AI is in crisis. According to an MIT Sloan Management Review article, “Black workers represent only 2.5% of Google’s entire workforce and 4% of Facebook’s and Microsoft’s. Gender comparisons are also stark: Globally, only 22% of AI professionals are female.”
Many in the sector think that innovation in AI means faster computers, larger teams, more complex algorithms, and more money. Some will even deny there is a bias problem in AI at all. However, true technology innovators like Tony Effik, know that the real way to make AI more effective, accurate, and ultimately useful for a better world, is through increased efforts towards diversity and equality.
Effik is the Co-Founder, with his wife – Perky Noah-Effik, of The Black and Brilliant Advocacy Network – a volunteer organization dedicated to creating a more diverse workforce by upskilling and coaching high potential talent. Effik also serves as a Managing Director at Google, and Adjunct Assistant Professor at Columbia University’s Department of Industrial Engineering and Operations Research.
Earlier this year, The Black and Brilliant Advocacy Network partnered with online learning platform Codecademy and a network of professional coaches and mentors, to launch an AI Accelerator built specifically to empower Black and Brown students with the hard and soft skills they need to pursue career opportunities in AI.
In this powerful interview, Effik shares his thoughts on the current diversity problem in AI as well as his hopes for the future. He also dives deeper into his work with Black and Brilliant, and the incredible volunteers and students that make this exciting program, including their AI Accelerator, a reality.
What is the inspiration that led to the founding of Black & Brilliant?
May 25, 2020 was a critical date – the date of the murder of George Floyd – which brought a critical awakeneing not just of the problem of racial injustice, but of the need to organize and act. The Black and Brilliant Advocacy Network was born in June that year as a reaction to the inequality that exists in the workplace. It started as a post on Linkedin and turned into a loose network of people collaborating around the world to create a more diverse workforce. We have people of all backgrounds across the USA, Canada, South Africa, UK, and many other countries contributing time, access, and ideas.
What is your career background working in AI?
I started my career managing strategy and analytics teams in the digital media space, and that evolved into managing data scientists and engineers building machine learning applications and optimization algorithms for digital media. Over the last 6 years or so, I have taught a graduate course at Columbia University’s Department of Industrial Engineering and Operations Research focussed on using optimization techniques in Digital Business Models. With the increasing importance of AI in the community, I have borrowed a phrase from James Briggs, an AI entrepreneur in our network, who says: AI is too important for AI Engineers alone to handle. We all need a seat at the table, and everyone should be represented.
Could you elaborate on the problem that this quote from Black & Brilliant About page: “challenging and eliminating the idea that far too many corporate leaders hold about the lack of a Black Talent and a Pipeline”
This is an often repeated statement by business leaders that turns the problem of racial injustice on its head, and says the problem is not hiring practices but instead the problem of Black talent. Even well meaning executives cite this as an explanation for lack of racial equity in hiring and career advancement.
We like to say talent is equally distributed, but opportunity is not. The issue isn’t lack of talent. The issue is lack of talent in the right places. Black talent is usually not driving in the fast lanes of the economy. The stats are 452,760 or 11.1% of college graduates are Black or African American, which is nearly inline with the 13% share of the total population. We need to help put more in the faster lanes, including fields like AI. In the United States, this starts with college.
- More African-Americans are completing high school and going to college than ever before. But according to research from Georgetown University, they study courses that lead to debt and low paying jobs.
- “African-Americans make up only a small percentage of some of the highest-paying majors, including those in STEM and business. They’re only 8 percent of engineering, 7 percent of mathematics and 5 percent of computer science majors. “
- Those who do study STEM courses are studying the worst paying STEM degrees which are civil and mechanical engineering. We need to get these people into the faster lanes of the economy where their aptitude and education can help drive this new economy (and diversify it)
- Even those Black professionals that make it into STEM careers are paid at significantly levels than their non-Black colleagues
How does the above false belief of a lack of Black talent manifest specifically in the AI sector?
We need to firstly understand that there are many roles in AI outside of the role of engineer. Some are technical and some non-technical: Product Manager, UX, Data Visualization. AI Ethicist, AI Lawyer, Analyst, Technical Program Manager, AI Sales Person, AI Training, Strategist, Data Engineer, QA, and many more. All of these roles need more diversity.
There is Black talent in adjacent industries and disciplines that could easily change lanes with transferable skills to AI. Many Black+ talents have not been able to accumulate the “badges” in the traditional areas that lead to Silicon Valley, e.g. type of school, network, language and culture fit, etc. This is the core belief that led to the AI Accelerator idea with Codecademy.
Could you speak a bit more about the new Codeacademy AI Accelerator?
The core argument is that lack of diversity is anchored on a concept I am calling Readiness – which is centered on the question: are you ready for a career in this space? There are hard and soft components, and we structured the AI Accelerator with Codecademy to address both.
Hard components are too often used as the excuse for holding people back. But platforms like Codecademy now democratize the hard components, allowing anyone to obtain these skills if they are motivated and supported, and most importantly inspired. That brings us to soft skills.
In the world of AI, we now need to recognize two types of soft skills:
- the perspective that diverse teams of engineers bring to responsible AI and
- the soft skills that are used to prevent brown and black talent entering that workforce
We know as an example there were many women in computer science until the early eighties until it started being marketed as a male domain. Something “soft” like career appropriateness kept women out, after initial success (source: NPR).
Soft skills include confidence, being inspired and feeling this is a domain for me and people like me. There is also a familiarity with the language and terrain of a particular industry — having what we call the “Cheat Codes” to do well in the recruitment and internal mobility processes. Many Black+ people don’t have these soft skills, because they haven’t been exposed to them or mentored by somebody on the inside. That’s why mentorship is a key part of the accelerator, in addition to upskilling via Codecademy.
Is Enterprise AI a part of this program?
Enterprise AI was not explicitly planned as part of the program, but it emerged regularly as a recurring topic. We covered topics that impact Enterprise AI where bias could creep in, such as image and semantic classification, visual recognition, skin tone detection, lack of non-Judeo-Christian names in libraries of voice systems for names common in the Black community, predictive analytics and recommendations systems that compound existing racial biases, geography as input such as zip codes, and many more. These are all considerations our coaches tackled.
Many of our coaches came from the Enterprise AI world or closely aligned worlds such as Data Engineering, B2B consulting and agencies, cybersecurity, ML Engineer, and many others, so they leaned heavily on this corpus of knowledge.
We structured the program into 4 modules
- Programming and Data Engineering
- AI Ethics & Social Justice
- User Experience Design/Research/Creativity:
- User Research and Usability Research
Why, from your experience, is it so important to have more Black technologists working in AI?
There are economic and social upsides and downsides across many spheres of AI that will uniquely impact the Black community.
Economic Upside – Digitization of the global economy is in full swing. AI is driving the next wave of digitization. The upside is that thousands, if not millions, of new jobs will be created in obvious spaces like engineering, but also in adjacent spaces such as law and ethics, project management, analytics, UX, venture capital and finance, and many more.
Economic Downside – There are hundreds of scary stories about AI and digital technology, but few dive deep into the impact on Black and Brown communities. There will be a disproportionate number of job losses among Black and Brown workers based on current employment patterns, for example, frontline workers. We need to upskill and mobilize these people
Social Downside – We are codifying existing social and racial biases into AI. In Black and Brilliant we have an AI subgroup that is focusing on something called the Black Data Gap that is amplifying inequalities
Having Black+ technologists on AI teams helps ensure group think doesn’t happen and that light is shone on blindspots, in very practical ways such as ensuring that there are rigorous checks and balances in building datasets, labelling, training, dataset versioning, model management, and in application design.
In a recent article titled, “Making Enterprise AI an Organizational Asset” the author writes, “The ability to become a true AI enterprise by successfully scaling and employing robust data methodology at all levels of the company is an organizational asset pivotal to the success of businesses of the future, no matter what industry” – how does diversity, specifically racial diversity, enhance (is necessary for) an effective and successful enterprise AI strategy?
I completely agree with the author. Scaling and employing robust data is pivotal to success. This is important for two reasons:
1. Representativeness in who you build for and who is doing the building of Enterprise AI systems:
Garbage in, garbage out. If there is a lack of representation in the data or a misunderstanding of that data then the predictive accuracy of the data model diminishes, you’re also not able to fully capitalize on the flywheel of data learning effects, and even worse you might cause damage. If for example you are using supervised learning and using humans to do data labelling to build a training dataset and those humans are biased then you corrupt the dataset. If the N count of the dataset is low or biased or if you don’t understand how the applications impact your organization or marginalized subsets of your consumers then the advantages of the system are outweighed by the risks.
2. Augmented Artificial intelligence
Racial diversity enhances Enterprise AI for one simple reason – the best technologies will be People-centered technologies. In the race towards digital transformation and competitive advantage the real winners are those who solve human-computer interaction problems most efficiently to augment human intelligence. You can’t gain that advantage if you only build for a portion of humanity, and do that building with talent from just a portion of the population. The gap between the most advanced builders of Enterprise AI systems and the laggards can quickly be closed. People-centered technologies will stay ahead and be most resilient because of product/market fit and the compounding effect of their data learning cycle. People come in a diverse array of backgrounds, so building for diversity by a diverse team allows businesses to stay ahead.
Could you speak a bit more on how “all #BlackandBrilliant initiatives are data and insight-driven”?
All of our initiatives start from an insight. The insight can come from qualitative or quantitative data or just simple observation. We analyse open rates and clicks to our newsletter, views on social media, and profile our membership, but we go beyond that and talk to people around the world to identify the most common problems and best solutions.
One of our most persistent insights we use is that the middle of most organizations is the actual battlefield. We’ve seen around the world that firms claim credit for diversity initiatives by hiring junior people. These junior people rarely make it to the middle and the few in the middle don’t make it to the top, and when it comes to hiring for the top business leaders complain about a lack of talent in the middle. We ran a range of programs centered on Breaking through the Middle. We also identified early on that many in our network were not aware of the opportunities and perils of AI, so we did a series of events on “Are you ready for AI?”
What are some examples of the impact the organization has made thus far?
- Our last cycle of the AI Accelerator graduated 85 people. Many have been able to make job changes and even career changes.
- In the UK, we worked with Facebook and others in the BrIM program, which is about Black representation In Marketing to set new guidelines and frameworks.
- We’ve done training programs on building more diverse organizations with firms like Doubleverify and Wunderman Thompson
- We’ve been working with Buzzfeed in a program called Get to know…which spotlights Black talent to challenge the notion of a lack of black talent.
- We co-led the Festival of Diversity event in the UK to help organizations map a path forward in the world of diversity.
- We partnered with the digital transformation consulting firm, R/GA to reimagine the careers fair to introduce a generation of Black+ talent to the world of digital transformation careers.
- We are helping mentor a cohort of strategists with the global agency network, BeenThereDoneThat.
- Conducted events that celebrated unrecognized Black+ talent with NBCUniversal.
You can support the great work Effik and the Black and Brilliant team are doing by getting involved via this link.
Credit: Source link