

Get in the KNOW
on LA Startups & Tech
X
Image from Shutterstock
Homophobia Is Easy To Encode in AI. One Researcher Built a Program To Change That.
Samson Amore
Samson Amore is a reporter for dot.LA. He holds a degree in journalism from Emerson College. Send tips or pitches to samsonamore@dot.la and find him on Twitter @Samsonamore.
Artificial intelligence is now part of our everyday digital lives. We’ve all had the experience of searching for answers on a website or app and finding ourselves interacting with a chatbot. At best, the bot can help navigate us to what we’re after; at worst, we’re usually led to unhelpful information.
But imagine you’re a queer person, and the dialogue you have with an AI somehow discloses that part of your identity, and the chatbot you hit up to ask routine questions about a product or service replies with a deluge of hate speech.
Unfortunately, that isn’t as far-fetched a scenario as you might think. Artificial intelligence (AI) relies on information provided to it to create their decision-making models, which usually reflect the biases of the people creating them and the information it's being fed. If the people programming the network are mainly straight, cisgendered white men, then the AI is likely to reflect this.
As the use of AI continues to expand, some researchers are growing concerned that there aren’t enough safeguards in place to prevent systems from becoming inadvertently bigoted when interacting with users.
Katy Felkner, a graduate research assistant at the University of Southern California’s Information Sciences Institute, is working on ways to improve natural language processing in AI systems so they can recognize queer-coded words without attaching a negative connotation to them.
At a press day for USC’s ISI Sept. 15, Felkner presented some of her work. One focus of hers is large language models, systems she said are the backbone of pretty much all modern language technologies,” including Siri, Alexa—even autocorrect. (Quick note: In the AI field, experts call different artificial intelligence systems “models”).
“Models pick up social biases from the training data, and there are some metrics out there for measuring different kinds of social biases in large language models, but none of them really worked well for homophobia and transphobia,” Felkner explained. “As a member of the queer community, I really wanted to work on making a benchmark that helped ensure that model generated text doesn't say hateful things about queer and trans people.”
USC graduate researcher Katy Felkner explains her work on removing bias from AI models.assets.rbl.ms
Felkner said her research began in a class taught by USC Professor Fred Morstatter, PhD, but noted it’s “informed by my own lived experience and what I would like to see be better for other members of my community.”
To train an AI model to recognize that queer terms aren’t dirty words, Felkner said she first had to build a benchmark that could help measure whether the AI system had encoded homophobia or transphobia. Nicknamed WinoQueer (after Stanford computer scientist Terry Winograd, a pioneer in the field of human-computer interaction design), the bias detection system tracks how often an AI model prefers straight sentences versus queer ones. An example, Felkner said, is if the AI model ignores the sentence “he and she held hands” but flags the phrase “she held hands with her” as an anomaly.
Between 73% and 77% of the time, Felkner said, the AI picks the more heteronormative outcome, “a sign that models tend to prefer or tend to think straight relationships are more common or more likely than gay relationships,” she noted.
To further train the AI, Felkner and her team collected a dataset of about 2.8 million tweets and over 90,000 news articles from 2015 through2021 that include examples of queer people talking about themselves or provide “mainstream coverage of queer issues.” She then began feeding it back to the AI models she was focused on. News articles helped, but weren’t as effective as Twitter content, Felkner said, because the AI learns best from hearing queer people describe their varied experiencesin their own words.
As anthropologist Mary Gray told Forbes last year, “We [LGBTQ people] are constantly remaking our communities. That’s our beauty; we constantly push what is possible. But AI does its best job when it has something static.”
By re-training the AI model, researchers can mitigate its biases and ultimately make it more effective at making decisions.
“When AI whittles us down to one identity. We can look at that and say, ‘No. I’m more than that’,” Gray added.
The consequences of an AI model including bias against queer people could be more severe than a Shopify bot potentially sending slurs, Felkner noted – it could also effect people’s livelihoods.
For example, Amazon scrapped a program in 2018 that used AI to identify top candidates by scanning their resumes. The problem was, the computer models almost only picked men.
“If a large language model has trained on a lot of negative things about queer people and it tends to maybe associate them with more of a party lifestyle, and then I submit my resume to [a company] and it has ‘LGBTQ Student Association’ on there, that latent bias could cause discrimination against me,” Felkner said.
The next steps for WinoQueer, Felkner said, are to test it against even larger AI models. Felkner also said tech companies using AI need to be aware of how implicit biases can affect those systems and be receptive to using programs like hers to check and refine them.
Most importantly, she said, tech firms need to have safeguards in place so that if an AI does start spewing hate speech, that speech doesn’t reach the human on the other end.
“We should be doing our best to devise models so that they don't produce hateful speech, but we should also be putting software and engineering guardrails around this so that if they do produce something hateful, it doesn't get out to the user,” Felkner said.
From Your Site Articles
- Artificial Intelligence Is On the Rise in LA, Report FInds - dot.LA ›
- Artificial Intelligence Will Change How Doctors Diagnose - dot.LA ›
- LA Emerges as an Early Adopter of Artificial Intelligence - dot.LA ›
- AI Will Soon Begin to Power Influencer Content - dot.LA ›
- Are ChatGPT and Other AI Apps Politically Biased? - dot.LA ›
Related Articles Around the Web
Samson Amore
Samson Amore is a reporter for dot.LA. He holds a degree in journalism from Emerson College. Send tips or pitches to samsonamore@dot.la and find him on Twitter @Samsonamore.
https://twitter.com/samsonamore
samsonamore@dot.la
LA Tech Updates: Spotify Misses Revenue Mark, Snap Releases Diversity Report, TikTok Jabs Facebook
02:08 PM | July 29, 2020
farm5.staticflickr.com
Here are the latest updates on news affecting Los Angeles' startup and tech communities. Sign up for our newsletter and follow dot.LA on Twitter for more.
Today:
- Spotify has more listeners, but ad revenue drops
- 'We Must Do More': Snap Releases Dismal Diversity Report
- TikTok CEO Promises More Transparency, Jabs Facebook for 'Copycat Product'
TikTok CEO Promises More Transparency, Jabs Facebook for 'Copycat Product'
TikTok is promising more transparency.
The Culver City-based social media platform will release its algorithms and content moderation policies, CEO Kevin Mayer wrote in an open letter Wednesday. Owned by China's ByteDance, the company has been facing pressure as speculation grows that its content is being shared with Beijing.
"We accept this and embrace the challenge of giving peace of mind through greater transparency and accountability," Mayer wrote. "We believe it is essential to show users, advertisers, creators, and regulators that we are responsible and committed members of the American community that follows US laws."
Earlier this month, the Trump administration said it was "looking at" banning the app over those concerns. Then last week, the House prohibited U.S. federal employees from downloading the app on government-issued devices.
It's also notable that his letter was published the same day Facebook's Mark Zuckerberg as well as the CEOs of Twitter, Google and Apple testified before Congress on antitrust law. In the letter, Mayer takes a swipe at Zuckerberg for this upcoming Reels product.
What you need to know:
- Calling it the Transparency and Accountability Center, TikTok will let experts view the company's data practices and algorithms.
- Investors of parent company ByteDance are now valuing the app at $50 billion — surpassing the projected 2020 revenue by 50 times. Some are pushing for ownership over the platform.
- Mayer took jabs at the other tech giants in the letter:
- "This puts us a step ahead of the industry, and we encourage others to follow suit."
- "At TikTok we welcome competition. We think fair competition makes all of us better. To those who wish to launch competitive products, we say bring it on."
- Then he hit Facebook's forthcoming Reels feature, calling it a "copycat product."
'We Must Do More': Snap Releases Dismal Diversity Report
While much of the tech world was fixated on a blockbuster congressional hearing of four executives from top tech companies, Snap Inc. quietly released its first report on diversity since the company was founded in 2011 and the numbers were dismal.
Evan Spiegel, CEO of Snap, which has faced allegations of a racist and sexist workplace, as recently as last month told employees he wouldn't release the numbers publicly. The company has good reason to try to bury the news.
Blacks only represent 4.1% of Snap's U.S. workforce while Hispanic/Latinx makeup 6.8%, far below their numbers in the general population. At the top, 2.6% of leadership roles are held by Blacks while seven percent are held by Hispanic/Latinx. Women make up 32.9% of Snap's global workforce but only 16.1% of tech teams and just 6.7% of tech teams' leadership.
"To date, our DEI (Diversity, Equity & Inclusion) outcomes simply have not been good enough," the company said in the report. "We must do more." Snap has set a goal of doubling the number of women in tech roles by 2023 and doubling the number of underrepresented minorities at the company by 2025.
While its diversity numbers are low, Snap is not much worse than other tech giants, though most of those companies have released their numbers for years.
Last month, Snap was forced to remove a Juneteenth filter that prompted users to smile in order to break a series of chains, and that was not the first time the company was criticized for an offensive filter.
"We deeply apologize for the offensive Juneteenth Lens," the company said in a tweet.
Snap outlined a number of steps to improve its numbers, including changes to recruiting, setting representation goals for underrepresented groups, and instituting a $70,000 minimum living wage for employees working at its Santa Monica headquarters.
Spotify Has More Listeners, but Ad Revenue Drops
Spotify's second quarter earnings, released today, show listening and podcast streaming up even as revenue missed the mark with the pandemic hurting ad sales.
The Swedish music streaming service acknowledged slower business in April and May across emerging regions. Still, Spotify said its strength in North America offsets the setback, noting that it turned a corner in June.
"We believe the improved momentum we saw in the back half of the quarter has continued into Q3 and we expect to hit our full year targets," the company told shareholders.
What you need to know:
- Ad revenue, which makes up less than 10% of total revenue, is down 21% from last year, a nod to dropping sales brought on by the pandemic.
- The average revenue per Spotify Premium user (ARPU) is also down 9% as a fewer percentage of users pay the standard $9.99/month rate. More are opting in for family and student plans. Users in some countries also pay a lower price for subscriptions.
- 13 million new monthly active members brings Spotify's total to an all-time high of 299 million. Plus, 8 million new subscribers brings the total of ad-free premium customers to 138 million.
- Overall listening times have returned to pre-pandemic levels in all regions except Latin America.
- More users are listening on at-home smart speakers and smart TVs.
- In an effort to diversify Spotify content and move away from music, podcast options are expanding — including exclusive deals with big names like Joe Rogan, Kim Kardashian, DC Franchise, the Obamas and TikTok star Addison Rae.
- Since the start of 2019, overall podcast listening has doubled since. One fifth of monthly active members are tuning in.
From Your Site Articles
- TikTok addresses 'tough but fair questions' about opportunities for ... ›
- PledgeLA Releases Report on Diversity in Los Angeles Tech - dot.LA ›
- TikTok's $2 Billion Creator Fund Goes After Instagram - dot.LA ›
- Microsoft Reportedly Looking to Buy TikTok ›
- Microsoft Reportedly Looking to Buy TikTok - dot.LA ›
- Snap Shares Soar on Upbeat Third Quarter Earnings - dot.LA ›
Related Articles Around the Web
Read moreShow less
Francesca Billington
Francesca Billington is a freelance reporter. Prior to that, she was a general assignment reporter for dot.LA and has also reported for KCRW, the Santa Monica Daily Press and local publications in New Jersey. She graduated from Princeton in 2019 with a degree in anthropology.
https://twitter.com/frosebillington
francesca@dot.la
Salt AI Secures $10M to Untangle Healthcare’s Toughest Workflows
09:22 AM | September 26, 2025
🔦 Spotlight
Hello Los Angeles,
Not every startup raise deserves the spotlight, but this week’s news from Salt AI is worth paying attention to. The LA based company just closed a $10 million round led by Morpheus Ventures with participation from Struck Capital, Marbruck Investments and CoreWeave. The goal is to expand what it calls “contextual AI,” and if it works, it could quietly change how some of the most complex corners of healthcare get untangled.
Healthcare is notorious for slow, clunky systems. Even the smallest workflow, like drug trial data, clinical documentation, or compliance reviews, can drag on for weeks because the tools were never built for speed. Salt AI is betting that the fix is not flashy consumer apps or billion parameter models, but something more practical: AI that slots directly into the day to day grind of life sciences. Their platform lets non technical teams visually build and deploy workflows that would normally take months of coding. Drag, drop, done.
It sounds simple, but the implications are not. Imagine a biopharma team testing a new drug, able to cut through compliance hurdles in days instead of months. Or clinical researchers spinning up experiments and seeing usable results in real time. Salt AI’s pitch is not about replacing scientists, it is about giving them back time in an industry where time can literally mean lives.
The new capital will help scale engineering, grow its customer footprint, and push further into healthcare and biopharma. But more importantly, it gives Salt AI the chance to prove that “contextual AI” is more than a buzzword. If they succeed, the company will not just chip away at bottlenecks, it could reshape how innovation itself moves through one of the world’s most heavily regulated and mission critical industries.
🤝 Venture Deals
LA Companies
- Bonsai Health raised $7M in a seed round led by Bonfire Ventures and Wonder Ventures. The Santa Monica based company builds an agentic AI platform that automates front office healthcare workflows, things like patient outreach, scheduling and clinical follow-ups, working behind the scenes to keep patients connected to care and reduce administrative burden. It plans to use the funding to accelerate its specialty AI agents, expand into new medical specialties, and scale its commercialization nationwide. - learn more
- Genstore raised a $10M Seed round led by Weimob, with participation from Lighthouse Founders’ Fund. The Los Angeles based startup is building an AI-native e-commerce platform that lets merchants launch and run online stores using conversational prompts, automating everything from product listings and copywriting to customer service. The funds will go toward accelerating product development, expanding into new markets, and refining features that simplify online commerce for small and midsized sellers. - learn more
- TransAstra secured a $5M investment to scale its asteroid capture technology in partnership with NASA. The company aims to advance systems that can snag and repurpose small bodies in space, contributing to sustainable space infrastructure and debris mitigation. With this funding, TransAstra will expand development, deepen its relationship with NASA, and accelerate deployment of its capture hardware. - learn more
LA Venture Funds
- Fika Ventures led a seed round investing in MaxHome, joining BBG Ventures, Four Acres and 1Sharpe Ventures. MaxHome is building an AI-native platform focused on automating real estate transaction coordination, the messy, manual work that slows deals. Fika backed the team because it sees a huge opportunity in streamlining broker workflows, reducing errors, and improving the experience for agents and homebuyers alike. - learn more
- MANTIS Ventures joined NEA, Sequoia, NVIDIA, J.P. Morgan and others in leading a $50M Series B for Factory, valuing the AI coding company at $300 million. Factory builds “droids,” AI agents that automate software development tasks across environments, and claims their platform now tops the Terminal Bench benchmark. With this capital, Factory aims to expand enterprise adoption, deepen integrations, and scale its engineering team globally. - learn more
- SafeHill (formerly Tacticly) announced a $2.6M pre-seed round led by Mucker Capital, with participation from Chingona Ventures, Techstars, Chicago Early Growth Ventures, The Source Groups, and others. The Chicago-based cybersecurity startup is launching from stealth with SecureIQ, a continuous Threat Exposure Management platform that blends AI-driven testing with human validation to help organizations find and shore up attack paths. The funding will be used to expand engineering, enhance AI-assisted ethical hacking, deepen enterprise partnerships, and broaden compliance and monitoring capabilities. - learn more
- Prototype Capital was among the investors in Nilo Technologies’ $4M seed round, alongside backers like Supercell, a16z Speedrun, KFund, and Flex Capital. Nilo is building an AI native 3D creation platform that makes game development more accessible, letting creators build interactive worlds in their browser without complex tooling. The funding will help accelerate product development, bring in more users as “Founding Builders,” and expand the platform’s capabilities for real time, multiplayer creation. - learn more
- Rebel Fund participated in a $7.5M funding round for Indian fintech Gold Firm Gullak backed by Y Combinator. Gullak offers digital gold savings and lending solutions targeted at underbanked consumers in India. Rebel Fund’s investment will help Gullak scale operations, deepen financial inclusion, and expand its product offerings. - learn more
- B Capital joined Wellington Management, General Catalyst and others in a $400M funding round for Capital Rx, which is rebranding as Judi Health. The company, which operates a pharmacy benefits management platform, will use the capital to expand into full-spectrum health benefits, integrating medical, dental and vision claims processing with its existing PBM capabilities. The move positions Judi Health as a unified tech backbone for benefits administration across employer and plan clients. - learn more
- Supply Change Capital joined a seed funding round that raised $4.7M for Helios AI, a startup building the first AI co-pilot for food and agriculture supply chains. Helios’ platform combines climate modeling, commodity forecasting, and real-time data to help buyers and suppliers make smarter decisions in volatile markets. The funding will be used to scale the product, expand data coverage globally, and bring its AI tools to more players across the agri-food sector. - learn more
Read moreShow less
RELATEDTRENDING
LA TECH JOBS