Latest News
Movies Anywhere, the Disney-owned app that lets users access their digitally owned movie collection from across services, is shutting down two features, “Screen Pass” and “Watch Together.”
Screen Pass, which launched in 2020, allows you to loan out three movies per month within the app, which recipients can then watch for up to 72 hours. Starting on May 1, users will no longer be able to share their purchased movies with friends and family. Cord Cutters News first noted the change.
“At Movies Anywhere, we are continually making changes to the website and app to help our users enjoy and grow their collections,” the company wrote on its website. “As the experience continues to evolve, we want to notify you that effective May 1, users will no longer be able to use the Screen Pass feature to send a Screen Pass. For Screen Passes sent prior to May 1, recipients will still be able to accept and finish watching the movie before their passes expire. As of June 1, the Screen Pass feature will no longer be supported.
In a separate post, Movies Anywhere announced that it would also be shutting down its “Watch Together” feature on June 1. Essentially a watch party capability, the feature is a synced viewing experience where you can send a room code or URL to up to nine friends to watch the same movie together at the same time.
The company didn’t state an exact reason for ending these features but said that it would focus on experiences that its users “are most passionate about” which includes expanding their movie collections and watching titles across multiple platforms. Basically, it sounds like not enough people were using Screen Pass or Watch Together.
TechCrunch reached out to Movies Anywhere for comment.
While co-viewing and movie sharing were once popular trends, especially during the pandemic, more people are going back to the movie theaters or having IRL watch parties in their homes. According to media measurement and analytics company Comscore, the 2023 box office topped $958 million in ticket sales as of February 27, compared to $98.7 million in 2021.
Also, the digital movie space just isn’t as trendy anymore and streaming services are getting their moment in the spotlight. The streaming market is a nearly $60 billion business, with giants like Netflix, Disney+ and HBO Max catering to millions of subscribers.
Movies Anywhere (formerly Disney Movies Anywhere) launched in 2014 and gives digital movie collectors a single hub where they can access all the movies they bought on iTunes, Vudu, Prime Video, YouTube and Xfinity, among other services. In 2017, it became jointly operated by Disney, Universal, WB, Sony Pictures and 20th Century Fox.
It’s most recent feature to launch was “My Lists,” an AI-powered feature that rolled out in 2021. My Lists automatically organizes movies together based on genre, cast, franchise and so on.
Digital locker app Movies Anywhere sunsets ‘Screen Pass’ and ‘Watch Together’ features by Lauren Forristal originally published on TechCrunch
Prog.ai users can build lists of top experts in specific disciplines, such as “large language models” or “computer vision,” and generate a leaderboard of top performers in any given field. Or they can submit a list of repositories and create a ranking of all the contributors by the number of commits that they have made.
Effectively, recruiters and companies can tailor their search to whatever parameters they want, including areas of skill, programming languages, and number of years of experience.
Prog.ai search example Image Credits: Prog.ai
But understanding code is only a part of Prog.ai’s offering.
A core selling-point for recruiters is the ability to connect with software developers, and for that Prog.ai packs a built-in email outreach engine, powered by sales engagement platform
Companies already have a wealth of tools at their disposal for headhunting technical talent, but a new startup wants to give recruiters a leg-up by bringing together the worlds of GitHub and LinkedIn to create a database of the most suitable candidates for a specific software development role — and it’s doing so by using AI to “infer” skills from code they’ve written.
Prog.AI, as the company is called, allows recruiters to search for developers based on their technical skills, libraries they have used, or simply the contributions they have made to projects on GitHub.
Founded out of San Francisco in 2022, Prog.AI is the brainchild of CEO Maria Grineva, who sold a previous data startup called Orb Intelligence to Dun & Bradstreet back in 2020; CTO Fedor Soprunov, previously a machine learning researcher at Russian tech titan Yandex; and product head Dmitry Pyanov, who has worked in product teams at companies including Yandex and Replika.
While hiring is the company’s primary focus initially, with its inaugural product opening for recruiters in closed beta this week, Grineva sees a broad gamut of use-cases beyond helping companies fill technical roles. This includes fostering developer relations, such as asking them to join a community or inviting them to contribute to an open source project; requesting their expertise for a specific problem; and even to help developer tool companies pitch their wares.
“This week we’re launching Prog.AI for tech recruiters, and in April we are going to extend our SaaS offering with Prog.AI for developer relations to help companies that build tools for developers to understand their TAM (total addressable market), learn more about their existing developer community, and reach their target audience,” Grineva explained to TechCrunch.
To help kickstart its commercial push, Prog.ai today announced that it has raised $1 million in pre-seed funding from Germany-based angel fund Angel Invest, Brooklyn Bridge Ventures, and a slew of angel backers including one of Spotify’s first employees and its former CTO Andreas Ehn.
So how does Prog.ai actually go about inferring skills from public source code? Well, in the first instance, the platform actions GitHub’s “git clone” command, which creates a copy millions of public repositories and branches. Prog.ai then analyzes each git commit, and inspects the code snippet, file-path, and the subject of the commit to figure out what it is about.
“For a given project, we can see who is the core architect, who develops the backend or frontend, who focuses on the UI/UX, who builds the QA and tests, and who are the technical writers,” Grineva said.
Prog.ai also pores over git actions such as pull requests including rejections and approvals, comments, and issue openings, which serves to help Prog.ai “understand” the different roles and engagement levels of the project contributors.
“We process not only famous open source projects, but also ‘pet’ projects, tests, forks, and even training projects from Coursera or Udemy that engineers keep public on GitHub,” Grineva added. “All together, we are processing about 1 billion commits on GitHub per year to get a very accurate profile of the skills of every engineer.”
Under the hood, Prog.ai leans on OpenAI’s GPT, tailoring the much-hyped language model on high-profile open source projects and StackOverflow articles to help it derive scores on code quality, for example.
Prog.ai profile example Image Credits: Prog.ai
Prog.ai users can build lists of top experts in specific disciplines, such as “large language models” or “computer vision,” and generate a leaderboard of top performers in any given field. Or they can submit a list of repositories and create a ranking of all the contributors by the number of commits that they have made.
Effectively, recruiters and companies can tailor their search to whatever parameters they want, including areas of skill, programming languages, and number of years of experience.
Prog.ai search example Image Credits: Prog.ai
But understanding code is only a part of Prog.ai’s offering.
A core selling-point for recruiters is the ability to connect with software developers, and for that Prog.ai packs a built-in email outreach engine, powered by sales engagement platform Reply.io.
“Users use our search to create a list of relevant candidates, and then they can create a personalized email sequence, mentioning candidates by name, referring to their projects, and explaining why they think a job position is a good fit for them,” Grineva said.
Prog.ai: Email outreach example Image Credits: Prog.ai
Recruiters will also probably want a more rounded view of a developer’s skills, education, and employment history, which they probably won’t get from GitHub. This is where LinkedIn enters the fray, with Prog.ai gleaning publicly-available data and aligning it with the corresponding individual from GitHub. And this is what Grineva says is the platform’s special sauce — by meshing data from two widely used platforms, it can build a finer-grained picture of potential candidates.
“I believe joining GitHub and LinkedIn profiles brings a lot of value, since engineers are typically not very good at promoting themselves and often don’t even have complete LinkedIn profiles,” Grineva said. “Furthermore, on LinkedIn, people self-describe themselves, which means that the information is subjective. Applying a standard methodology to infer the skills of all engineers based on their actual code contributions not only removes the subjectivity, but also means that companies will be able to evaluate candidates uniformly.”
Of course, none of this is offers a perfect recruitment conduit. Bringing together two gargantuan, disparate data sets is no easy feat, and there is likely a lot of room for error here, with similar names and histories raising the potential for conflating profiles. And that’s assuming that a person has a LinkedIn profile in the first place, which they absolutely might not. But under the hood, Grineva said they have put measures in place that go some way toward addressing at lease some of those potential pitfalls.
“Matching two large datasets is not an easy task, since the information people make available on GitHub can be sparse, with many engineers choosing to be anonymous on GitHub,” Grineva explained. “We have built a proprietary fuzzy-matching system that takes into account not only names, usernames and email addresses, but also matches places of work, expertise, interests.”
On top of that, Grineva said that they use computer vision to compare profile avatars across platforms, which while not fool-proof on its own, serves as an extra tool alongside its other verification mechanisms.
At the time of writing, Prog.ai claims to have the contact information from around 70% of all profiles in its database, which obviously means that 30% are lacking that crucial data. To that point, Grineva said that while they hope to improve its contact detail coverage as it expands, its potential use-cases won’t always revolve around reaching out.
“Another important use case is data-enrichment,” she said. “Customers can look up full candidate profile by GitHub handle, LinkedIn URL or contact email — in this case, we can only match to those 70% where we have the email.”
There’s also the giant elephant in the room here: isn’t Prog.ai simply facilitating “cold-callers” looking to contact developers en-masse?
“There is a risk, but it’s important to first recognize that recruiters are already trying to cold-call developers and this is currently happening via other tools, as well as some tech recruiters manually extracting contact information directly out of GitHub,” Grineva said. “That said, recruiters are currently doing this with bad or limited insights about the developers they are reaching out to, which means that the outreach is not personalized and often the opportunity is not a fit for the developers. As a result, these emails come across as spam.”
For those on the receiving end of a Prog.ai-powered reachout campaign, Grineva noted that the platform is “fully GDPR compliant,” and developers are able to ask it to remove or edit their profiles, as well as opt-out entirely from email outreach.
It’s still early days for Prog.ai and it’s experimenting with different plans, but the company is essentially operating a SaaS-based subscription model, with pricing based on the number of contacts a user accesses. This starts at “free” for up to 100 contacts per month, all the way up to a “recruiter” plan which is $530 per month for advanced search features and 3,000 contacts. It also offers an enterprise plan with custom pricing, which is available on request.
There’s also no ignoring the myriad other hiring solutions out there, spanning everything from LinkedIn’s very own Talent Solutions product, through Zoominfo, SeekOut, TalentOS, and HireEZ. But Grineva says Prog.ai’s focus purely on technical talent, and its GitHub scanning smarts, is what sets it apart from the crowd. In turn, this could mean better-targeted headhunting efforts, where a recruiter and candidate’s goals are more closely aligned.
“Being an engineer myself, I receive a lot of messages from recruiters that are not relevant for me and see this problem first-hand,” Grineva said. “I believe that this is primarily a data quality issue: recruiters just don’t have enough information about me to match me to interesting opportunities. Our goal is to reduce the level of noise developers receive today. By providing recruiters with better information, we believe that this will be a win-win for both developers and recruiters.”
Prog.ai wants to help recruiters find technical talent by inferring skills from GitHub code by Paul Sawers originally published on TechCrunch
Leslie Feinzaig, a venture capitalist, likes that her venture firm, Graham & Walker, sounds like an old, stodgy law firm. But apart from the name, there’s nothing really stodgy about it: Her fund exclusively invests in female- and nonbinary-founded startups.
It’s a relatively new name for her firm, which was originally called Female Founders Alliance. Feinzaig rebranded in 2021 in an effort to attract a more diverse set of founders and check-writers into her portfolio.
“The number one risk that we fall into is inadvertently stamping our own portfolio with a diversity signal,” she said. “And I mean that in the negative context of the word: We want our founders to stand on their own for being amazing founders. So what do we need to do? We need to become a super, high-signal VC.” In her view, that meant departing from a name that made her firm sound like it was making “diversity investments” and finding a name that didn’t include gender as a brand.
Now, she said, when she enters a room, “It’s very different to be Leslie, the CEO of Female Founder Alliance, than Leslie, managing director of Graham & Walker. Nobody questions it; it sounds like it belongs.”
That said, the investor still found a way to insert the mission into the nameKatharine Graham was the first female Fortune 500 CEO, and Madam C. J. Walker was the first female self-made millionaire.
The goal of being a VC is to generate returns for limited partners, and there’s an understanding that a diverse startup ecosystem will lead to better outcomes for all. Balancing those two, for female VCs, has often manifested in different, often frustrating ways.
A new generation of female venture capitalists is ditching institutional firms to start their own or steadily rising through leadership ranks. According to a survey analyzed by TC+, the share of women represented in director and principal positions has significantly increased over the past two years, while the percentage of women in higher-level positions, such as managing general partner or senior managing director, stands below 25% and has for the past two years. The ranks are diversifying. Slowly.
To put it simply: More women in venture means that bias and strategic branding are increasingly relevant for a larger fraction of check-writers.
For female VCs bias is a branding issue by Natasha Mascarenhas originally published on TechCrunch
https://techcrunch.com/2023/03/03/venture-capital-firm-rebrand/
Time is running out for you to score the biggest savings on passes to TechCrunch Disrupt 2023 — the original and always-evolving conference dedicated to early-stage startups. Beat the March 10 deadline, and you’ll save up to $1,000 on General Admission, Founder, and Investor passes. Students and nonprofits — grab a deeply discounted pass for just $95!
Get ready to join 10,000 attendees in San Francisco on September 19–21. Hit up the Disrupt stage for in-depth interviews with top-tier founders, CEOs, investors and tech-savvy celebrities — like Serena Williams and Kevin Hart — actively building or investing in startups.
You asked for more say in the programming and we heard you loud and clear. Last year, TechCrunch readers voted for the breakout sessions and roundtable discussions they wanted at Disrupt. It was a massive success, and we’re doing it again this year. Stay tuned — details to come!
In addition to amazing speakers, breakout sessions and roundtables, we’re bringing TC Sessions — our popular stand-alone industry events — to Disrupt! We’ll have six new stages featuring industry-specific programming tracks. Lean in to get the latest news, networking, topics, trends, startups and VCs within your sector. Check out the stages and the tech they cover.
Imagine the multiple cross-collaboration opportunities just waiting to be discovered — and now they’re all under one big roof!
TechCrunch Disrupt takes place in San Francisco on September 19–21, but if you want the best price and want to save up to $1,000, you need to buy your pass before prices increase. The deadline is Friday, March 10 at 11:59 p.m. PT, but why wait? Buy your Disrupt pass today!
Is your company interested in sponsoring or exhibiting at TechCrunch Disrupt 2023? Contact our sponsorship sales team by filling out this form.
Super early-bird savings to TC Disrupt ending soon by Lauren Simonds originally published on TechCrunch
https://techcrunch.com/2023/03/03/super-early-bird-savings-end-soon-techcrunch-disrupt-2023/
Hello and welcome back to Equity, a podcast about the business of startups, where we unpack the numbers and nuance behind the headlines.
This week Mary Ann Azevedo, Becca Szkutak, and Alex Wilhelm gathered to riff through the week’s biggest startup and venture news. A big thank you to Becca for stepping in while Alex was on leave, and a note before we dive into topics that Natasha will be back on the podcast next week!
Now, here’s what we got into:
And that is all we had time to chew on, friends. We will talk to you soon!
For episode transcripts and more, head to Equity’s Simplecast website.
Equity drops at 7:00 a.m. PT every Monday, Wednesday and Friday, so subscribe to us on Apple Podcasts, Overcast, Spotify and all the casts. TechCrunch also has a great show on crypto, a show that interviews founders, one that details how our stories come together and more!
Why hasn’t generative AI come up with something easier to say than ‘generative AI’? by Rebecca Szkutak originally published on TechCrunch
It’s only been a couple of months since OpenAI’s ChatGPT exploded into the public consciousness, and it already feels like our news feeds will never be the same again.
Whether it’s headlines about AI startups securing massive funding rounds or Twitter threads about how you should be using ChatGPT, the AI news cycle is well and truly here. Sorry, web3, you had your 15 minutes of fame.
Going from all-out rage prompted by the FTX fiasco to ChatGPT setting off the red alert at Google HQ made for a sudden, even shocking shift in the tech news cycle. Crypto publication Decrypt pointed out the focus hasn’t shifted only for the media: JPMorgan’s e-Trading Edit report noted that institutional traders are also looking carefully at AI while blockchain begins to lose its allure.
In this environment, it’s going to be extremely tempting for tech startups to quickly slap the words “AI” and “machine learning” wherever they’re vaguely applicable and dial up the newsworthiness of a given announcement or market insight.
Actually, that might not be a bad idea. In fact, it’s a huge opportunity to miss.
If AI-related coverage can get a new, unknown brand into its target publications today, it could help get the brand’s pitch deck in front of potential investors tomorrow.
Clearly, AI stories are going to have a relatively easier time catching reporters’ attention in this climate. That said, the need to differentiate messaging within the AI vertical is going to rise considerably with the influx of similar pitches heading to reporters’ inboxes.
The question is whether tech startups should shift their PR messaging toward AI-related topics. Such an approach is a given for startups that actually focus on AI: ChatGPT has paved the way and now they can reap the industrywide rewards. But for companies where AI was previously No. 4 on the list of proof points, machine learning capabilities should merge into the main hook of the announcement.
Startups that don’t have much to do with AI will likely fear accusations of “jumping on the bandwagon” if they wade into the discussion. Startups might think they should avoid the topic altogether unless they’re an all-out AI firm. The logic is for their PR messaging to stick closer to their core technology or brand mission and prioritize the longer-term benefits of clear positioning.
Startup PR professionals should be jumping on the AI bandwagon by Ram Iyer originally published on TechCrunch
https://techcrunch.com/2023/03/03/startup-pr-professionals-should-be-jumping-on-the-ai-bandwagon/
India’s central bank has slapped Amazon Pay’s India unit with a fine of over $373,300 for non-compliance with local guidelines surrounding know your customer and prepaid payment instruments norms.
The Reserve Bank of India said in a statement (PDF) that Amazon Pay (India) was non-compliant with certain provisions of the guidelines on prepaid payment instruments (required by a company to operate as a digital wallet) and know your customer issued by the central bank on KYC requirements, but did not elaborate precisely which rules were violated.
“Accordingly, notice was issued to the entity advising it to show cause as to why penalty should not be imposed for non-compliance with the directions. After considering the entity’s response, RBI concluded that the aforesaid charge of non-compliance with RBI directions was substantiated and warranted imposition of monetary penalty,” it said in a statement.
The fine comes at a time when the Indian central bank is toughening its compliance requirements for fintech and Big Tech firms in the country as it cracks down on money laundering and predatory business practices.
India is a key market for Amazon, which has deployed over $7 billion in the country over the past decade.
“We remain deeply committed to operating as per regulatory guidelines and maintaining a high compliance bar, while we innovate on behalf of our customers to offer them a safe and convenient payments experience. We continue to work closely with the authorities to share our commitment with them,” an Amazon spokesperson said in a statement.
India’s central bank slaps penalty on Amazon’s payments unit by Manish Singh originally published on TechCrunch
https://techcrunch.com/2023/03/03/reserve-bank-of-india-amazon-pay/
Japanese marketing tech firm Geniee, part of the SoftBank Group, has paid about $70 million in cash to acquire the revenue optimization platform Zelto (formerly known as AdPushup), a person familiar with the matter said, delivering 40 times return to a number of angel investors in the startup that began its journey in India.
The acquisition is a remarkable turnaround for Zelto, the 10-year-old firm that provides content creators and web publishers with tools to generate more revenue by tapping dozens of advertising exchanges, which faced near-death experiences twice in its journey. In 2014, the startup almost ran out of cash. Later, it scrambled with its product offerings after its marquee service struggled to make inroads, Zelto founder and chief executive Ankit Oberoi told TechCrunch in an interview.
The two firms announced the deal on Thursday, but did not disclose the terms of the deal. The source requested anonymity sharing private information. Oberoi of Zelto, which has been profitable for several years and raised less than $2.5 million in external funding during its startup journey, declined to disclose the terms of the deal. Elle, NDTV, Cnet, PCMag, Mashable, and GSMArena are among some of the customers of AdPushup, Zelto’s marquee offering, according to the company’s website.
“It’s been a rollercoaster ride. We had our fair share of challenges. But I think those challenging times were actually helpful,” said Oberoi. “After the launch, customers were leaving. Half of the team sort of left. But it forced us to look at the fundamental value we deliver to customers. And when we almost ran out of cash in 2014, we were forced to become frugal and profitable.”
The acquisition won’t significantly change how Zelto operates, he said. “It’s business as usual for us. But now that we have the support of a SoftBank-backed firm, we will invest a lot more in our growth and expand to Southeast Asian markets that we were not exploring earlier.” Zelto, which maintains a large team in India and started its journey in the country, identifies the U.S. as its headquarters and largest market.
“I am very happy to have this partnership that will greatly advance the purpose of Zelto and Geniee. My first encounter with Ankit dates back seven years. At the time, it was still a small company, but I invested in Ankit on the intuition that it would become a leading global entrepreneurship in the industry,” said Tomoaki Kudo, chief executive of Geniee, in a statement.
“I also feel a deep sense of respect for the fact that they have achieved high growth in the highly competitive cutting-edge markets of North America and India. In the future, we would like to learn from the North American and Indian markets, Zelto’s technology, services, knowledge, and corporate culture, and strongly promote the globalization of Geniee to realize its purpose.”
Japan’s Geniee acquires AdPushup-operator Zelto for $70 million by Manish Singh originally published on TechCrunch
https://techcrunch.com/2023/03/03/softbank-geniee-acquires-zelto-adpushup-for-70-million/
Twitter’s paid plan Twitter Blue is now available to more than 20 new countries in Europe. These countries include Netherlands, Poland, Ireland, Belgium, Sweden, Romania, Czech Republic, Finland, Denmark, Greece, Austria, Hungary, Bulgaria, Lithuania, Slovakia, Latvia, Slovenia, Estonia, Croatia, Luxembourg, Malta, and Cyprus.
This expansion makes the social network’s subscription service available in more than 35 countries across the world. Under Elon Musk, Twitter Blue was first launched back in December at an $8 per month price point with the Blue verification mark for paying users.
Later, the company introduced features like the ability to post 60-minute long videos, and 4,000-character long tweets, and get priority in conversations. The plan also has some legacy features like an edit tweet functionality, a thread reader, and bookmark folders.
To increase subscribers, Twitter launched an annual plan at $84 per year earlier in January. In the last few weeks, Twitter has also launched the annual plan on both iOS and Android priced at $114.
While Musk has been banking on subscription plans to bring in a ton of revenue, early results haven’t been encouraging. According to estimates and reports, the new Twitter Blue service only has less than 300,000 subscribers.
Earlier this week, many users had problems accessing various parts of Twitter including timelines, searches, and direct messages. These issues occurred after Musk reportedly fired more than 200 people in the latest round of job slashing.
Twitter Blue expands to more than 20 countries by Ivan Mehta originally published on TechCrunch
https://techcrunch.com/2023/03/02/twitter-blue-expands-to-more-than-20-countries/
Developing cutting-edge AI systems like ChatGPT requires massive technical resources, in part because they’re costly to develop and run. While several open source efforts have attempted to reverse-engineer proprietary, closed source systems created by commercial labs such as Alphabet’s DeepMind and OpenAI, they’ve often run into roadblocks — mainly due to a lack of capital and domain expertise.
Hoping to avoid this fate, one community research group, EleutherAI, is forming a nonprofit foundation. The organization today announced it’ll found a not-for-profit research institute, the EleutherAI Institute, funded by donations and grants from backers, including AI startups Hugging Face and Stability AI, former GitHub CEO Nat Friedman, Lambda Labs and Canva.
“Formalizing as an organization allows us to build a full time staff and engage in longer and more involved projects than would be feasible as a volunteer group,” Stella Biderman, an AI researcher at Booz Allen Hamilton who will co-run the EleutherAI Institute, told TechCrunch in an email interview. “In terms of a nonprofit specifically, I think it’s a no-brainer given our focus on research and the open source space.”
EleutherAI started several years ago as a grassroots collection of developers working to open source AI research. Its founding members — Connor Leahy, Leo Gao and Sid Black — wrote the code and collected the data needed to create a machine learning model close to OpenAI’s text-generating GPT-3, which at the time was getting a lot of press.
The company curated and open sourced The Pile, a collection of datasets designed to be used to train GPT-3-like models to complete text, write code and more. And it released several models under the Apache 2.0 license, including GPT-J and GPT-NeoX, language models that for a while fueled an entirely new wave of startups.
To train its models, EleutherAI relied mostly on the TPU Research Cloud, a Google Cloud program that supports projects with the expectation that the results will be shared publicly. CoreWeave, a U.S.-based cryptocurrency miner that provides cloud services for AI workloads, also supplied compute resources to EleutherAI in exchange for models its customers can use and serve.
EleutherAI grew quickly. Today, over 20 of the community’s regular contributors are working full-time, focusing mainly on research. And over the past 18 months, EleutherAI members have co-authored 28 academic papers, trained dozens of models and released ten codebases.
But the fickle nature of its cloud providers sometimes forced EleutherAI to scuttle its plans. Originally, the group had intended to release a model roughly the size of GPT-3 in terms of the number of parameters, but ended up shelving that roadmap for technical and funding reasons. (In AI, parameters are the parts of the model learned from historical training data and essentially define the skill of the model on a problem, such as generating text.)
In late 2022, EleutherAI became well-acquainted with Stability AI, the now-well-financed startup behind the image-generating AI system Stable Diffusion. Along with other collaborators, it helped to create the initial version of Stable Diffusion. And since then, Stability AI has donated a portion of compute from its AWS cluster for EleutherAI’s ongoing language model research.
After another big patron — Hugging Face — approached EleutherAI and nonprofit discussions kicked off, Biderman says. (Many EleutherAI staff were involved with the company’s BigScience effort, which sought to train and open source a model akin to GPT-3 over the course of a year.)
“EleutherAI has largely focused on large language models that are architecturally similar to ChatGPT in the past, and will likely continue to do so,” Biderman said. “Beyond training large language models, we are excited to devote more resources to ethics, interpretability and alignment work.”
One might wonder whether the involvement of commercially motivated ventures like Stability AI and Hugging Face — both of which are backed by substantial venture capital — might influence EleutherAI’s research. It’s a natural assumption — and it’s even evidence-backed. At least one study shows a direct correlation between donations and the likelihood that nonprofits speak up about a proposed government rule.
Biderman asserts that the EleutherAI Foundation will remain independent and says she doesn’t see a problem with the donor pool so far.
“We don’t develop models at the behest of commercial entities,” Biderman said. “If anything, I think that having a diverse sponsorship improves our independence. If we were fully funded by one tech company, that seems like a much bigger potential issue from our end.”
Another challenge the EleutherAI Foundation will have to overcome is ensuring its coffers don’t run dry. OpenAI is a cautionary tale; after being founded as a nonprofit in 2015, the company later transitioned to a “capped-profit” structure in order to fund its ongoing research.
Broadly speaking, nonprofit initiatives to fund AI research have been a mixed bag.
Among the success stories is the Allen Institute for AI (AI2), founded by the late Microsoft co-founder Paul Allen, which aims to achieve scientific breakthroughs in AI and machine learning. There’s also the Alan Turing Institute, the U.K.-based, government-funded national institute for data science and machine learning. Smaller promising efforts include AI startup Cohere’s Cohere For AI (despite its corporate ties) and Timnit Gebru’s Distributed AI Research, a global distributed research organization.
But for every AI2, there’s former Google chairman Eric Schmidt’s fund for AI research. Over $125 million in size, it attracted fresh controversy after Politico reported that Schmidt wields an unusually heavy sway over the White House Office of Science and Technology Policy.
Time will tell which direction the EleutherAI Foundation ultimately takes. Likely, the mission will evolve and change over time — in positive ways, we can only hope.
Stability AI, Hugging Face and Canva back new AI research nonprofit by Kyle Wiggers originally published on TechCrunch
