Peter Mousaferiadis gave a brilliant keynote ‘Diversified We Grow: Unlocking Diversity in the Age of AI’ delivered at Big Data & AI World, London, March 2024
The human poulation explosion!!
Years 100k- 10k 123 - 33- 14 - 13 - 35
From 1927 to 2022 we have gone from 2billion to 8 billion people
In less than 100 years we’ve quadrupled.
Are we at the top of a bell curve?
Where to now?
Growth through Culture - what is culture ?
The human has grown through being part of tribes and a common culture !
Culture is a mighty force, a fire that can warm and protect but can also burn and destroy.
UNESCO identifies culture as:
- a driver of sustainable development;
- an eradicator of poverty;
- key to quality education, and;
- key to building socially cohesive and peaceful communities.
Culture is also a significant enabler of innovation.
After all, innovation thrives on the diversification of ideas. What better way to foster innovation than by bringing together the diverse perspectives of different cultures?
So, what is culture?
Well, the term is extremely slippery, and its definition depends on what context and what type of culture we are referring to.
Simply put, culture is the ways in which a particular group of people live. Culture includes shared knowledge, values, customs, physical objects and social norms.
It provides the frames, the contexts, the lenses, through which we all see the world.
Different Cultures create conflicts and conflict costs
What is at stake? According to the Global Peace Index in 2022 alone we spent US$17.5 trillion dealing with conflict. According to UNESCO, about 75% of the world’s conflicts have an ethno-linguistic and religious cultural dimension.
The big cost is lives, livelihoods and human happiness.
Inclusiveness
How do we make sure everyone who wants to be included is, and no one is left behind? How do we give visibility to everyone?
Globalisation has made it a more-than-human challenge to understand ‘the other’. Do we need the aid of technology to do this ?
Connectivity Internet Social Media and AI
Today, with the near ubiquity of the internet we are finding ourselves in a super-diverse world where time and space have been compressed.
More than 65% of the world’s population has internet access and Meta (formerly known as Facebook) launched in 2004 (as The Facebook) now has more than 3 billion active users – nearly 56% of all internet users.
ChatGPT attracted a million users in just 5 days after going public on 30 November 2022. It now has more than 180 million users.
And Yet Peace has been on a steady decline since.
Migration and conflict
Humanity can be defined by an ongoing flux of migration and conflict. Peter’s grandparents were banished 101 years ago from the Ottoman Empire, from the area today known as Türkiye, and his parents migrated to Australia in the ’50s for a better life.
Who among us cannot be defined by a journey?
Ai and diversity
AI is a logical extension of human adaptability and sharing and quest for speed and knowledge.
Can it solve the challenges of globalisation: the intercultural misunderstandings and exclusions that are costing us so much in terms of human happiness and dollars?
Or will it, unchecked, exacerbate current conflicts and inequities?
The ability to understand and relate to the other is now more important than ever.
Human identity is complex because humans are multifaceted. As with a Rubik’s cube, it is not a successful outcome if I ‘solve’ one aspect of identity and leave the rest in disorder.
Our challenges require a holistic approach.
We need to factor in all attributes, not only gender, disability or racialised attributes but simultaneously all relevant identity markers if we’re to avoid the pushback and myriad unintended consequences of doing this work selectively. Gender, for example, is important but it is no more important than other human attributes when considering who gets left out or treated as lesser.
Data on its own is meaningless. It needs to be sorted into categories (datasets) to enable us to identify meaningful information.
We Don’t Know What We Don’t Know
How do you measure diversity - Diversity Atlas
What we can measure we can manage
Looking Under the Covers at Meta - they don’t have a handle on measuring culture
I looked under the covers at Meta.
This is a paper they published titled ‘A diverse, large benchmark for measuring fairness and robustness in audio/vision/speech models’.
In this paper they write, ‘The seven race groups used are White, Black, Indian, East Indian, Southeast Asian, Middle East, and Latino, and the dataset is reasonably balanced across these groups. However, race is seen as a social construct and its use in categorization exercises may be problematic.’
Say no more!
We need nuanced data that goes beyond appearance and nation states and broad geographic boundaries to define one’s culture and consider other areas of identity.
What would someone from a Kurdish background in Türkiye have in common with a Cantonese person from Southern China? A lot, but there is great dissimilarity, too. How can AI account for these nuances?
If we’re categorised into broad groups, AI will persist in making decisions under the assumption that we all begin from the same social and cultural standpoint.
How do we ensure that AI assists us on a healthy path towards greater understanding, compassion and joy and not more conflict, polarisation and misery?
Confluence is the lifeblood of culture and AI is a powerful expression of confluence.
If the development of AI is in the tight grip of a powerful few who can attract the necessary funding, IT IS LIKELY we will see stagnation in parts of the AI field. This could manifest as increased conflict and polarisation, accompanied by a collective sense of hopelessness, inertia, and a feeling of history repeating itself.
The alternative is learning from our past.
Diversity in the Age of AI: The future lies in the past
Diversify to grow
Cultural diversity is as important to sustaining humanity as biodiversity is for sustaining the environment. Great depths of human knowledge are encoded in each specific culture. We need to cherish these diverse expressions of humanity.
If AI is our collective future then we need more AI platforms, made by more people, reflecting the diversity of our world.
How Do We Mindfully Benefit from AI?
AI is rapidly creating breakthroughs, as in the medical field with AlphaFold, developed by Google’s DeepMind, which almost every pharmaceutical industry is now using.
In terms of DEI, AI has improved accessibility for people with disabilities through speech recognition, text-to-speech and visual recognition. AI is also helpful for bridging communication gaps between different language speakers.
ChatGPT 4, for instance, can communicate in more than 50 languages. Thanks to AI, barriers are tumbling down that may have once prevented employers from using someone’s skills.
I often use the new chatbots to help me communicate my ideas. They are like new members of staff.
Each of the language models, ChatGPT, Claude, Gemini and Pi, for instance, have different strengths, weaknesses and capacities, because they’ve each been shaped by different data inputs – just like real people but a lot better read, faster and may I add ever so slightly prone to hallucinating. AI chatbots can assist but can’t replace real humans because AI cannot reproduce the emotional depth, personal experiences and unique perspectives that humans bring to their work.
No company today can afford to think AI will compensate for a representatively diverse workforce or that AI is somehow free of bias and erasures. All the big AI developers in the Anglosphere have been criticised for bias, sometimes from within their own organisations. Even though the most recent iteration of ChatGPT has the capacity to communicate in 50 languages, it still excludes thousands of other languages representing value-based systems from across the globe. This is not a holistic system.
Female-founded AI startups currently receive only 2% of funding deals in the UK and US. Ada Lovelace was arguably the first ever computer programmer. Is this field less equitable for women than it was nearly 200 years ago? Who else is missing out on funding deals? Their absence as developers deprives the AI field of much-needed diversity and creates a lot of questions in my mind.
There’s never been a more important time to advocate for genuine representation in the technology industry. An equitable allocation of funding would lead to a more vibrant, less biased and more human future for AI.
If we have diversity in tech, we have healthy tech.
My final word is a call to work together, tread carefully and recognise the importance of every voice moving forward with AI.
ABOUT THE AUTHOR:
Peter Mousaferiadis
Peter Mousaferiadis has had an extensive career working in the cultural & creative industries. He has had a career as a conductor, creative director and producer and is considered a thought leader in culture as a driver of progress, social cohesion, and innovation. In 2002, he founded the internationally recognised organisation Cultural Infusion, which builds global harmony through intercultural action within education, ICT & the arts. Peter is a winner of numerous international awards and in 2021 was a winner of the Global Business and Interfaith Peace Award. His flagship product Diversity Atlas, a data-driven tool for measuring and understanding diversity in organizations is now being used by organisations throughout the world.
Tracking diversity equity and inclusion
Diversity Atlas and the Global Database of Humanity
Diversity Atlas is underpinned by the Global Database of Humanity, which comprises more than 42,000 human attributes, including every known language and dialect, secular and non-secular tradition, ethnic group and country of birth, gender, age, sexual orientation, sex at birth, position level, position type and many other dimensions, which can shed light on the correlation between experience and identity like never before.
Importantly, Diversity Atlas avoids lazy, outdated racial categories to allow people to gain a clear view of where inequities are playing out in any given context – which may or may not be along ethnic, religious and/or appearance lines. In other words, we don’t group people into broad, unnuanced categories like ‘African’, ‘Asian’, ‘White’ or ‘Black’.
Crude and broad categories of one kind or another were once necessary for people in leadership roles because they had to rely on humans to count groups of people and sort them into categories with assigned meanings. Now that we have machines to do this work there is no excuse for them not to take a nuanced approach.
The granular insights from Diversity Atlas help organisations identify stakeholders, shape Diversity, Equity and Inclusion (DEI) strategies and inform business goals by effectively leveraging the skills and knowledge of their people. It sheds light on organisational identity by answering the question who are we? Our work relies on AI working in combination with data scientists and cultural experts to process the vast amounts of intersectional data our analytics tool generates.
AI can assist with identifying patterns, however it cannot replace the humans whose job it is to understand the information a particular organisation needs. This work requires the experience of living as a human in the real world and the sort of intuition, compassion, creativity and critical thinking that comes from that. Even Pi.AI told me that.