Is Gender Bias in AI real?

Shaking a bottle of ketchup can make it 1,000 times thinner due to the spherical tomato particles turning into thinner ellipses, at which point it all flows out and drenches your fries.

What’s in store:

• AI - is the gender bias real?

• China has a new AI model, and the Chinese president is behind it.

• Around the web: Lego unveils the Great Deku Tree set from Zelda, featuring characters from Ocarina of Time and Breath of the Wild.

• Jurny is one of the hottest AI startups!

Read time: 6 minutes

GENDER BIAS IN AI

Is AI Gender-biased?

Gender bias in AI is a big issue affecting the fairness and effectiveness of the technology.

But how can a tool like ChatGPT have any biases, isn’t it just a bot? Well, the answer can be found in the data it was trained on.

AI inherits its knowledge from its creators and data, which can lead to unfair outcomes. For example, if an AI bot was trained on a subreddit of mostly male Redditors and their conversations, it would skew towards this type of communication, as that is what it has been exposed to more.

In general, AI should reflect our society as a whole - and with many believing women still have distinct disadvantages - certainly in some countries more than others - then the AI will reflect this real-world bias as “normal”.

This bias can appear in various applications, such as natural language processing (NLP), facial recognition, and voice assistants.

For instance, NLP might link certain jobs or traits with specific genders, facial recognition systems may struggle with underrepresented groups, and voice assistants using feminine voices can reinforce stereotypes.

Many AI training datasets lack diversity, leading to skewed results.

Another example is that if a facial recognition dataset mainly includes lighter-skinned individuals, the AI may struggle with darker-skinned faces.

Implicit biases from developers can also influence AI models.

Developers might accidentally encode gender stereotypes into an AI hiring tool if it’s trained on biased historical data.

The language and design choices in virtual assistants can also reflect and perpetuate gender norms.

Smart tech, dumb bias

Translation systems often associate jobs like 'nurse' with women and 'doctor' with men.

AI hiring tools can favour male candidates if trained on biased data, and image recognition software often shows fewer women in high-status roles, reinforcing gender biases.

This bias in AI could limit women's job opportunities, as biased hiring tools may favour male candidates, excluding a higher percentage of women before a human ever sees the application. This has already happened at Amazon.

Stereotyped language models can marginalise certain genders, and biases in critical areas like healthcare and finance can worsen inequalities.

These biases could also erode public trust in AI and reinforce societal prejudices.

The solution?

Using varied datasets to reduce stereotypes in AI is essential to prevent these inherent biases.

This involves incorporating data that represents a wide range of identities and experiences.

Promoting diversity in the AI development community helps ensure different viewpoints are considered.

Establishing guidelines for fair AI development by creating robust ethical frameworks and governance structures is also crucial for mitigating gender bias.

Educating end-users about potential biases in AI systems empowers them to advocate for fairer technologies.

By implementing these practices we can ensure AI benefits everyone equally.

In brief:

• Diverse Datasets: Essential for fair AI outcomes.

• Inclusive Practices: Crucial in AI development.

• Ethical Standards: Key for accountability.

Tackling gender bias in AI requires a concerted effort from researchers, policymakers, industry leaders, and users.

Together, we can create AI systems that truly reflect society's diversity and treat everyone fairly, regardless of gender.

I can’t stop thinking of a chatbot trained exclusively on Reddit data *shudders*.

Vote for live results and see results + opinions from yesterday at the bottom of the email.

AI STARTUPS

 Is AI About to Disrupt Hospitality?

This AI company is set to disrupt the $4.1T hospitality industry, and this is the chance for Mindstream readers to invest before the round closes on June 18th.

Jurny is an AI platform automating hotel and short-term rental operations, targeting over $355B in inefficiencies.

Partnered with industry giants like Airbnb, Vrbo, and Expedia, Jurny’s technology is fully automating operations for thousands of properties globally.

• In 2023, Jurny’s customer base grew 5x and it processed $35M+ in bookings.

• Featured by CNBC’s Squawk Box for its AI innovations and impact on hospitality, as well as by Forbes, Bloomberg, Skift, and many others.

• Successfully raised more than $12M from leading VCs and 1,200+ individual investors.

Secure your stake - the round closes June 18th! →

On This Day

May 31st 1859 - The clock tower at the Houses of Parliament, which houses Big Ben, starts keeping time. This iconic landmark is often used as a symbol of the United Kingdom and particularly London.

Riddle Me This

Invisible life force, intangible might,

I course unseen, powering devices that light

Our modernised world with abundant energy -

An ethereal drive, technological synergy.

My catalytic spark births circuits alive,

Animating silicon boards to survive and thrive.

Without form or shape, I am everything,

The enigmatic essence that makes CPUs sing.

A binding life current in our data-age plane,

Electrifying the clouds, wires, and information veins.

Able to be tamed, shaped, unleashed or reined,

What ephemeral power keeps our tech enchained?

Nifty Numbers

The Evolution of Robotics

This infographic traces the fascinating evolution of robot shapes from the 1950s to the present. It showcases the journey from early bulky humanoid robots inspired by science fiction, through the rise of practical industrial robots and mobile robots in the late 20th century, to modern sophisticated humanoid, animal-inspired, and adaptable soft robots.

AI IN CHINA

China’s new AI model

China has launched a new AI chatbot, trained on President Xi Jinping's political philosophy, "Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era."

This shows Beijing's ongoing efforts to control AI's role in informing internet users in China.

The chatbot is trained with official literature from the Cyberspace Administration of China (CAC), ensuring content aligns with the government’s views and stays professional.

Currently, it's used at a research centre under the internet regulator.

The chatbot can answer questions, create reports, summarise information, and translate between Chinese and English.

It might be released for wider use in the future.

Key points:

• The AI chatbot is trained on Xi Jinping's political philosophy.

• It must adhere to socialist values and avoid subverting state power.

• AI in China faces challenges due to limited Chinese language data sets for AI training.

China has made extensive efforts to promote Xi’s ideas across various platforms, including books, news apps, and educational tools like the Study Xi Strong Nation app.

The CAC requires generative AI to embody core socialist values and avoid content that undermines state power.

Challenges remain due to limited Chinese language data sets, often requiring models to use English language data, which may not align with China's speech norms.

Companies like Baidu and Alibaba strictly control their models to avoid sensitive topics.

To support developers, the Cyber Security Association of China has released a database of "high-quality and trustworthy data" for model training, heavily based on government and official publications.

Just what we needed—an AI chatbot with political opinions!

Mindstream Picks

A critical security flaw in the TP-Link Archer C5400X gaming router, tracked as CVE-2024-5035, has been patched in firmware version 1_1.1.7.

Aramco’s shares rose slightly on news from Reuters about a potential $10 billion share sale next month.

Deep Blue Aerospace has secured substantial funding and plans to launch and land China's first reusable rocket, Nebula-1, by the end of the year.

A new Lego set includes the Great Deku Tree from Zelda games, Ocarina of Time and Breath of the Wild, with interchangeable green summer foliage and pink spring blossoms.

Don’t Miss - Russia’s top metal producers are using Tether’s USDT for cross-border transactions with Chinese clients to avoid U.S. sanctions. While the volume of these trades remains unclear, the sources close to the companies claim the current alternatives to stablecoins are slower or even worse, risking an overseas bank account that could be frozen.

AI Art

Our Image of the Day

Daily Image Prompt

Submit your artwork to Mindstream →

We get a lot of submissions, but we do look at every single one! So please don’t hesitate to send us your art.

Yesterday’s Poll

“Do you think global AI safety institutes will lead to meaningful change?”

Yes, it's a step in the right direction. - 64% ✅

No, it's just talk. - 36%

Reader’s Opinions

“As always, talk will dominate until a major disaster occurs, then hopefully the best ideas from all the talk will be implemented.” - jim

“I believe overall AI is good but there needs to be WAY more regulation than there is.” - tshields

Submit your opinions in our daily poll for the chance to be featured!