AI Gender Bias, Ethics and Deepfakes

Greetings!
This month, we're looking at what everyone seems to be looking at, too - artificial intelligence, but with a different view.

This edition covers three critical updates: AI gender bias and sexism, AI ethics, and AI’s role in the rapid spread of deepfake technology. While I intended our first newsletter to be more upbeat, recent events have shifted our focus.

First, I share my recent LinkedIn exchange with a Google exec, which was the catalyst for putting AI ethics on the list this month.

We then examine OpenAI CEO Sam Altman's controversy over ChatGPT's voice assistant, Sky, and its implications for sexism in AI.

Lastly, we explore the alarming rise of deepfake technology, especially in pornography, and its impact on society, noting the growing scrutiny from Capitol Hill.

Though these topics are challenging, they are essential for non-tech leaders to understand the current landscape.

Next month we’ll covering Blockchain and AI, We will revisit AI energy use and AI regulation updates in September.

Trend 1: AI Ethics

Frequently Avoided Questions

It's feeling very 1990s out there with some bubbling signs of "irrational exuberance," one of my favorite phrases from the Greenspan Fed era. I'm having deja vu right about now.

Even though AI ethics and regulations issues are making news now, I've been asking FAQs (Frequently Avoided Questions) on LinkedIn for over a year. The post below is from early 2023, the Wonder Years phase of ChatGPT's early rollout. My question got a thumbs-up from Arun (but no response.)

I smile at my naivete now. He was probably thinking, Where do I even start?

AI Ethics, In Brief 

The central dilemmas around AI involve complex, interrelated questions about privacy, jobs, and human freedom of action. There are concerns that AI will worsen bias, perpetuate gender stereotypes, and undermine privacy. ChatGPT sums up the basic dilemma below.

Who is working on AI Ethics? 

All Tech is Human maintains an impressive database of 629 ethics-in-tech-related non-profits. If you run one, you can add your organization, too. Google's position on responsible AI practices is worth a scan. You might find the section on Fairness particularly interesting.

In American society, we fund tech startups with millions and leave it to NPOs, activists and regulators to sort out the ethics, usually after the fact.

That's why you'll find United Way covering an article from Newsweek about Sam Altman's recent co-founding of a new AI Ethics Council with Operation HOPE CEO John Hope Bryant. 

They will build a "non-binding AI governance model that ensures that traditionally underrepresented communities have a voice in the technological revolution."

Bryant said, "Poor people, underserved folks don't get a voice in moments of change. Racism and sexism should not come into play." 

I agree, but bias is already here, big time. It would be more effective if tech leadership initiated these discussions earlier in AI's development. 

Also, this: Why are only three women on the founding nine-member council?   

Speaking of sexism…

Trend 2: It's giving the Ick- sexism and AI gender bias

The male-dominated tech industry has been under fire for years for sexism. Globally, women make up less than one-third of employees in the tech sector,  only 8% of Chief Technology Officers in the US, and only 22% of AI workers.

On the flip side, one way companies are using AI to help reduce gender bias is in hiring. The Boston Consulting Group notes that companies using AI tools to remove bias from the interview process are seeing hiring improvements in gender diversity.

However, we have a long way to go. In the age of AI, gender bias and sexism are present in some of the industry's early products, services, and leaders. One high-profile incident from the tech sexism scrapbook relates to the industry’s preference for chatbots with submissive female voices and involves the CEO of OpenAI, Sam Altman. Here's the story:

Sam and Scarlett

Sam Altman asked Scarlett Johansson, a prominent actor and celebrity, to be the voice of ChatGPT's voice assistant. Sam is a big fan of the 2013 movie Her, where Scarlett did the voiceover for a lonely man's AI companion.

Scarlett said "no" to being the voice of ChatGPT. 

So when OpenAI released their demo of "Sky," the voice of ChatGPT, and Sky sounded just like Scarlett, the actor was livid.

Sam denied all wrongdoing even while taking Sky offline ASAP. Scarlett's lawyers asked for a sit-down with Sam and his lawyers. Media coverage was swift and sharp. This post from CNN details some of the legal considerations.

I’m so curious! It seems that Altman should have known better; why didn’t he? Did OpenAI's legal team or the company's female CTO, Mira Murati, try to give him a heads-up and save him from an expensive public PR debacle?

The Problem with Sky: Women, Society and OpenAI

As if allegedly stealing a popular celebrity's voice after she said no isn't bad enough, many social commentators raised doubts about Sky's personality. Sky’s will likely be the most influential bot in human history (so far), and her personality wasn't exactly ushering in a new era for women's empowerment in tech and society. 

In the New York Times, Alissa Wilkinson wrote that Sky's style is "lightly flirtatious, wholly attentive woman who's ready to serve the user's every whim, at least within the limits of her programming."

This article from Fast Company gives a good overview of OpenAI's less-than-stellar reputation for, well, Openness.  

I asked ChatGPT about the Sky-Sam-Scarlett controversy, and its parting comment was: 

OpenAI's legal team is addressing the situation, but the ethical concerns remain significant.

Smart bot.

Trend 3: AI Deepfake Tech Abuse

AI gender bias has a problem even worse than voice stealing - it's like digital human trafficking. People using AI to create deepfake porn can steal someone's (usually a woman, sometimes a man, or a child’s) face and body images, animate them with AI, and resell them. Some contend that non-consensual deepfake manipulation is a form of digital gender violence.

Deepfake porn is the fastest growing use of generative AI, so much so that the industry is now under fire from Alexandria Ocasio-Cortez, the fiery, outspoken US House Representative from New York. She introduced legislation in March 2024 to criminalize non-consensual deepfake activity.

What is Deepfake Tech? 

AI deepfake technology utilizes AI to superimpose a person's likeness, such as their face or voice, onto another individual or image. Before AI, deepfakes required extensive visual assets and time to create. With AI, users only need a couple of images and a smartphone app to create deepfakes.

Deepfake media is called "deepfake" because the image or video looks real, with the ability to portray someone saying or doing things they never did.

The charts below from Humanorai.io show the effect of AI on deepfake technology adoption, based on a comprehensive study of more than 30 deepfake applications and non-application platforms.

Government and media watchers have significant concerns about the misuse of deepfakes to create fake news or impersonate public figures. But the most significant threat right now is from deepfake porn, the fastest-growing segment of AI deepfake technology use. 

According to research from Sensity AI, a visual threat intelligence company that has been tracking deepfake online videos since 2019, 90% to 95% are non-consensual pornography, and 90% are non-consensual pornography of women. 

Deepfake porn services target both movie stars and everyday people, mostly women. Ex-wives, ex-girlfriends, and crushes are popular targets. Deepfake child pornography is on the rise, too. 

What do you think? The world would survive without deepfake technology. Is it necessary at all? Do the benefits justify the downside?

Wrapping up July

While regulating the usage of AI is important going forward, AI is not the reason why gender bias exists or why pornography is hugely popular and profitable. Stereotypes are the basis of gender bias, and the demand for pornography is rooted in a complex blend of toxic cultural messaging and brain chemicals like dopamine.

AI is a tool— it can help improve society or make things much worse. In reality, we will likely see both outcomes. We aim to provide a balanced view of AI's impact, showcasing positive developments, innovative solutions, and individuals making a difference in the tech world. Our goal is to inform non-tech leaders about the challenges and opportunities in AI. Thank you for reading!

FAQs

1. What is the controversy surrounding OpenAI CEO Sam Altman and ChatGPT's voice assistant?

• OpenAI CEO Sam Altman faced backlash for allegedly modeling ChatGPT's voice assistant, Sky, after Scarlett Johansson's voice without her consent. Despite her refusal to participate, Sky's demo sounded remarkably similar to Scarlett's, prompting legal action and significant media scrutiny. Critics also noted Sky's overly flirtatious personality, raising concerns about sexism in AI products.

2. What is deepfake porn, and why is it spreading so fast?

• Deepfake porn involves using AI to superimpose a person's likeness onto explicit content without their consent. This technology has become the fastest-growing segment of AI deepfakes, with 90-95% of deepfake videos being non-consensual pornography, predominantly targeting women. The ease of creating deepfakes with just a few images and a smartphone app has contributed to its rapid spread, drawing attention from lawmakers like Alexandria Ocasio-Cortez.

3. Why are AI ethics important, and what are the central dilemmas?

• AI ethics are crucial for balancing the benefits of AI advancements with potential risks to privacy, employment, and human agency. Central dilemmas include data privacy, surveillance, bias, and job displacement. Addressing these issues requires transparent, equitable, and responsible AI development to protect individual rights and promote societal well-being.

4. Who is working on AI ethics?

• Numerous organizations are dedicated to AI ethics. All Tech is Human, for example, maintains a database of 629 ethics-in-tech-related non-profits. Additionally, initiatives like the AI Ethics Council co-founded by Sam Altman and Operation HOPE CEO John Hope Bryant aim to ensure diverse voices are heard in AI development. However, there is ongoing criticism about the inclusivity and effectiveness of these efforts.

5. How is sexism manifesting in the AI industry?

• The AI industry, like the broader tech sector, has been criticized for sexism. Incidents like the controversy with OpenAI's Sky highlight how AI products can perpetuate gender biases. The male-dominated nature of the industry often leads to products and decisions that do not adequately consider women's perspectives or the broader impact on gender equality.

6. What can be done to address deepfake porn and its impact?

• Addressing deepfake porn requires a multi-faceted approach, including stricter regulations, technological solutions to detect and prevent deepfakes, and increased public awareness. Legislative efforts, such as those led by Alexandria Ocasio-Cortez, are crucial for creating legal frameworks to combat non-consensual deepfake content and protect victims.