Connect with us

Uncategorized

An AI saw a cropped photo of AOC. It autocompleted her wearing a bikini.

Published

on

Language-generation algorithms are known to embed racist and sexist ideas. They’re trained on the language of the internet, including the dark corners of Reddit and Twitter that may include hate speech and disinformation. Whatever harmful ideas are present in those forums get normalized as part of their learning.

Researchers have now demonstrated that the same can be true for image-generation algorithms. Feed one a photo of a man cropped right below his neck, and 43% of the time, it will autocomplete him wearing a suit. Feed the same one a cropped photo of a woman, even a famous woman like US Representative Alexandria Ocasio-Cortez, and 53% of the time, it will autocomplete her wearing a low-cut top or bikini. This has implications not just for image generation, but for all computer-vision applications, including video-based candidate assessment algorithms, facial recognition, and surveillance.

Ryan Steed, a PhD student at Carnegie Mellon University, and Aylin Caliskan, an assistant professor at George Washington University, looked at two algorithms: OpenAI’s iGPT (a version of GPT-2 that is trained on pixels instead of words) and Google’s SimCLR. While each algorithm approaches learning images differently, they share an important characteristic—they both use completely unsupervised learning, meaning they do not need humans to label the images.

This is a relatively new innovation as of 2020. Previous computer-vision algorithms mainly used supervised learning, which involves feeding them manually labeled images: cat photos with the tag “cat” and baby photos with the tag “baby.” But in 2019, researcher Kate Crawford and artist Trevor Paglen found that these human-created labels in ImageNet, the most foundational image data set for training computer-vision models, sometimes contain disturbing language, like “slut” for women and racial slurs for minorities.

The latest paper demonstrates an even deeper source of toxicity. Even without these human labels, the images themselves encode unwanted patterns. The issue parallels what the natural-language processing (NLP) community has already discovered. The enormous datasets compiled to feed these data-hungry algorithms capture everything on the internet. And the internet has an overrepresentation of scantily clad women and other often harmful stereotypes.

To conduct their study, Steed and Caliskan cleverly adapted a technique that Caliskan previously used to examine bias in unsupervised NLP models. These models learn to manipulate and generate language using word embeddings, a mathematical representation of language that clusters words commonly used together and separates words commonly found apart. In a 2017 paper published in Science, Caliskan measured the distances between the different word pairings that psychologists were using to measure human biases in the Implicit Association Test (IAT). She found that those distances almost perfectly recreated the IAT’s results. Stereotypical word pairings like man and career or woman and family were close together, while opposite pairings like man and family or woman and career were far apart.

iGPT is also based on embeddings: it clusters or separates pixels based on how often they co-occur within its training images. Those pixel embeddings can then be used to compare how close or far two images are in mathematical space.

In their study, Steed and Caliskan once again found that those distances mirror the results of IAT. Photos of men and ties and suits appear close together, while photos of women appear farther apart. The researchers got the same results with SimCLR, despite it using a different method for deriving embeddings from images.

These results have concerning implications for image generation. Other image-generation algorithms, like generative adversarial networks, have led to an explosion of deepfake pornography that almost exclusively targets women. iGPT in particular adds yet another way for people to generate sexualized photos of women.

But the potential downstream effects are much bigger. In the field of NLP, unsupervised models have become the backbone for all kinds of applications. Researchers begin with an existing unsupervised model like BERT or GPT-2 and use a tailored datasets to “fine-tune” it for a specific purpose. This semi-supervised approach, a combination of both unsupervised and supervised learning, has become a de facto standard.

Likewise, the computer vision field is beginning to see the same trend. Steed and Caliskan worry about what these baked-in biases could mean when the algorithms are used for sensitive applications such as in policing or hiring, where models are already analyzing candidate video recordings to decide if they’re a good fit for the job. “These are very dangerous applications that make consequential decisions,” says Caliskan.

Deborah Raji, a Mozilla fellow who co-authored an influential study revealing the biases in facial recognition, says the study should serve as a wakeup call to the computer vision field. “For a long time, a lot of the critique on bias was about the way we label our images,” she says. Now this paper is saying “the actual composition of the dataset is resulting in these biases. We need accountability on how we curate these data sets and collect this information.”

Steed and Caliskan urge greater transparency from the companies who are developing these models to open source them and let the academic community continue their investigations. They also encourage fellow researchers to do more testing before deploying a vision model, such as by using the methods they developed for this paper. And finally, they hope the field will develop more responsible ways of compiling and documenting what’s included in training datasets.

Caliskan says the goal is ultimately to gain greater awareness and control when applying computer vision. “We need to be very careful about how we use them,” she says, “but at the same time, now that we have these methods, we can try to use this for social good.”

Lyron Foster is a Hawaii based African American Musician, Author, Actor, Blogger, Filmmaker, Philanthropist and Multinational Serial Tech Entrepreneur.

Continue Reading
Comments

Uncategorized

Freemium isn’t a trend — it’s the future of SaaS

Published

on

As the COVID-19 lockdowns cascaded around the world last spring, companies large and small saw demand slow to a halt seemingly overnight. Enterprises weren’t comfortable making big, long-term commitments when they had no clue what the future would hold.

Innovative SaaS companies responded quickly by making their products available for free or at a steep discount to boost demand.

While Zoom gets all the attention, there were hundreds of free SaaS tools to help folks through the pandemic. Pluralsight ran a #FreeApril campaign, offering free access to its platform for all of April. Cloudflare made its Teams product free from March until September 1, 2020. GitHub went free for teams in April and slashed the price of its paid Team plan.

A selection of new free, free trial and low-priced offerings from leading SaaS companies. Image Credits: Kyle Poyar/OpenView.

The free products were aimed squarely at end users — whether it be a developer, individual marketer, sales rep or someone else at the edge of an organization. These end users were stuck at home during the pandemic, yet they desperately needed software to power their working lives.

End users prefer to do the vast majority of their research online before ever talking to a sales rep, making free products the ideal way to reach them.

End users prefer to do the vast majority of their research online before ever talking to a sales rep, making free products the ideal way to reach them. Many end users want to jump straight into a product, no hassle or credit card or budget approval required.

After they’ve set up an account and customized it for their workflow, end users have essentially already made a purchase decision with their time — all without ever feeling like they were in an active buying cycle.

An end user-focused free offering became an essential SaaS survival strategy in 2020.

But these free offerings didn’t go away as lockdowns loosened up. SaaS companies instead doubled down on freemium because they realized that doing so had a real and positive impact on their business. In doing so, they busted the outdated myths that have held 82% of SaaS companies back from offering their own free plan.

Myth: A free offering will cannibalize paying customers

GoDaddy is a digital behemoth, known for being a ’90s-era pioneer in web domains as well as for their controversial Super Bowl ads. The company has steadily diversified into business software, now generating roughly $700 million in ARR from its business applications segment and reaching millions of paying customers. There are very few businesses that would see greater potential revenue cannibalization from launching a free product than GoDaddy.

But GoDaddy didn’t let fear stop them from testing freemium when lockdowns set in. Freemium started out as a small-scale experiment in spring 2020 for the websites and marketing product. GoDaddy has since increased the experiment to 50% of U.S. website traffic, with plans to scale to 100% of U.S. traffic and open availability to other markets in 2021.

Continue Reading

Uncategorized

Metafy adds $5.5M to its seed round as the market for games coaching grows

Published

on

This morning Metafy, a distributed startup building a marketplace to match gamers with instructors, announced that it has closed an additional $5.5 million to its $3.15 million seed round. Call it a seed-2, seed-extension or merely a baby Series A; Forerunner Ventures, DCM and Seven Seven Six led the round as a trio.

Metafy’s model is catching on with its market. According to its CEO Josh Fabian, the company has grown from incorporation to gross merchandise volume (GMV) of $76,000 in around nine months. That’s quick.

The startup is building in public, so we have its raw data to share. Via Fabian, here’s how Metafy has grown since its birth:

From the company. As a small tip, if you want the media to care about your startup’s growth rate, share like this!

When TechCrunch first caught wind of Metafy via prior seed investor M25, we presumed that it was a marketplace that was built to allow esports pros and other highly capable gamers teach esports-hopefuls get better at their chosen title. That’s not the case.

Don’t think of Metafy as a marketplace where you can hire a former professional League of Legends player to help improve your laning-phase AD carry mechanics. Though that might come in time. Today a full 0% of the company’s current GMV comes from esports titles. Instead, the company is pursuing games with strong niche followings, what Fabian described as “vibrant, loyal communities.” Like Super Smash Brothers, its leading game today in terms of GMV generated.

Why pursue those titles instead of the most competitive games? Metafy’s CEO explained that his startup has a particular take on its market — that it focuses on coaches as its core customer, over trainees. This allows the startup to focus on its mission of making coaching a full-time gig, or at least one that pays well enough to matter. By doing so, Metafy has cut its need for marketing spend, because the coaches that it onboards bring their own audience. This is where the company is targeting games with super-dedicated user bases, like Smash. They fit well into its build for coaches, onboard coaches, coaches bring their fans, GMV is generated model.

Metafy has big plans, which brings us back to its recent raise. Fabian told TechCrunch any game with a skill curve could wind up on Metafy. Think chess, poker or other games that can be played digitally. To build toward that future, Metafy decided to take on more capital so that it could grow its team.

So what does its $5.5 million unlock for the startup? Per its CEO, Metafy is currently a team of 18 with a monthly burn rate of around $80,000. He wants it to grow to 30 folks, with nearly all of its new hires going into its product org, broadly.

TechCrunch’s perspective is that gaming is not becoming mainstream, but that it has already done so. Building for the gaming world, then, makes good sense, as tools like Metafy won’t suffer from the same boom/bust cycles that can plague game developers. Especially as the startup becomes more diversified in its title base.

Normally we’d close by noting that we’ll get back in touch with the company in a few quarters to see how it’s getting on in growth terms. But because it’s sharing that data publicly, we’ll simply keep reading. More when we have a few months’ more data to chew on.

Continue Reading

Uncategorized

Snap to launch a new Creator Marketplace this month, initially focused on Lens Creators

Published

on

Snap on Wednesday announced its plan to soon launch a Creator Marketplace, which will make it easier for businesses to find and partner with Snapchat creators, including Lens creators, AR creators and later, prominent Snapchat creators known as Snap Stars. At launch, the marketplace will focus on connecting brands and AR creators for AR ads. It will then expand to support all Snap Creators by 2022.

The company had previously helped connect its creator community with advertisers through its Snapchat Storytellers program, which first launched into pilot testing in 2018 — already a late arrival to the space. However, that program’s focus was similar to Facebook’s Brand Collabs Manager, as it focused on helping businesses find Snap creators who could produce video content.

Snap’s new marketplace, meanwhile, has a broader focus in terms of connecting all sorts of creators with the Snap advertising ecosystem. This includes Lens Creators, Developers and Partners, and then later, Snap’s popular creators with public profiles.

Snap says the Creator Marketplace will open to businesses later this month to help them partner with a select group of AR Creators in Snap’s Lens Network. These creators can help businesses build AR experiences without the need for extensive creative resources, which makes access to Snap’s AR ads more accessible to businesses, including smaller businesses without in-house developer talent.

Lens creators have already found opportunity working for businesses that want to grow their Snapchat presence — even allowing some creators to quit their day jobs and just build Lenses for a living. Snap has been further investing in this area of its business, having announced in December a $3.5 million fund directed toward AR Lens creation. The company said at the time there were tens of thousands of Lens creators who had collectively made over 1.5 million Lenses to date.

Using Lenses has grown more popular, too, the company had noted, saying that more than 180 million people interact with a Snapchat Lens every day — up from 70 million daily active users of Lenses when the Lens Explorer section first launched in the app in 2018.

Now, Snap says that over 200 million Snapchat users interact with augmented reality on a daily basis, on average, out of its 280 million daily users. The majority (over 90%) of these users are 13 to 25-year-olds. In total, users are posting over 5 billion Snaps per day.

Snap says the Creator Marketplace will remain focused on connecting businesses with AR Lens Creators throughout 2021.

The following year, it will expand to include the community of professional creators and storytellers who understand the current trends and interests of the Snap user base and can help businesses with their ad campaigns. The company will not take a cut of the deals facilitated through the Marketplace, it says.

This would include the creators making content for Snap’s new TikTok rival, Spotlight, which launched in November 2020. Snap encouraged adoption of the feature by shelling out $1 million per day to creators of top videos. In March 2021, over 125 million Snapchat users watched Spotlight, it says.

Image Credits: Snapchat

Spotlight isn’t the only way Snap is challenging TikTok.

The company also on Wednesday announced it’s snagging two of TikTok’s biggest stars for its upcoming Snap Originals lineup: Charli and Dixie D’Amelio. The siblings, who have gained over 20 million follows on Snapchat this past year, will star in the series “Charli vs. Dixie.” Other new Originals will feature names like artist Megan Thee Stallion, actor Ryan Reynolds, twins and influencers Niki and Gabi DeMartino, and YouTube beauty vlogger Manny Mua, among others.

Snap’s shows were watched by over 400 million people in 2020, including 93% of the Gen Z population in the U.S., it noted.

 

Continue Reading

Trending