Artificial Intelligence Is Brilliant But Biased!

In today’s world, Artificial Intelligence (AI) is at the center of technology talks. AI tools have changed our professional and daily lives, finding a place in many industries. Recently, OpenAI and its chatbot ChatGPT have taken up the headlines, as the number of industries and companies looking to make use of its capabilities increases exponentially.

Such AI tools are trained using a large quantity of high-quality data. This includes OpenAI’s ChatGPT model, for which the exact amount of data collected is not known but is believed to consist of hundreds of billions of words or even more! Once acquired, the data is filtered and processed. After this, advanced machine learning (ML) techniques train the model to learn how words and phrases are connected in order to generate text. Essentially, it’s the quality and size of the dataset as well as the training that has a major impact on how the tool performs.

These AI tools coupled with the promise and capabilities of Web3, could produce novel and innovative products and services. Although the Web3 market size is tough to ascertain given its size is still small and it’s still evolving, it is expected to grow at a quick pace.

While its market size is believed to be around $2 billion in 2021, analysts and research firms believe it will rise to $80 billion by 2030, with a CAGR (Compound Annual Growth Rate) of around 45%.


However, a key issue is that bias in AI can produce unfair and discriminatory results in key sectors like employment, housing, finances, criminal justice and others.

Reportedly, in 2018, Amazon scrapped an AI recruiting tool that displayed bias against women. Amazon used resumes, which consisted of mainly male candidates received over 10 years, to train the tool. This led to the AI downrating resumes with words like “female” and “woman”.

Similarly, in 2019, researchers reported that an outcome-predicting AI algorithm showed bias against African American patients by producing false reports, as the tool was trained using predominantly Caucasian patient data.

Meredith Broussard, an NYU professor and former software developer, shared her thoughts, saying, “The problems I'm really concerned with inside AI are racism, sexism, and ableism. Structural discrimination and structural inequality exist in the world and are visible inside AI systems. When we increasingly rely on these systems to make social decisions or mediate the world we’re perpetuating those biases.”


The Web3 industry faces issues with diversity too. Reportedly, as of 2022, only 26.7% of technology jobs were held by women of which 56% were women of color, with a lower rate for executive positions. Numerous analysts found that less than 5% of Web3 startups had female founders, all of which could lead to male and Caucasian founders inadvertently ignoring data bias.

In the end, Web3’s decentralized form combined with AI’s lack of diversity in training could increase the bias in its results. This could be overcome by increasing inclusiveness and diversity in training data and involved personnel.

What do you think can be done to increase inclusiveness and diversity to achieve unbiased AI tools? Let us know in the comments below!