Speaking at the 16th annual Oslo Freedom Forum on Wednesday, the remorseful Twitter cofounder expressed how conflicted he felt at his own creation born out of the noble ideals of open-source protocols before becoming corrupted by the realities of needing to survive in a competitive marketplace. 

And he had a warning: Social media and internet companies that grew into trillion-dollar behemoths by discovering content in the vast, unorganized expanse of the internet and curating it to maximize user engagement are just the first manifestation of this danger. 

What Facebook and Google were for the Web, companies like OpenAI are for the coming era of artificial general intelligence.

AI tools will know us better than we know ourselves and, either through design or default, influence our thinking at a subconscious level, even if their algorithms are made transparent. 

“This is going to sound a little bit crazy, but I think the free speech debate is a complete distraction right now. I think the real debate should be about free will,” he said in Norway’s capital of Oslo, home of the Nobel Peace Prize.

“We are being programmed. We are being programmed based on what we say we’re interested in, and we’re told through these discovery mechanisms what is interesting—and as we engage and interact with this content, the algorithm continues to build more and more of this bias.” 

Dorsey, who left the board of microblogging platform BlueSky to endorse Elon Musk’s rival X last month, received moral support from the serial entrepreneur who purchased his creation for $44 billion in November 2022, only to refashion it in his own image, complete with mass firings and a total strategic rebranding. 

“Yeah, Jack is right,” the Tesla CEO posted on Wednesday.

According to Dorsey, revealing the underlying code, as Musk has done with X, to create transparency and build trust won’t help either.

Even if an algorithm is open-source, it remains effectively a black box in his view.

Not only is it impossible to model and predict how it works and what it’s going to show you in any given instance, but it can be changed at any time, the Twitter cofounder argued.

“Because people have become so dependent on it, it’s actually changing and impacting the agency we have,” Dorsey warned.

“We can resist it all we want, but it knows us better than we know us, because we tell it our preferences implicitly and explicitly all the time, and it just feels super dangerous to continue to rely on that.” 

Restore agency through a marketplace of algorithms

The only answer, in Dorsey’s view, is to create a marketplace of algorithms that return choice to users about which black boxes they find the most trustworthy while granting them the ability to seamlessly switch from one to another at a moment’s notice depending on their needs—or even build their own.

This need to find a solution only becomes even more pressing as we move forward, since what a neural network like OpenAI’s GPT-4 is at its most basic level is an algorithm harnessing data scraped from the internet using hundreds of billions of parameters. 

In the past, the risk was ameliorated through the constant publishing of AI science to benefit all of humankind—open-sourcing the research models, in other words. 

Yet these days have all but ended as OpenAI’s November 2022 release of ChatGPT fired the starting shot in a race among a handful of corporations to commercialize the technology.

Yup

— Elon Musk (@elonmusk) May 7, 2024

“Five companies are building tools that we will all become entirely dependent upon,” Dorsey said. “And because they’re so complicated, we have no idea how to verify the correctness, we have no idea how to verify how they work, [or] what they’re actually doing,” 

Dorsey didn’t name them, but he likely meant OpenAI along with partner Microsoft, Google’s DeepMind, Meta, and Amazon-backed Anthropic. 

He may have also potentially included xAI, as he warned against becoming dependent on any one CEO, such as Sam Altman—but, ironically, Elon Musk as well.

Whether the owner of X and xAI meant to endorse that as well, however, is questionable.