Plug & Play: Speaking on Generative AI

Talking Privacy, Data Ownership, Biases, and Worker Employment

A common theme in many conversations at most conferences today involves meeting a stranger, exchanging a few words about your excitement regarding LLMs, donning a thought-leader hat, and then posing a question like:

So, what do you think about Data Privacy with OpenAI?

Recently, I donned my thought-leader hat and seized the opportunity to discuss Privacy, Data Ownership, and other similar topics.

We don't have data privacy - it's time to get used to it...

Seriously, this is 2023. Facebook, Google and Apple know where we shop, who we meet, and where we work. We've all made trade-offs concerning privacy, and largely, privacy has lost.

Sure, on a day-to-day basis we don’t see anything that would be alarming. Facebook does NOT suggest you new dating partners. Google does NOT suggest me a new work place despite knowing how many hours I work (“become an entrepreneur, they said” ??).

However, just because we don't see it today doesn't mean it's not possible. The reality for these organizations is that their biggest barrier to privacy violations isn't so much rules or regulations, but rather self-imposed governance to avoid potential backfires. When you have a billion users, almost any algorithmic change is bound to upset someone. Even if the error rate is <1%. Suggesting a new dating partner to 1 billion facebook users could result in 20 million of those suggestions observed by the current partner of a target facebook user…

And as we move into the AI age, it is going to be much easier to segment users to avoid such errors. So, like it or not, it is very likely that whatever unspoken contract we as a society had with big tech, will no longer exist.

The Real New Problem: Who Owns the Data?

For more than a decade now we’ve contributed to Wikipedia, StackOverflow, GitHub, etc - imagining that this is our social content and reputation. Then suddenly, it turns out that we weren't so much contributing to those platforms and building our reputations as we were feeding a third-party AI Large Language Model. Talk about feeling of “being used” the morning after…

Now, I can get over my small social profiles not becoming relevant, but we’re talking about multi-billion dollar content sharing companies that largely have become irrelevant overnight. Why would I spend time searching StackOverflow today for code examples when ChatGPT can just write me an answer?

And if we now all recognize that the incentive to build social “public” profiles has disappeared on current platforms, then what kind of incentives are we going to see on future versions of StackOverflow to incentivize people to contribute their knowledge? Unless, of course, we imagine a reality in which the new input into LLMs is the output of existing LLMs.

The world is biased. Deal with it!

So much conversations about biased AI that sometimes I wonder if maybe the people who bring this up grew up on a different planet. Newspapers, TV talk shows, Radio, education curriculum - all are highly biased. Show me an unbiased source of information and I will show you a dozen times how it is biased. AI is not any different.

Perhaps the bigger question is: what are the new implications due to the size of the potential audience? ChatGPT is the fastest growing service. It has grown faster than Google and Facebook. The exposure of its biases is unprecedented. Should we be worried?

I don’t think so! The reality of things is that we’re about to see hundreds of new foundational models being built. I admire Sam Altman’s confidence, but there are very few protections in tech against competition. To use Google and Facebook as examples, brand and network effects have been the ways to protect those companies from competition. And, when looked through that prism, pretty much every speech Sam has given since ChatGPT’s release has been an attempt for OpenAI to preserve its Brand. That’s great, but there are no network effects - not yet, anyways. Give me a better product and I will use that.

Employment, there will be more of it, not less

I really get tired explaining this over and over. In our business people also ask me whether I believe with AI there will be less data developers, and I will always tell them one thing:

Historically, when tasks become easier due to technological advancements, more people engage in that work, not fewer.

Over a hundred years ago, we, as people, introduced technology and automation into the Ford factory. What was the result? Do we see less people working on cars?

No, we have more factories, more cars manufactured, and more people cumulatively working across all those factories.

Yes, the nature of the work changes. However, technological progress creates more jobs, not fewer.

If a year ago you needed to hire a native English speaker to do good content copywriting, now you can hire people from South Asia, and the quality of their output, when supported by AI, will be just as good as that of a native English speaker. More people will be doing the job.

Something to think about… Happy Friday!

About the Author

In my former life I was a Data Scientist. I studied Computer Science and Economics for undergrad and then Statistics in graduate school, ending up at MiT, and, among other places, Looker (joined in 2014), a business intelligence company, which Google acquired for its Google Cloud in 2019.