Most of the massive companies that work in A.I. have chosen to work in a closed source model, meaning they keep the source code to themselves.
The rest of us can play with their toys, but we can’t change them or see how they work.
Others, including Facebook parent company, Meta, have chosen to go “open source.” And they’re really going for it, saying earlier this month that they want to open source artificial general intelligence. (Non-techy explanation of AGI here.)
What’s the difference and what does it mean for our kids?
Open Source
Open source advocates believe everyone should have a chance to see — and participate in — the way A.I. is developed.
Open source is only the path to safety, they say. We need to keep any one entity from running away with the future of humanity.
One of A.I.’s founding fathers, Yann LeCun, is now the chief A.I. scientist at Meta and one of the world’s most vocal open source advocates.
He explained, “There's going to be a future, probably not so far away, where the vast majority, if not all, of our interactions with the digital world will be mediated by AI systems. You don't want those things to be under the control of a small number of companies in California.”
Confusingly, A.I. front runner OpenAI, changed its mind on the topic and now operates under a closed source system.
In 2015, Elon Musk started OpenAI (creator of ChatGPT) with many of the people who still run it today, and they named it OpenAI as an explicit nod to the open source model. Musk left, just three years later, over disagreements with everyone else and tweeted this last year:
So, why did OpenAI make the switch?
Closed Source
Closed source advocates believe A.I. would be dangerous — catastrophic even — if put in the wrong hands.
Closed source is the only path to safety, they say. We need to keep the bad guys from running away with the future of humanity.
When asked directly, a different OpenAI founder said this, “We were wrong. Flat out, we were wrong. If you believe, as we do, that at some point, AI — AGI — is going to be extremely, unbelievably potent, then it just does not make sense to open-source. It is a bad idea... I fully expect that in a few years it’s going to be completely obvious to everyone that open-sourcing AI is just not wise.”
It’s a valid debate, one where it’s easy to see both sides. Should the keys to our future belong to the tech bros or the tech bros and the terrorists?
One interesting twist late last month has to do with the Senate hearings on social media and its negative effects on children and teens.
Meta (A.I. open source leader) is being pressured to — and is now vowing to — enforce stricter monitoring, guardrails, and other measures. Kind of, you know, opposite to the kumbaya idea that an open, sharing community working together will keep us all safe.
The truth is: open source models, including Meta’s most recent model, are already out there. Some of them are so good and so small they can fit on your phone or even your Apple Watch. Really.
So, everyone in the world has the ability to play with, learn from, change, and build upon the A.I. programs that are out in public today.
But only if you have the resources to do so and you know what you’re doing.
Back to education
Last month, we talked about the fact that U.S. state school systems were (very) slowly starting to roll out their plans for bringing A.I. to the classroom. And that bringing A.I. literacy to everyone is a matter of equity.
Well, sending our kids into the world — with some of them in the know and some of them not — is effectively answering the open vs closed source question before even considering it.
Without schools actively teaching it yet and sophisticated open source models already available today, it is open to those who can afford to learn it and closed to those who can’t.
Where should we be playing?
As we know, we’re all using A.I. every day and have been for years.
But if you want to dip your toes into the world of generative A.I. — where chatbots can work with you to create (or generate) text, images, audio, and video — and you want to you do so with your kids, should you use a closed source model or an open one?
It’s up to your specific set of circumstances.
If you and your kid aren’t used to things like coding or if you’re looking for more guardrails and protections, use one of the closed source models getting all the attention today.
But if you have a kid who knows how to work with computers on a deeper level — or you want them to — playing in an open-source sandbox will let them run faster and farther than in a closed source one.
As we talked we about last time, both the budding techies and the budding not-so-techies are critical to keeping us whole in an A.I. world. And we’ll need both of these segments to keep those tech bros and the terrorists at bay.
