If social media were the equivalent of a weapon of mass destruction, Jack Dorsey would be the digital era’s Robert Oppenheimer.
Speech at the 16th Annual General Meeting Oslo Freedom Forum On Wednesday, the Twitter co-founder expressed regret, explaining how conflicted he felt that his work, born from the lofty ideals of an open-source protocol, had been corrupted by the realities of surviving in a competitive marketplace.
And he warned: The social media and internet companies that have grown into multi-trillion dollar behemoths by finding content in the internet’s vast, sprawling space and curating it to maximize user engagement are only the first sign of this danger.
What Facebook and Google were to the web, companies like OpenAI are to the coming age of artificial general intelligence.
AI tools know us better than we know ourselves, and even if their algorithms are made transparent, they will influence our thinking on a subconscious level, either through design or default.
“It may sound a bit crazy, but I think the debate about freedom of speech is completely confusing at the moment. The real debate should be about free will,” he said in Oslo, the Norwegian capital and birthplace of the Nobel Peace Prize.
“we ProgrammedWe are programmed based on what we say we’re interested in, and we’re taught what’s interesting through these discovery mechanisms. And as we engage and interact with this content, the algorithms continue to build more and more of this bias.”
Dorsey, who stepped down from the board of microblogging platform Blue Sky last month in favor of Elon Musk’s rival company X, received moral support from the serial entrepreneur who acquired his creation for $44 billion in November 2022, but he also carried out mass layoffs and a complete strategic rebrand to remake it in his own image.
“Yes Jack is right,” the Tesla CEO posted. Wednesday.
Dorsey said opening up the underlying code to create transparency and build trust, as Musk did with X, doesn’t help.
Even if the algorithm is open source, in his view it remains effectively a black box.
The Twitter co-founder argued that not only is it impossible to model and predict how it will work and what it will show in any given situation, but it is also subject to change at any time.
“As people have become so reliant on social media, it is actually changing and impacting the power we have,” Dorsey warned.
“We can resist as much as we like, but because we are constantly communicating our preferences, both implicitly and explicitly, we feel that they know us better than we know ourselves, and that it would be extremely dangerous to continue to rely on that.”
Restoring agency through the algorithmic marketplace
In Dorsey’s view, the only answer is to create a marketplace of algorithms that lets users choose which black boxes they trust most, while also giving them the ability to seamlessly switch between them instantly as needed, or even build their own.
The need to find a solution will only become more urgent in the future, because neural networks like OpenAI’s GPT-4 are, at the most basic level, algorithms that use hundreds of billions of parameters to leverage data collected from the internet.
Until now, risks have been mitigated by continually publishing AI science that benefits all humanity, i.e. by open-sourcing research models.
But with OpenAI’s release of ChatGPT in November 2022, marking the start of a race by a small number of companies to commercialize the technology, those days are largely over.
Yeah
— Elon Musk (@elonmusk) May 7, 2024
“Five companies are building tools that we’re all going to be totally dependent on,” Dorsey said, “and they’re so complicated that we have no idea how to verify their accuracy. We have no idea how to verify how they work.” [or] What are they actually doing?”
Dorsey didn’t name any companies, but was likely referring to OpenAI and its partners Microsoft, Google’s DeepMind and Meta, and the Amazon-backed Anthropic.
He has warned against relying on CEOs like Sam Altman and, ironically, Elon Musk, so xAI may have been on the list.
But I doubt that the owners of X and xAI were going to approve it.