Month: December 2022

Released two years ago, OpenAI’s remarkably capable, if flawed, GPT-3 was perhaps the first to demonstrate that AI can write convincingly — if not perfectly — like a human. The successor to GPT-3, most likely called GPT-4, is expected to be unveiled in the near future, perhaps as soon as 2023. But in the meantime, OpenAI has quietly rolled out a series of AI models based on “GPT-3.5,” a previously-unannounced, improved version of GPT-3.

GPT-3.5 broke cover on Wednesday with ChatGPT, a fine-tuned version of GPT-3.5 that’s essentially a general-purpose chatbot. Debuted in a public demo yesterday afternoon, ChatGPT can engage with a range of topics, including programming, TV scripts and scientific concepts.

According to OpenAI, GPT-3.5 was trained on a blend of text and code published prior to Q4 2021. Like GPT-3 and other text-generating AI, GPT-3.5 learned the relationships between sentences, words and parts of words by ingesting huge amounts of content from the web, including hundreds of thousands of Wikipedia entries, social media posts and news articles.

Rather than release the fully trained GPT-3.5, OpenAI used it to create several systems fine-tuned for specific tasks — each available through the OpenAI API. One — text-davinci-003 — can handle more complex instructions than models built on GPT-3, according to the lab, and is measurably better at both long-form and “high-quality” writing.

According to OpenAI data scientist Jan Leike, text-davinci-003 is similar but not identical to InstructGPT, a family of GPT-3-based models released by OpenAI earlier this year that are less likely to generate problematic (e.g., toxic and highly biased) text while more closely aligning with a user’s intent. Text-davinci-003 — and by extension GPT-3.5 — “scores higher on human preference ratings” while suffering from “less severe” limitations, Leike said in a tweet.

That appears to be the case anecdotally. Data scientists at Pepper Content, a content marketing platform, report that text-davinci-003 “performs better in understanding the ‘context’ behind a request and then using that to produce better content” while “hallucinating” less than GPT-3-based models. (Where it concerns text-generating AI, hallucination refers to an AI writing inconsistent, factually incorrect statements.)

In a test on OpenAI’s Playground website, which provides a UI frontend for the models, the Pepper Content team fed several prompts to text-davinci-003 and a model based on GPT-3 (text-davinci-002). Given “What is the philosophy behind WeWork?,” the GPT-3.5-based text-davinci-003 generated this:

WeWork is a global workspace provider that believes people are the most important asset in any organization. The philosophy of WeWork is to create a collaborative environment that enables people to work together in a flexible and efficient way. The company encourages collaboration and productivity, while providing a comfortable and inspiring space.

WeWork is also committed to being a socially responsible organization, by finding ways to reduce its environmental impact, by providing meaningful work experiences, and by promoting diversity and inclusion. WeWork also strives to create meaningful experiences for its members, through its unique community-based programming, events and activities. The company believes that when people work together in an inspiring and collaborative environment, they can achieve more and create meaningful change.

It’s not perfect — note the excess commas and repetitiveness. But the copy’s certainly more engaging than what the GPT-3-based text-davinci-002 produced:

At WeWork, we believe in openness, collaboration, and community. We have reimagined what a workspace can be by bringing together a global community of creators, entrepreneurs, and startups — anyone looking to build something meaningful and transform the world. We are proud to be your neighbour, coworker, and friend.

GPT-3.5 is also better at generating blog posts, it seems. Here’s what the Pepper Content team got when they prompted text-davinci-003 to write a post about picking a sofa:

If you are looking to give your green living room a fresh look, adding the the best sofa for the space is a great way to start. However, there is more to choosing the perfect sofa than just finding a piece that looks good. You want to find a piece that is comfortable and fits your green living room décor.

In this blog, we will discuss how to pick sofa far your green living room. We will look at the different types of sofas available, the design factors to consider when making your selection, and how to choose a sofa that fits your style and budget. We will also provide tips on how to maintain and care for your new sofa. By the end of this blog, you will have all the information you need to pick the perfect for your green living room.

Again, it isn’t foolproof. GPT-3.5 oddly added the bit about a “green living room.” But also again, GPT-3 is more basic and less grammatical in its generation:

Sofa is one of the most basic requirements in a living room. It’s not just a piece of furniture but an important part of the décor of your living room. So, what should be the criteria while picking a sofa? If you are wondering about this then stay with me as I discuss the different aspects of the sofa would help you in picking the best one for yourself.

Experiments beyond Pepper Content’s suggest that GPT-3.5 tends to be much more sophisticated and thorough in its responses than GPT-3.

For example, when YouTube channel All About AI prompted text-davinci-003 to write a history about AI, the model’s output mentioned key luminaries in the field, including Alan Turing and Arthur Samuelson, while text-davinci-002”s did not. All About AI also found that text-davinci-003 tends to have a more nuanced understanding of instructions, for instance providing details such as a title, description, outline, introduction and recap when asked to create a video script.

A hallmark feature of text-davinci-003/GPT-3.5’s generations is wordiness, as it turns out. In an analysis, scientists at startup Scale AI found text-davinci-003/GPT-3.5 generates outputs roughly 65% longer than text-davinci-002/GPT-3 under identical prompts.

Perhaps less useful but nonetheless entertaining, text-davinci-003/GPT-3.5 is better at composing songs, limericks and rhyming poetry than its predecessor. Ars Technica reports that commenters on Y Combinator’s Hacker News forum used text-davinci-003 to write a poem explaining Albert Einstein’s theory of relativity and then re-write the poem in the style of John Keats. See:

If you want to understand Einstein’s thought
It’s not that hard if you give it a shot
General Relativity is the name of the game
Where space and time cannot remain the same
Mass affects the curvature of space
Which affects the flow of time’s race
An object’s motion will be affected
By the distortion that is detected
The closer you are to a large mass
The slower time will seem to pass
The farther away you may be
Time will speed up for you to see

The Scale AI team even found that Text-davinci-003/GPT-3.5 has a notion of meters like iambic pentameter. See:

O gentle steeds, that bear me swift and

sure

Through fields of green and pathways so

obscure,

My heart doth swell with pride to be with

you

As on we ride the world a-fresh to view

The wind doth whistle through our hair so

free

And stirs a passion deep inside of me.

My soul doth lift, my spirits soar on high,

To ride with you, my truest friend, am I

Your strength and grace, your courage and

your fire,

Inspire us both to go beyond our sire.

No earthly bonds can hold us, only fate,

To gallop on, our wond’rous course create

Relatedly, text-davinci-003/GPT-3.5 is wittier — at least subjectively. Asking text-davinci-002/GPT-3 to “tell a joke” usually yields this:

Why did the chicken cross the road? To get to the other side.

Text-davinci-003/GPT-3.5 has cleverer responses:

Q: What did the fish say when it hit the wall? A: Dam!

Q: What did one ocean say to the other ocean? A: Nothing, they just waved.

Scale AI had it explain Python code as Eminem, a feat which text-davinci-002/GPT-3 couldn’t accomplish:

Yo, so I’m loopin’ through this list

With each item that I find

I’m gonna print out every letter in each one

of them

Dog, Cat, Banana, Apple, I’m gonna get’em

all with this rhyme

So why is GPT-3.5 so much better than GPT-3 in these particular areas? We can’t know the exact answer without additional details from OpenAI, which aren’t forthcoming; an OpenAI spokesperson declined our request for comment. But it’s safe to assume that GPT-3.5’s training approach had something to do with it. Like InstructGPT, GPT-3.5 was trained with the help of human trainers who ranked and rated the way early versions of the model responded to prompts. This information was then fed back into the system, which tuned its answers to match the trainers’ preferences.

Of course, it doesn’t make GPT-3.5 immune to the pitfalls to which all language models eventually succumb. Because GPT-3.5 merely relies on statistical regularities in its training data rather than a human-like understanding of the world, it’s still prone to, in Leike’s words, “mak[ing] stuff up a bunch.” It also has limited knowledge of the world after 2021 because the training data is more sparse after that year. And its safeguards against toxic output can be straightforwardly circumvented.

Still, GPT-3.5 and its derivative models demonstrate that GPT-4 — whenever it arrives — won’t necessarily need a huge number of parameters to best the most capable text-generating systems today. (Parameters are the parts of the model learned from historical training data and essentially define the skill of the model on a problem.) While some have predicted that GPT-4 will contain over 100 trillion parameters — nearly 600 times as many as GPT-3 — others argue that emerging techniques in language processing, like those seen in GPT-3.5 and InstructGPT, will make that enormous jump unnecessary.

One of those techniques could involve browsing the web for greater context, a la Meta’s ill-fated BlenderBot 3.0 chatbot. John Shulman, a research scientist and co-founder of OpenAI, told MIT Tech Review that OpenAI is continuing work on a language model it announced late last year, WebGPT, that can go and look up information on the web (via Bing) and give sources for its answers. At least one Twitter user appears to have found evidence of the feature undergoing testing for ChatGPT.

OpenAI has another reason to pursue lower-parameter models as it continues to evolve GPT-3: huge costs. A 2020studyfromAI21 Labs pegged the expenses for developing a text-generating model with only 1.5 billion parameters at as much as $1.6 million. OpenAI has raised over $1 billion to date from Microsoft and other backers, and it’s reportedly in talks to raise more. But all investors, no matter how big, expect to see returns eventually.

While anticipation builds for GPT-4, OpenAI quietly releases GPT-3.5 by Kyle Wiggers originally published on TechCrunch

In the older days of the web, an app called Trillian emerged to solve the problem of allowing users to connect to their friends across multiple messaging services, like AIM, Jabber, Yahoo! Messenger, IRC and others.

Lately, I’ve been longing for a Trillian-like service that would allow me to keep up with Twitter’s fleeing user base, the way that Trillian once connected users to multiple IM services.

I’m not sure that would be a good thing, though.

In the days since the Elon Musk Twitter takeover, a good chunk of Twitter’s network has exited to other apps. That leaves Twitter users who are trying to stay connected with friends with a lot of open browser tabs and new apps stuffed into iPhone folders.

The problem is, the market hasn’t yet picked a winner for “new Twitter” — and it’s possible it never will.

It’s a tough problem to solve. Reproducing Twitter’s magic isn’t as simple as re-creating a microblogging tool with similar functionality — a heart, a retweet, an @ reply and threads. Twitter’s biggest draw was its community, not its feature set. And it’s fair to say that network has been diminished in recent days — despite Musk’s tweets to the contrary. Even if Twitter is gaining users, as he claims, it’s losing others at the same time.

If Musk succeeds, Twitter’s sheer numbers may grow, but it won’t be the same Twitter it was before. Much like the company itself, when the original people are gone, so is the vibe, the spark, the sense of belonging. MySpace, after all, is still around. But it’s not MySpace anymore.

Or, in other words: though Twitter lives on, in some ways, it’s already gone.

You’ve probably felt it, too. Every day since the Musk takeover, much of my Twitter engagement now comes from crypto trolls who are tagging me into threads and bizarro, hashtagged praise for Musk from people’s accounts that aren’t quite discernible from bots. Meanwhile, real conversations are happening elsewhere — and with increasing frequency.

Twitter, meanwhile, seems to be doubling down on its reward-and-punishment system for only the most performative, most snarky, most zingy of posters, led by twit-in-chief Musk. Ultimately, it seems, Twitter may live on as an entertainment network, if not a community, as a result.

In the meantime, many of Twitter’s OG users have fled to different parts of the web.

Those with more technical expertise may have managed to navigate the complicated onboarding process to find a home on Mastodon, but it’s not a Twitter clone despite the initial similarities. Universal search doesn’t exist. Direct messages work differently. Quote tweets aren’t a thing. The mobile app is slow. Adding friends is wonky. And there are so many rules. (You’re even supposed to add content warnings to talk about Twitter, it’s so despised over there! And yet Twitter users’ collective sadness over its seeming downfall is all we want to talk about now. )

Elsewhere, some of Twitter’s media crowd have been trying out Post.News, which is being advised by tech journo Kara Swisher. But it’s not launched publicly yet and there’s no mobile app. Plus, the network itself, with its journalist base, is sort of …serious and boring. (Sorry, but it wasn’t media Twitter that made Twitter fun!)

Some young people left Twitter for Hive, whose 3-person team can’t keep the darned thing functioning. Currently, it’s offline as the team addresses significant security concerns. The company’s lack of transparency around staffing, funding, security, privacy and other key aspects of its business — not to mention its tendency to call everyone “besties” — doesn’t inspire much trust, either.

Then there are the numerous alt Twitters: Cohost, CounterSocial, Tribel, Koo, and so on.

Oh, and I guess we’re using Tumblr again, too? Which may one day connect to Mastodon folks? Substack wants to be a Twitter now? And Discord?

But wait! There’s more. Twitter co-founder Jack Dorsey is building his own decentralized Twitter called BlueSky, which is both a protocol and an app. But BlueSky says it won’t use ActivityPub — which means it won’t be able to talk to Mastodon. Sigh.

Even as all the alt Twitter sampling continues, many Twitter users haven’t actually fully left — they’ve just reduced their postings or are cross-posting. They’ve got one foot out of the door but keep looking back over their shoulder, as if asking themselves, should I really leave? Is now the time? (Points at self). After all, it seems there’s still too much going on over on Twitter to truly sign out forever.

I’m tired. This problem is in need of a solution. Or at least a decision. Can we convene a vote, please?

Eventually — hopefully — the issue will resolve itself.

Maybe we’ll collectively decide that we actually don’t choose which companies to use based on our morals and values, as it turns out. (Or else our local mom-and-pop stores would have our business, not Amazon Prime!) Maybe, with heads hung low, we’ll go back to Twitter and pretend we never left.

Or maybe we’ll all finally gravitate towards one network over the others, which will then gain enough traction to consider itself the real “Twitter 2.0.”

Or maybe we’ll decide we don’t need a Twitter 2.0, and stay dispersed across smaller networks. Maybe that’s the right move. Maybe it’s time to retire the concept of the global town square, the dunk factory, and the algorithm.

Maybe we should find a group of people we actually want to talk to — not perform for — and leave Twitter behind.

What’s a Twitter user to do? by Sarah Perez originally published on TechCrunch

Launch provider Rocket Lab is establishing a subsidiary to handle its sensitive U.S. government business, like launching spysats and experimental military spacecraft. Rocket Lab National Security LLC will handle most of the Defense Department stuff going forward to save the rest of the company a bit of grief.

The main arm of Rocket Lab, or perhaps it would be more appropriate to call it the body, has done plenty of business with the U.S. government already, putting tests from DARPA and the National Reconnaissance Office into orbit. It’s kind of inevitable if you’re a launch provider — governments in general are big customers, and the U.S. in particular.

But wow, you want to talk about picky? Try getting a contract to launch a top secret satellite that costs $200 million! There are all kinds of hoops you have to jump through. But then they want to launch right away, price is no object — as if you’ve just got rockets lying around!

Of course in Rocket Lab’s case that might well be true, but the company clearly decided that it would be smart to contain the “bureaucratic requirements” and other red tape, clearances, etc to a specialized subsidiary that can work more closely with its national security clients and partners. It’s likely there was already a considerable firewall within the company, since commercial activity is in many ways fundamentally different from government contracts. Now it’s truly its own business unit (or perhaps that is not the correct term of art, but you know what I mean).

“Top of the list for national security is reliability and responsiveness, something we’ve delivered on across multiple missions already. With Rocket Lab National Security we’re building on this strong heritage to deliver tailored capabilities that evolve as the nation’s needs do,” said Brian Rogers, leader of the company’s government launch services department, in a press release.

Less high-touch missions for civilian agencies like NASA and NOAA would likely use the “vanilla” Rocket Lab’s services, and indeed defense projects that don’t come with too many strings attached can probably save a buck or two by going that way as well. But compared with launch costs ten years ago it’s all a rounding error.

The launch company’s work designing spacecraft and on more complex missions like CAPSTONE remain part of the main company as well. Until the military starts launching space lasers to cislunar space, anyway.

Rocket Lab carves off defense-focused subsidiary for national security customers by Devin Coldewey originally published on TechCrunch

The cybercriminals behind the Medibank ransomware attack have published what appears to be the rest of the data stolen from the Australian health insurance giant.

The attackers, which are believed to be linked to the Russian-backed REvil ransomware gang, posted an update to its dark web blog in the early hours of Thursday morning, saying: “Happy Cyber Security Day!!! Added folder full. Case closed.”

The dark web blog was unavailable at the time of writing, but according to Medibank, the “full” folder contained six zipped files of raw data. At more than six gigabytes in size, the cache is much larger than any of the attackers’ previous Medibank leaks. Medibank confirmed in November that the attackers took 9.7 million customers’ personal details and health claims data for almost 500,000 customers.

The Medibank cybercriminals previously published data including customers’ names, birth dates, passport numbers, information on medical claims and sensitive files related to abortions and alcohol-related illnesses. Portions of the data seen by TechCrunch also appear to include correspondence between the cybercriminals and Medibank CEO David Koczkar, including a message in which the hackers threaten to leak “keys for decrypting credit cards,” despite Medibank’s assertion that no banking or credit card details were accessed.

The cybercriminals claimed they published the data after Medibank refused to pay their $10 million ransom demand, which was later reduced to $9.7 million, or $1 per affected customer.

Medibank said on Thursday that it is in the process of analyzing the latest leaked data but said it “appears to be the data we believed the criminal stole.”

“While our investigation continues there are currently no signs that financial or banking data has been taken,” Medibank said. “And the personal data stolen, in itself, is not sufficient to enable identity and financial fraud. The raw data we have analyzed today so far is incomplete and hard to understand.”

Although it’s believed the hackers have released all of the data stolen from Medibank, the company added that it expects “the criminal to continue to release files on the dark web.”

The Australian health insurance giant is urging customers to be vigilant with all online communications and transactions and to be alert for phishing scams related to the breach. Medibank added that to strengthen its security, it has this week added two-factor authentication in its contact centers to verify the identity of customers.

While Medibank is taking steps to shore up its cybersecurity, the company could face major financial penalties after the Australian parliament this week passed legislation that paves the way for businesses to be fined up to $50 million for repeated or serious data breaches.

Australia’s data and privacy watchdog, the Office of the Australian Information Commissioner (OAIC) on Thursday announced that it had begun an investigation into the personal information handling practices of Medibank. The OAIC — also investigating the recent Optus breach — said its investigation will focus on whether Medibank took reasonable steps to protect the personal information they held from misuse, interference, loss, unauthorized access, modification or disclosure.

“If the investigation finds serious and/or repeated interferences with privacy in contravention of Australian privacy law, then the Commissioner has the power to seek civil penalties through the Federal Court of up to $2.2 million for each contravention,” the OAIC said.

News of the investigation comes after the Australian Federal Police (AFP) said in November that it knows the identity of the individuals responsible for the attack on Medibank. The agency declined to name the individuals but said the police believe that those responsible for the breach are based in Russia, though some affiliates may be in other countries. The Russian Embassy in Canberra rebuffed the allegations.

Though their identities remain unknown, the attackers responsible already appear to be moving on from the Medibank hack. In recent days the group has posted new victims to its dark web blog, including New York-based medical group Sunknowledge Services and the Kenosha Unified School District.

Medibank hackers declare ‘case closed’ as trove of stolen data is released by Carly Page originally published on TechCrunch

Apple iPhone users with iOS 16.1 software can now share car keys in their Wallet with non-iPhone users, starting with Google Pixel devices. In the future, that capability will extend to other devices with Android 12+, according to Apple.

Apple says it has been working with the Internet Engineering Task Force and other industry players to establish a standard for sharing digital keys across platforms. Keys can be shared via email, text message and WhatsApp.

When Apple initially launched digital car keys in 2020, iPhone users could share their keys through iMessage, Apple’s instant messaging service.

Of course, key sharing only works on cars that are compatible with digital car keys. Apple didn’t share a list of compatible cars, and instead directed users to contact their car manufacturer or dealership for an answer. To our knowledge, there are several 2021 and 2022 models of BMW that can be unlocked and used via digital car keys on iPhones, Google Pixels, Samsung Galaxy devices. Both Apple’s and Google’s digital car keys also recently arrived on the 2022 Kia Niro, the 2022 Genesis G90 and the all-electric 2022 Genesis GV60.

Apple’s digital car keys can usually be added to the Apple Watch (Series 5 or later or the Apple Watch SE) with the latest version of watchOS, according to a post from Apple. It’s not clear if this cross-platform capability will be available for Apple Watch users just yet.

Apple’s digital car keys are one of the company’s many features that allow users to carry less stuff around. Perhaps one of the company’s most popular examples of this is Apple Pay, which syncs a user’s credit, debit or prepaid card to their phone. In addition, iPhone users in certain states can now carry their driver’s license or state ID in their Wallet, and in select Hyatt hotels, users can unlock their rooms with a digitally stored key.

iPhone users can share car keys in Wallet with non-iPhone users by Rebecca Bellan originally published on TechCrunch

It might seem like Lensa AI sprung up over night when suddenly, your friends are all posting artistic renditions of themselves that they generated on the app. But while the Lensa AI app itself has been around since 2018, its viral “magic avatars” feature launched in late November, boosting the app to the #1 spot on the iOS App Store‘s competitive “Photo & Video” charts. For comparison, YouTube is #3 and Instagram is #4 on the charts at the time of publication.

Lensa AI works by inviting users to upload 10-20 photos of themselves. Using the open source Stable Diffusion model, the app processes your photos to generate avatars of you that look like they were created by a digital artist.

When you download Lensa AI, you’re immediately greeted with a pop-up inviting you to join a seven-day free trial to use their AI editing tools — if you don’t cancel in time, you’ll be charged $39.99 for unlimited use of the app for a year. You can bypass this screen without committing to the trial, but the free version of the app is very limited and doesn’t include the viral magic avatar feature. Even if you don’t subscribe to the unlimited plan, you’ll be hit with another in-app purchase screen if you try to make your very own magic avatar. At the lowest price tier, you’ll have to pay $3.99 for 50 unique avatars (5 variations of 10 different styles). But hey, at least that’s cheaper than a blue check!

According to app analytics firm SensorTower, Lensa AI has amassed about 22.2 million worldwide downloads and almost $29 million in consumer spending since its launch in 2018. In the last month, the app has seen a significant spike with the release of magic avatars. In November, the app was downloaded 1.6 million times, up 631% from 219,000 downloads in October. The U.S. is Lensa AI’s largest market, generating 58% of consumer spending, but the app is especially popular in Brazil right now. Of all November downloads, 31% were from Brazil, where installs climbed 24,450% month over month.

Viral photo editing apps have a shady history, as some apps have been found to be vectors for malware. In other cases, users have worried about what happens to the photos they upload into these apps. These concerns came up around Russia-based AI editor FaceApp, which later made a statement that it might store updated photos in the cloud for “performance and traffic reasons,” but that most images are deleted within 48 hours.

Prisma Labs, the team behind Lensa AI, told TechCrunch that it uses AWS cloud services to process user’s photos. As soon as an AI model is trained on a user’s photos, the images are immediately deleted.

Lensa AI climbs the App Store charts as its ‘magic avatars’ go viral by Amanda Silberling originally published on TechCrunch

A little over a year after buzzy coding bootcamp Lambda School rebranded as Bloom Institute of Technology, the venture-backed startup is conducting massive layoffs, according to sources. The workforce reduction, per people familiar with the matter, has impacted half of the company’s staff across content, product, data and engineering teams. The layoff is expected to have impacted around 88 employees, using metrics provided in BloomTech’s 2022 diversity report metrics.

This is the company’s third round of known layoffs since the COVID-19 pandemic began. During BloomTech’s last big layoff, in April 2021, BloomTech CEO Austin Allred admitted that it’s been difficult to make his for-profit company’s vision of “incentive-aligned education work.” In a now-deleted tweet posted earlier today, Allred quoted a prior statement of gratitude for employees pushing through the work despite being attacked from all sides, adding “even more so today.”

The name change to BloomTech appeared to be part of the company’s attempt to get things back on track. At the time, the startup updated its tuition payment options to introduce an outcomes-based loan. The financing instrument allows students to take a loan with zero dollars upfront, and then get 110% of their tuition refunded, including fees and interest from an approved lender, if they are unable to secure a job in web development or computer science within the next year.

At the time, the move meant that BloomTech was expanding beyond the original vision of scaling income-sharing agreements (ISAs), the controversial financing vehicle that it helped pioneer. ISAs, which are used between 90% to 100% of students within BloomTech’s cohorts, will continue to be offered as an option, with some changes.

Fast forward to today, and the current state of the tech labor market doesn’t make BloomTech’s mission any easier. The widespread layoffs within tech companies have unlocked a massive market of trained professionals looking for their next gig; all potential competition for BloomTech graduates hoping to stand out against ex-Twitter and ex-Stripe employees. Career Karma, a platform that helps connect students to coding bootcamps, cut staff this year while Flockjay pivoted away from its bootcamp pitch in 2021.

Market uncertainty is certainly one factor that impacts outcomes, but BloomTech has also been embroiled within a number of lawsuits over the past few years led by students who claim to have been misled by the institution. BloomTech’s pitch is that it can help students land tech jobs, but questions have arisen about what happens when students get jobs after graduating from BloomTech’s program that are unrelated to their studies there, and whether they still owe the company a share of their resulting income.

These tensions have grown over the years. Most recently, reports one education publication, a student accused BloomTech of intentionally misrepresenting its job placement rates. Last year, leaked documents obtained by Business Insider raised questions about the company inflating its efficacy and hyping up a curriculum that didn’t upskill folks at the level expected.

Allred did not immediately respond to requests for comment.

BloomTech, previously Lambda School, cuts half of staff by Natasha Mascarenhas originally published on TechCrunch

Despite a joint statement between Ye (fka Kanye West) and Parler in October noting that the two had reached an agreement for the rapper to buy the social network, that will not come to pass, Parler owner Parlement Techonologies said today.

Parler’s owner and Ye have “mutually” parted ways without closing the deal, the company said via emailed statement to TechCrunch. It sounds like the deal was also never really formalized beyond at best a memorandum-of-understanding, given the wording of the official statement, which follows:

Parlement Technologies has confirmed that the company has mutually agreed with Ye to terminate the intent of sale of Parler. This decision was made in the interest of both parties in mid-November. Parler will continue to pursue future opportunities for growth and the evolution of the platform for our vibrant community.

Also worth noting that Parlement says the deal actually died in mid-November, so it’s been sitting on this info for at least half a month now. So what’s changed?

Well, there is the fact that earlier today West praised Adolf Hitler and the Nazis during an interview with none other than Alex Jones — himself a person who was recently found liable for nearly $1 billion in damages for spreading horrible lies about the incredibly tragic Sandy Hook school shooting.

There is that.

Kanye West isn’t buying Parler after all by Darrell Etherington originally published on TechCrunch

While this year’s stock market decline was swift, it was also widespread, with very few companies escaping the downturn. But current market conditions haven’t caused the same uniform trajectory for startups.

When public-market stock prices started to fall, everyone reminded themselves that it would take a few months to see the real impact on the private market — historically a six-month lag. But data from Caplight, a fintech that looks to make secondary trading more transparent, found that late-stage startups weren’t really following a singular trend.

The sample set of startups that Caplight examined includes the 10 highest-valued venture-backed companies, including recognizable names like Canva, ByteDance, and Stripe. We are focusing on the changes to the share prices at which these companies have been traded on the secondary market. Those prices are, in turn, derived from a company’s valuation set during secondary trades.

The data found that some of these late-stage startups’ valuations fell in line with the public market, while some started to drop off in 2021, before the public markets tanked, and others are still seeing their valuation creep up. While we don’t know precisely why, when, and how hard each company’s valuation is getting hit — if at all — there are a few observations worth noting.

Startup valuations are declining — but not consistently by Rebecca Szkutak originally published on TechCrunch

After launching as a pilot late last year, Discord will allow more servers to offer paid memberships in exchange for special server-specific perks.

Plenty of Discord communities were already using third-party services like Patreon to offer access to premium content, but the expanded feature will allow Discord-centric communities to manage that process from within the app itself. With the monetization option built into Discord, anyone who runs a server can create customized membership tiers that offer special roles and other benefits, like early access to merch drops or exclusive content.

Discord now calls the feature “server subscriptions” and it does what’s on the label. To be eligible, server owners must be located in the U.S. (for now, at least) and be in good standing on the platform with no recent violations of its platform policies and community guidelines.

The subscriptions offer server owners quite a bit of flexibility on what their monetization will look like. They can set subscriptions to be anywhere from $2.99 to $199.99 and decide which kind of perks to associate with the tiers they come up with. Discord’s cut of subscription revenue is 10%.

Unlike Twitter, which shoehorned a paid premium option late in the game, Discord users are already used to paying for special features on the app. Nitro, the premium version of Discord, bundles HD streaming, bigger file uploads and custom emoji for monthly subscribers. Discord devotees can also pay to “boost” their favorite servers, unlocking community-wide perks that everyone can use.

Beyond giving server owners ways to monetize their content, Discord is also adding more things for users to do in servers beyond text chat, audio and livestreaming. A few months ago, the company announced a suite of new interactive experiences that run within Discord servers including chess, poker and a way for community members to watch YouTube videos together.

Discord also announced a directory for apps — stuff like RPG mini-games and new sneaker drop alerts that run within servers (most users have historically called them “bots.”) Some hand-picked developers can sell premium subscriptions to apps, giving Discord’s broader community more ways to make money within the app itself.

Discord opens up paid subscriptions so servers can sell premium perks by Taylor Hatmaker originally published on TechCrunch