Bias in AI: Let’s demand the algorithm dances to our tune
In the week that the world met to discuss the future of AI, the bid for Open AI by Elon Musk, the high-priest of "anti-woke", reminded us precisely why independent governance, safety, and oversight are non-negotiable for the future deployment of AI.
The efforts of the AI Action Summit in Paris to focus on tackling AI bias and diversity through collaborative oversight could not come at a better time.
But are the words on the communiqué really enough for the scale of the challenge facing us?
As a technology, there is no doubt about the potential of AI to improve equity and fairness.
In today’s world, those who stand to gain the most are those who are still the most disadvantaged. A case in point, discussed at the summit, is the Global South. With fewer legacy systems and greater potential for adaptability, AI can help leapfrog traditional barriers in healthcare, education, and finance, providing telemedicine to remote areas, multilingual education resources, and financial access for the unbanked.
These transformative benefits don’t come without risks. So, unlike the gingerbread man riding across the river on the back of the predatory fox which he believes to be his saviour, we need to think ahead to what the end game looks like.
Again, looking at the Global South, a reliance on foreign-developed AI models risks creating a new kind of digital dependency. Critical data and decision-making infrastructure will be outsourced to external tech giants. The surrender of privacy, economic power, and the extraction of local resources for foreign innovation with little reinvestment are all potentially part of the trade.
These contrasting dynamics vividly illustrate why systemic, global oversight of AI is urgently needed. And, as both the UK government and Dario Amodei, CEO of Anthropic, have pointed out, given the rapid pace and disruptive scope of AI, what’s needed is action today.
This isn’t because of some notional future risk of inequality. It is because the AI models we have today already contain bias and lack the systemic governance to self-correct and improve.
For this reason, I think that to embrace the future, we need to learn from the mistakes of the past.
If there is one thing we have learnt from the recent waves of technological advances—like social media, internet commerce, and workplace tech— it’s that when left unchecked they wreak unintended havoc.
We have also learnt that meaningful regulation and oversight are nearly impossible to put in place against the commercial and political might of the major tech businesses.
It is, then, the individuals and society who bear the costs and deal with any unintended consequences.
The current momentum behind AI makes its wholesale adoption inevitable. But it should not be accompanied by the current "proceed until apprehended" approach which, we know will only deepen systemic bias which is already so pernicious and entrenched.
For developers of AI they must prove that these systems are not only bias-free but also transparent and accountable to human oversight.
Those who deploy AI have a role too. As the UK law firm Hill Dickinson showed in its stand against the runaway use of ChatGPT – which it has blocked unfettered use of AI in its business – unchecked reliance on AI without safeguards or consideration of future consequences is not good for the firm or its people. We do have the power to pause.
This is not about trying to hold back the inevitable deluge but ensuring that we are best equipped to keep our heads above the water and free of the currents which will drag us down.
The way we do this is by ensuring that where we have influence, we put that to work, ensuring the narrative and practice of blind AI adoption is called out for all its risks and bias.
For AI to succeed for us all, transparency, human oversight, and enforceable standards must be part of the system, not optional add-ons for a later date.
If we fail to make that case, we only need look at the unchecked power that Musk, Zuckerberg and Bezos have over our lives for a glimpse of what the future looks like.
And like the fairytale of the gingerbread man, it’s a story with a predictable ending.
