Connect with us


What to expect from AI in 2023



As a rather commercially successful author once wrote, “the night is dark and full of terrors, the day bright and beautiful and full of hope.” It’s fitting imagery for AI, which like all tech has its upsides and downsides.

Art-generating models like Stable Diffusion, for instance, have led to incredible outpourings of creativity, powering apps and even entirely new business models. On the other hand, its open source nature lets bad actors to use it to create deepfakes at scale — all while artists protest that it’s profiting off of their work.

What’s on deck for AI in 2023? Will regulation rein in the worst of what AI brings, or are the floodgates open? Will powerful, transformative new forms of AI emerge, a la ChatGPT, disrupt industries once thought safe from automation?

Expect more (problematic) art-generating AI apps

With the success of Lensa, the AI-powered selfie app from Prisma Labs that went viral, you can expect a lot of me-too apps along these lines. And expect them to also be capable of being tricked into creating NSFW images, and to disproportionately sexualize and alter the appearance of women.

Maximilian Gahntz, a senior policy researcher at the Mozilla Foundation, said he expected integration of generative AI into consumer tech will amplify the effects of such systems, both the good and the bad.

Stable Diffusion, for example, was fed billions of images from the internet until it “learned” to associate certain words and concepts with certain imagery. Text-generating models have routinely been easily tricked into espousing offensive views or producing misleading content.

Mike Cook, a member of the Knives and Paintbrushes open research group, agrees with Gahntz that generative AI will continue to prove a major — and problematic — force for change. But he thinks that 2023 has to be the year that generative AI “finally puts its money where its mouth is.”

Prompt by TechCrunch, model by Stability AI, generated in the free tool Dream Studio.

“It’s not enough to motivate a community of specialists [to create new tech] — for technology to become a long-term part of our lives, it has to either make someone a lot of money, or have a meaningful impact on the daily lives of the general public,” Cook said. “So I predict we’ll see a serious push to make generative AI actually achieve one of these two things, with mixed success.”

Artists lead the effort to opt out of data sets

DeviantArt released an AI art generator built on Stable Diffusion and fine-tuned on artwork from the DeviantArt community. The art generator was met with loud disapproval from DeviantArt’s longtime denizens, who criticized the platform’s lack of transparency in using their uploaded art to train the system.

The creators of the most popular systems — OpenAI and Stability AI — say that they’ve taken steps to limit the amount of harmful content their systems produce. But judging by many of the generations on social media, it’s clear that there’s work to be done.

“The data sets require active curation to address these problems and should be subjected to significant scrutiny, including from communities that tend to get the short end of the stick,” Gahntz said, comparing the process to ongoing controversies over content moderation in social media.

Stability AI, which is largely funding the development of Stable Diffusion, recently bowed to public pressure, signaling that it would allow artists to opt out of the data set used to train the next-generation Stable Diffusion model. Through the website, rightsholders will be able to request opt-outs before training begins in a few weeks’ time.

OpenAI offers no such opt-out mechanism, instead preferring to partner with organizations like Shutterstock to license portions of their image galleries. But given the legal and sheer publicity headwinds it faces alongside Stability AI, it’s likely only a matter of time before it follows suit.

The courts may ultimately force its hand. In the U.S. Microsoft, GitHub and OpenAI are being sued in a class action lawsuit that accuses them of violating copyright law by letting Copilot, GitHub’s service that intelligently suggests lines of code, regurgitate sections of licensed code without providing credit.

Perhaps anticipating the legal challenge, GitHub recently added settings to prevent public code from showing up in Copilot’s suggestions and plans to introduce a feature that will reference the source of code suggestions. But they’re imperfect measures. In at least one instance, the filter setting caused Copilot to emit large chunks of copyrighted code including all attribution and license text.

Expect to see criticism ramp up in the coming year, particularly as the U.K. mulls over rules that would that would remove the requirement that systems trained through public data be used strictly non-commercially.

Open source and decentralized efforts will continue to grow

2022 saw a handful of AI companies dominate the stage, primarily OpenAI and Stability AI. But the pendulum may swing back towards open source in 2023 as the ability to build new systems moves beyond “resource-rich and powerful AI labs,” as Gahntz put it.

A community approach may lead to more scrutiny of systems as they are being built and deployed, he said: “If models are open and if data sets are open, that’ll enable much more of the critical research that has pointed to a lot of the flaws and harms linked to generative AI and that’s often been far too difficult to conduct.”


Image Credits: Results from OpenFold, an open source AI system that predicts the shapes of proteins, compared to DeepMind’s AlphaFold2.

Examples of such community-focused efforts include large language models from EleutherAI and BigScience, an effort backed by AI startup Hugging Face. Stability AI is funding a number of communities itself, like the music-generation-focused Harmonai and OpenBioML, a loose collection of biotech experiments.

Money and expertise are still required to train and run sophisticated AI models, but decentralized computing may challenge traditional data centers as open source efforts mature.

BigScience took a step toward enabling decentralized development with the recent release of the open source Petals project. Petals lets people contribute their compute power, similar to Folding@home, to run large AI language models that would normally require an high-end GPU or server.

“Modern generative models are computationally expensive to train and run. Some back-of-the-envelope estimates put daily ChatGPT expenditure to around $3 million,” Chandra Bhagavatula, a senior research scientist at the Allen Institute for AI, said via email. “To make this commercially viable and accessible more widely, it will be important to address this.”

Chandra points out, however, that that large labs will continue to have competitive advantages as long as the methods and data remain proprietary. In a recent example, OpenAI released Point-E, a model that can generate 3D objects given a text prompt. But while OpenAI open sourced the model, it didn’t disclose the sources of Point-E’s training data or release that data.

OpenAI Point-E

Point-E generates point clouds.

“I do think the open source efforts and decentralization efforts are absolutely worthwhile and are to the benefit of a larger number of researchers, practitioners and users,” Chandra said. “However, despite being open-sourced, the best models are still inaccessible to a large number of researchers and practitioners due to their resource constraints.”

AI companies buckle down for incoming regulations

Regulation like the EU’s AI Act may change how companies develop and deploy AI systems moving forward. So could more local efforts like New York City’s AI hiring statute, which requires that AI and algorithm-based tech for recruiting, hiring or promotion be audited for bias before being used.

Chandra sees these regulations as necessary especially in light of generative AI’s increasingly apparent technical flaws, like its tendency to spout factually wrong info.

“This makes generative AI difficult to apply for many areas where mistakes can have very high costs — e.g. healthcare. In addition, the ease of generating incorrect information creates challenges surrounding misinformation and disinformation,” she said. “[And yet] AI systems are already making decisions loaded with moral and ethical implications.”

Next year will only bring the threat of regulation, though — expect much more quibbling over rules and court cases before anyone gets fined or charged. But companies may still jockey for position in the most advantageous categories of upcoming laws, like the AI Act’s risk categories.

The rule as currently written divides AI systems into one of four risk categories, each with varying requirements and levels of scrutiny. Systems in the highest risk category, “high-risk” AI (e.g. credit scoring algorithms, robotic surgery apps), have to meet certain legal, ethical and technical standards before they’re allowed to enter the European market. The lowest risk category, “minimal or no risk” AI (e.g. spam filters, AI-enabled video games), imposes only transparency obligations like making users aware that they’re interacting with an AI system.

Os Keyes, a Ph.D. Candidate at the University of Washington, expressed worry that companies will aim for the lowest risk level in order to minimize their own responsibilities and visibility to regulators.

“That concern aside, [the AI Act] really the most positive thing I see on the table,” they said. “I haven’t seen much of anything out of Congress.”

But investments aren’t a sure thing

Gahntz argues that, even if an AI system works well enough for most people but is deeply harmful to some, there’s “still a lot of homework left” before a company should make it widely available. “There’s also a business case for all this. If your model generates a lot of messed up stuff, consumers aren’t going to like it,” he added. “But obviously this is also about fairness.”

It’s unclear whether companies will be persuaded by that argument going into next year, particularly as investors seem eager to put their money beyond any promising generative AI.

In the midst of the Stable Diffusion controversies, Stability AI raised $101 million at an over-$1 billion valuation from prominent backers including Coatue and Lightspeed Venture Partners. OpenAI is said to be valued at $20 billion as it enters advanced talks to raise more funding from Microsoft. (Microsoft previously invested $1 billion in OpenAI in 2019.)

Of course, those could be exceptions to the rule.

Jasper AI

Image Credits: Jasper

Outside of self-driving companies Cruise, Wayve and WeRide and robotics firm MegaRobo, the top-performing AI firms in terms of money raised this year were software-based, according to Crunchbase. Contentsquare, which sells a service that provides AI-driven recommendations for web content, closed a $600 million round in July. Uniphore, which sells software for “conversational analytics” (think call center metrics) and conversational assistants, landed $400 million in February. Meanwhile, Highspot, whose AI-powered platform provides sales reps and marketers with real-time and data-driven recommendations, nabbed $248 million in January.

Investors may well chase safer bets like automating analysis of customer complaints or generating sales leads, even if these aren’t as “sexy” as generative AI. That’s not to suggest there won’t be big attention-grabbing investments, but they’ll be reserved for players with clout.


A network of knockoff apparel stores exposed 330,000 customer credit cards



If you recently made a purchase from an overseas online store selling knockoff clothes and goods, there’s a chance your credit card number and personal information were exposed.

Since January 6, a database containing hundreds of thousands of unencrypted credit card numbers and corresponding cardholders’ information was spilling onto the open web. At the time it was pulled offline on Tuesday, the database had about 330,000 credit card numbers, cardholder names, and full billing addresses — and rising in real-time as customers placed new orders. The data contained all the information that a criminal would need to make fraudulent transactions and purchases using a cardholder’s information.

The credit card numbers belong to customers who made purchases through a network of near-identical online stores claiming to sell designer goods and apparel. But the stores had the same security problem in common: any time a customer made a purchase, their credit card data and billing information was saved in a database, which was left exposed to the internet without a password. Anyone who knew the IP address of the database could access reams of unencrypted financial data.

Anurag Sen, a good-faith security researcher, found the exposed credit card records and asked TechCrunch for help in reporting it to its owner. Sen has a respectable track record of scanning the internet looking for exposed servers and inadvertently published data, and reporting it to companies to get their systems secured.

But in this case, Sen wasn’t the first person to discover the spilling data. According to a ransom note left behind on the exposed database, someone else had found the spilling data and, instead of trying to identify the owner and responsibly reporting the spill, the unnamed person instead claimed to have taken a copy of the entire database’s contents of credit card data and would return it in exchange for a small sum of cryptocurrency.

A review of the data by TechCrunch shows most of the credit card numbers are owned by cardholders in the United States. Several people we contacted confirmed that their exposed credit card data was accurate.

TechCrunch has identified several online stores whose customers’ information was exposed by the leaky database. Many of the stores claim to operate out of Hong Kong. Some of the stores are designed to sound similar to big-name brands, like Sprayground, but whose websites have no discernible contact information, typos and spelling mistakes, and a conspicuous lack of customer reviews. Internet records also show the websites were set up in the past few weeks.

Some of these websites include:


If you bought something from one of those sites in the past few weeks, you might want to consider your banking card compromised and contact your bank or card provider.

It’s not clear who is responsible for this network of knockoff stores. TechCrunch contacted a person via WhatsApp whose Singapore-registered phone number was listed as the point of contact on several of the online stores. It’s not clear if the contact number listed is even involved with the stores, given one of the websites listed its location as a Chick-fil-A restaurant in Houston, Texas.

Internet records showed that the database was operated by a customer of Tencent, whose cloud services were used to host the database. TechCrunch contacted Tencent about its customer’s database leaking credit card information, and the company responded quickly. The customer’s database went offline a short time later.

“When we learned of the incident, we immediately contacted the customer who operates the database and it was shut down immediately. Data privacy and security are top priorities at Tencent. We will continue to work with our customers to ensure they maintain their databases in a safe and secure manner,” said Carrie Fan, global communications director at Tencent.

Read more:

Continue Reading


All Raise CEO steps down again



Less than a year after assuming the role, All Raise CEO Mandela SH Dixon has stepped down from her position at the nonprofit. The entrepreneur, who previously ran Founder Gym, an online training center for underrepresented founders, said in a blog post that the decision was made after she realized “being in the field working directly with entrepreneurs everyday” is her passion. Dixon said that she will be exploring new opportunities in alignment with that.

Her resignation is effective starting February 1st, 2023. She will remain an advisor to the Bay Area-based nonprofit.

This is the second chief executive to leave All Raise since it was first founded in 2017. In 2021, Pam Kostka resigned as the helm of the nonprofit to rejoin the startup world as well; Kostka is now an operator in residence and limited partner at Operator Collective, according to her LinkedIn. With Dixon gone, Paige Hendrix Buckner, who joined the outfit as chief of staff nine months ago, will step in as interim CEO. In the same blog post, Buckner wrote that “Mandela leaves All Raise in a strong position, and I’m grateful for the opportunity to continue the hard work of diversifying the VC backed ecosystem.”

Dixon did not immediately respond to comment on the record. It is unclear if All Raise is immediately kicking off a permanent CEO search.

The nonprofit has historically defined its goals in two ways: first, it wants to increase the amount of seed funding that goes to female founders from 11% to 23% by 2030, and, second, it wants to double the percentage of female decision-makers at U.S. firms by 2028.

In previous interviews, Dixon said that the company will work on creating explicit goals around what impact it wants to have for historically overlooked individuals. The data underscores the challenge ahead. Black and LatinX women receive disproportionately less venture capital money than white women; non-binary founders can also face higher hurdles when seeking funding, as All Raise board member Aileen Lee noted in the blog post.  The nonprofit has created specific programs for Black and Latinx founders but has not disclosed a specific goal for the cohort yet. These disconnects can be lost if not tracked. All Raise’s last impact report was published in 2020 and they’re working on bringing that analysis back, Lee tells TechCrunch in an interview.

“All Raise is in great hands with Paige as interim leader and we’ve got a lot of exciting things that we’re shaping and scaling,” Lee said. “We have to all continue to link arms to try and continue to make improvements for our industry…we’ve made good progress that we can’t let up.”

Since launch, the nonprofit has raised $11 million in funding, and opened regional chapters in New York, Boston, Los Angeles, Chicago, DC and, soon, Miami.

Continue Reading


Shopping app Temu is using TikTok’s strategy to keep its No. 1 spot on App Store



Temu, a shopping app from Chinese e-commerce giant Pinduoduo, is having quite the run as the No. 1 app on the U.S. app stores. The mobile shopping app hit the top spot on the U.S. App Store in September and has continued to hold a highly-ranked position in the months that followed, including as the No. 1 free app on Google Play since December 29, 2022. More recently, Temu again snagged the No. 1 position again on the iOS App Store on January 3 and hasn’t dropped since — even outpacing competitor Shein’s daily installs in the U.S.

Offering cheap factory-to-consumer goods, Temu provides access to a wide range of products, including fast fashion, and pushes users to share the app with friends in exchange for free products, which may account for some of its growth. However, the large majority of its new installs come from Temu’s marketing spend, it seems.

When TechCrunch covered Temu’s rise in November, the app had then seen a little more than 5 million installs in the U.S., according to data from app intelligence firm Sensor Tower, making the U.S. its largest market. Now, the firm says the app has seen 5 million U.S. installs this January alone, up 19% from 4.2 million in the prior 22 days from December 10 through December 31.

According to Sensor Tower estimates, Temu has managed to achieve a total of 19 million lifetime installs across the U.S. App Store and Google Play, more than 18 million of which came from the U.S.

The growth now sees Temu outpacing rival Shein in terms of daily installs. In October, Temu was averaging around 43,000 daily installs in the U.S., the firm said, while Shein averaged about 62,000. In November, Temu’s average daily installs grew to 185,000 while Shein’s climbed to 70,000 and last month, Temu averaged 187,000 installs while Shein saw about 62,000.

The shopping app’s fast rise recalls how the video entertainment platform TikTok grew to become the most downloaded app worldwide in 2021, after years of outsized growth. The video app topped 2 billion lifetime downloads by 2020, including sister app Douyin in China, Sensor Tower said. Combined, the TikTok apps have now reached 4.1 billion installs.

Like Temu, much of TikTok’s early growth was driven by marketing spend. The video app grew its footprint in the U.S. and abroad by heavily leveraging Facebook, Instagram, and Snapchat’s own ad platforms to acquire its customers. TikTok was famously said to have spent $1 billion on ads in 2018, even becoming Snap’s biggest advertiser that year, for instance.

By investing in user acquisition upfront, TikTok was able to gain a following which then improved its ability to personalize its For You feed with recommendations. Over time, this algorithm became very good at recognizing what videos would attract the most interest thanks to this investment, turning TikTok into one of the most addictive apps in terms of time spent. As of 2020, kids and teens began spending more time watching TikTok than they did on YouTube. And earlier this month, Insider Intelligence data indicated all TikTok users in the U.S. were now spending an average of nearly 1 hour per day on the app (55.8 minutes), compared with just 47.5 minutes on YouTube, including YouTube TV.

While Temu is nowhere near TikTok’s sky-high figures, it appears to be leveraging a similar growth strategy. The company is heavily investing in advertising to acquire users, which it uses to personalize the shopping experience. One of Temu’s key features, in fact, is its own sort of For You page that encourages users to browse trending items “Selected for You.” In addition to gamification elements, Temu also puts heavy emphasis on recommending shops and products on its home page, which is informed by its user data.

But the app’s growth doesn’t seem to be driven by social media. While the Temu hashtag (#temu) on TikTok is nearing 250 million views, that’s not really a remarkable number for an app as big as TikTok where something like #dogs has 120.5 billion views. (Or, for a more direct comparison, #shein has 48.3 billion views.) That suggests Temu’s rise isn’t necessarily powered by viral videos among Gen Z users or influencer marketing, but rather more traditional digital advertising.

According to Meta’s ad library, for instance, Temu has run some 8,800 ads across Meta’s various platforms just this month. The ads promote Temu’s sales and its extremely discounted items, like $5 necklaces, $4 shirts, and $13 shoes, among other deals. These ads appear to be working to boost Temu’s installs, allowing the app to maintain its No. 1 slot on the App Store’s “Top Free” charts, which are heavily influenced by the number of downloads and download velocity, among other things.

Of course, having a high number of downloads doesn’t necessarily mean Temu’s app will maintain a high number of monthly active users. Nor does it mean those users won’t churn out of the app after their initial curiosity has been abated. Still, Temu’s download growth saw it ranking as the No. 1 “Breakout” shopping app by downloads in the U.S. for 2022, according to’s year-end “State of Mobile” report. ( calculates “Breakout” apps in terms of year-over-year growth across iOS and Google Play.)

Because Temu’s growth is more recent, the app did not earn a position on the Top 10 apps in 2022 in either the U.S. or globally in terms of downloads, consumer spend, or monthly active users, on this report. Instead, most of those spots still went to social media apps, streamers, and dating apps like Bumble and Tinder. The only retailer to find a spot on these lists was Amazon, which was the No. 7 app worldwide by active users and the No. 8 most downloaded in the U.S.

Temu’s marketing investment may not pay off as well as TikTok’s did, though, as other discount shopping apps saw similar growth only to later fail as consumers found that, actually, $2 shirts and jeans were deals that were too good to be true. Wish famously fumbled as consumers grew frustrated with long delivery times, fake listings, missing orders, poor customer service, and other things consumers expect from online retail in the age of Amazon.

Temu today holds a 4.7-star rating on the U.S. App Store, but those ratings have become less trustworthy over the years due to the ease with which companies can get away with fake reviews. Dig into the reviews further and you’ll find similar complaints to Wish, including scammy listings, damaged and delayed deliveries, incorrect orders and lack of customer service. Without addressing these issues, Temu seems more likely to go the way of Wish, not TikTok, no matter what it spends.

Continue Reading