should we be worried about? Some answers in the MIT Technology Review here.

While all of the above challenges are important to think about, I’m not sure they’re urgent yet.

Until then, GenAI can safely and harmlessly do various valuable things. Like how it accelerated RFP responses for one of our UAE-based clients.

Accelerating proposals; multiplying market reach

Last week, I’d promised to share a case study with you. Here it is.

As an ambitious pilot to their organization-wide GenAI strategy, a UAE-based event management company asked us to automate their proposal creation process. Here’s the story about how we did that to straight up slash 20% of the proposal-making time.

Read case study.

Until next time, stay tuned.

Best,
Anshuman Pandey

P.S. Gemini Experimental 1121 is now available on Tune Chat