Tech News

New York City public schools have restricted access to ChatGPT, the AI system that can generate text on a range of subjects and in various styles, on school networks and devices. As widely reported this morning and confirmed to TechCrunch by a New York City Department of Education spokesperson, the restriction was implemented due to concerns about “[the] negative impacts on student learning” and “the safety and accuracy” of the content that ChatGPT produces.

“While the tool may be able to provide quick and easy answers to questions, it does not build critical-thinking and problem-solving skills, which are essential for academic and lifelong success,” the spokesperson told TechCrunch via email, adding that the restricted access came in response to requests from schools.

It’s not a ban per se. The New York City public school system is using the same filter forChatGPT that it uses to block other apps and websites — e.g. YouTube and Facebook — on school property. Individual schools can request to have ChatGPT unblocked, and the spokesperson said that the New York City Department of Education would “welcome” the opportunity to have a conversation with OpenAI, the startup behind ChatGPT, about how the tool could be adapted for education.

As for OpenAI, when reached for comment, a company spokesperson said that OpenAI is developing “mitigations” to help anyone spot text generated by ChatGPT. That’s significant. While TechCrunch reported recently that OpenAI was experimenting with a watermarking technique for AI-generated text, it’s the first time OpenAI has confirmed that it’s working on tools specifically for identifying text that came from ChatGPT.

“We made ChatGPT available as a research preview to learn from real-world use, which we believe is a critical part of developing and deploying capable, safe AI systems. We are constantly incorporating feedback and lessons learned,” the OpenAI spokesperson said. “We’ve always called for transparency around the use of AI-generated text. Our policies require that users be up-front with their audience when using our API and creative tools … We look forward to working with educators on useful solutions, and other ways to help teachers and students benefit from AI.”

ChatGPT has an aptitude for answering questions on topics ranging from poetry to coding, but one of its biggest flaws is its ability to sometimes give answers that sound convincing but aren’t factually true. That ledQ&A coding site Stack Overflow to temporarily ban users from sharing content generated by the AI, saying that ChatGPT made it too easy for users to flood the site with dubious answers. More recently, the International Conference on Machine Learning, one of the world’s largest AI and machine learning conferences, announced a prohibition on papers that include text generated by ChatGPT and other like AI systems for fear of “unanticipated consequences.”

In education, the debate has revolved largely around the cheating potential. Google “ChatGPT to write school papers,” and you’ll find plenty of examples of educators, journalists and students testing the waters by wielding ChatGPT to complete homework assignments and standardized essay tests. Wall Street Journal columnist Joanna Stern used ChatGPT to write a passing AP English essay, while Forbes staffer Emma Whitford tapped it to write two college essays in 20 minutes. Speaking to The Guardian, Arizona State University professor Dan Gillmor gave ChatGPT one of the assignments he typically gives his students and said that he would’ve given the AI’s essay “a good grade.”

Plagiarism is another concern. Like other text-generating AI systems, ChatGPT — which is trained on public data, usually collected without consent — can sometimes regurgitate this information verbatim without citing any sources. That includes factual inaccuracies, as alluded to earlier, as well as biased — including blatantly racist and sexist — perspectives. OpenAI continues to introduce filters and techniques to prevent problematic generations, but new workarounds pop up every day.

Despite those limitations, some educators see potential in generative AI technologies like ChatGPT. In a recent piece for Stanford’s Graduate School of Education website, Victor Lee, associate professor of education at Stanford, noted that ChatGPT may help students “think in ways they currently do not,” for example by helping them discover and clarify their ideas. Teachers may benefit from ChatGPT as well, he posits, by generating many examples for students of a narrative where the basic content remains the same but the style, syntax or grammar differ.

“ChatGPT may [allow] students to read, reflect and revise many times without the anguish or frustration that such processes often invoke, [while] teachers can use the tool as a way of generating many examples and nonexamples of a form or genre,” Lee said in a statement. “Obviously, teachers are less delighted about the computer doing a lot of legwork for students. And students still need to learn to write. But in what way, and what kinds of writing? A … side effect of this new medicine is that it requires all of us to ask those questions and probably make some substantive changes to the overarching goals and methods of our instruction.”

As NYC public schools block ChatGPT, OpenAI says it’s working on ‘mitigations’ to help spot ChatGPT-generated text by Kyle Wiggers originally published on TechCrunch

Update, 1/5/23, 4:30 PM ET: Twitch reports that the site is in the process of recovering, though some users are still experiencing issues.

The popular live streaming service Twitch is experiencing an outage for the second time this week. Around 3 PM ET on Thursday afternoon, many creators who were live on Twitch were cut off in the middle of their streams.

“We are aware of issues with the site and our teams are actively working on a resolution,” Twitch Support tweeted. “Thank you for your reports and patience whilst we work on it — we will keep you updated here.”

Through out the incident, Twitch’s status page indicated that all systems were operational.

Not all streamers were affected, and some were able to restart their streams within about twenty minutes of the outage. But the timing of these issues sparked frustration among fans.

Just two days prior, on January 3, a Twitch outage temporarily made it seem as though some creators had deleted their accounts. This was because Twitch was having trouble loading followed channels on both desktop and mobile.

Twitch confirmed on its status page that the outage “could manifest in chat not working, login difficulties, search impaired, etc.”

At the time of publication, we have seem some affected creators reboot their streams, but Twitch has not indicated via Twitter or its website that Thursday’s issue is fully resolved.

Twitch did not immediately respond for comment about the cause of these issues.

Twitch experiences an outage for the second time in a week by Amanda Silberling originally published on TechCrunch

If you’re a gamer who plays on their Android phone a lot, the Razer Edge might be the dedicated platform you’re looking for — and the device will be available later this month starting at $400, the company announced at CES.

The last few years have been interesting for handheld gaming as, following the smash success of the twitch, companies have adopted various approaches to making something better than a phone but not quite as involved as a full-on console. Valve’s Steam Deck has been the most visible of these attempts, but phone gaming conversion kits like the Backbone have also gained in popularity.

Razer is a force in PC gaming but has had only mixed success in the mobile realm, with its branded phones and accessories not making much of a splash. The Edge is its biggest play yet, an Android handheld dedicated solely to gaming — no calls or texts here. It amounts to a modified version of its Kishi phone-mounted controller with extra features and the 5G “phone” permanently attached.

The device itself is definitely enticing for anyone who plays serious games on Android: with a 6.8″ 2,400×1,080 AMOLED screen running at a 144Hz refresh rate, it actually outstrips what most games on the platform support. But of course software tends to grow to fit the container it is put in, so we can expect games that natively support this display mode as well as the few that already do (like Steam Link, but that’s kind of cheating).

As a compact, dedicated gaming platform it seems promising, but the real question is whether, at $400 for the wi-fi version and more for the 5G one, anyone will shell out when for that price you could get a Steam Deck or Switch. Android has plenty of good games on it, but compared with the library available on PC and Switch it may come off as somewhat lacking. And of course you might already have dropped a few bills on your current gaming phone — which you could a controller for much more cheaply.

If, however, you are banking on the continued success of game streaming services like Geforce Now, perhaps a compact dedicated handheld like this, with a great screen and 5G (on Verizon only at present, though), is the way to go. You’ve got until January 26, when it ships, to make your decision.

We’ll see about getting ourselves a little hands-on time with the device at CES and see if it lives up to the hype.

Razer’s 5G Edge gaming handheld hits on January 26 for $400 by Devin Coldewey originally published on TechCrunch

TikTok has quietly added video scrubbing thumbnails to its platform this week, as the company continues to beef up its video features. The new addition makes it easier for users to find a specific part of a longer video. The feature was first spotted by social media consultant Matt Navarra.

Since the feature is starting to roll out this week, it might not be available for everyone just yet. It appears that the video scrubbing thumbnails are currently only available for newer uploads, as I was unable to see them on older videos. Although users have had the option to rewind or fast forward long videos on TikTok for quite some time now, the addition of video scrubbing thumbnails makes this process more intuitive.

For example, say you’re following a TikTok recipe video and want to get to a specific part of the video, in the past you would have had to blindly scrub through the video. Now, you will be able to see where exactly you want to fast forward to via the new thumbnails.

It was a matter of time before TikTok added this feature, given that it’s slowly been inching into YouTube’s territory. YouTube has had scrubbing thumbnails for what feels like forever, so it makes sense for TikTok to offer them on its platform as well, especially as it continues to present itself as a viable option for longer-form video creators who normally post content on YouTube.

Ever since TikTok rolled out the ability for users to upload videos up to 10 minutes in length, it has been working to improve the viewing experience for users. Last month, TikTok began testing a new horizontal full screen mode with select users globally.

Data has shown that kids and teens now spend more time watching TikTok than YouTube. This has been the case since June 2020, when TikTok began to outrank YouTube in terms of the average minutes per day people ages 4 through 18 spent accessing these two competitive video platforms. By continuing to enhance its viewing experience, TikTok is inching further into YouTube’s domain.

TikTok adds video scrubbing thumbnails to make it easier to find specific parts of videos by Aisha Malik originally published on TechCrunch

For the past several CESes, Qualcomm’s big announcements have largely revolved around automotive news. There will likely still be a fair bit of that this week, as well, but the company just announced an upcoming mobile feature in this down time between the Snapdragon Summit and MWC.

The Southern California chipmaker is partnering with Iridium to bring satellite messaging to select Android devices, starting with those running its new flagship SoC, Snapdragon 8 Gen 2. The technology utilizes the 5G Modem-RF Systems on the Qualcomm hardware, coupled with Iridium’s satellite constellation to deliver emergency messaging, following in the recent footsteps of Apple and T-Mobile.

Garmin is in the mix here, as well, adding support by way of its Response feature built atop the Iridium network for emergency messages in remote locations where no carrier coverage exists. That technology was built with hikers in mind, though certainly there are plenty of other scenarios in which such SOS features could prove a literal life saver.

“Garmin Response supports thousands of SOS incidents each year and has likely saved many lives in the process,” Garmin VP Brad Trenkle says in a release, “and we are looking forward to collaborating with Qualcomm Technologies and Iridium to help people connect to emergency services no matter where life takes them.”

Applications include two-way emergency texting and messaging applications beyond SMS, using Iridium’s L-band. With today’s news, Qualcomm is opening access to the feature up to OEMs. The first batch of smartphones to get the feature will be arriving this year. I’d anticipate some more concrete announcements on that front right around MWC.

Qualcomm partners with Iridium to bring satellite messaging to Android phones by Brian Heater originally published on TechCrunch

To get a roundup of TechCrunch’s biggest and most important crytpo stories delivered to your inbox every Thursday at 12 p.m. PDT,subscribe here.

Welcome back to Chain Reaction and Happy New Year! As we kick off 2023, I’m curious to see if the crypto industry will be living the, “New Year, New Me” motto or if history will repeat itself. Hopefully the former. Anyways, let’s get into it.

The New York Attorney General filed a lawsuit on Thursday against Alex Mashinsky, the co-founder and former CEO of crypto lending platform Celsius, which filed for Chapter 11 bankruptcy in July 2022.

The attorney general claims Mashinsky defrauded hundreds of thousands of investors for billions of dollars worth of cryptocurrency. This announcement follows a bankruptcy judge ruling from Wednesday that Celsius owns users’ interest-bearing crypto accounts thanks to its fine print.

Separately, FTX founder and former CEO Sam Bankman-Fried plead not guilty to all eight counts of U.S. criminal charges on Tuesday.

Prior to the announcement, he was expected to plead not guilty, so this wasn’t a huge shocker to most, but this decision could turn into a lengthy legal battle. Bankman-Fried could face up to 115 years in jail if convicted on all charges. His trial date has been set for October 2, 2023…so stay tuned.

In other big news, Coinbase reached a $100 million settlement after New York financial regulators found the crypto exchange to be violating anti-money laundering laws by failing to conduct adequate background checks.

More details below.

The week in web3

Coinbase reaches $100M settlement over background check failures

As mentioned above, New York financial regulators have found that the popular cryptocurrency exchange Coinbase violated anti-money-laundering laws by failing to conduct adequate background checks. Coinbase will pay a $50 million fine to the New York State Department of Financial Services and is also required to spend $50 million on improving its compliance program.

Bankruptcy judge rules Celsius Network owns users’ interest-bearing crypto accounts

The verdict gives Celsius ownership of the $4.2 billion in cryptocurrency that users deposited into its high-interest Earn program, according to a 45-page filing from the U.S. Bankruptcy Court Southern District of New York on Wednesday. Celsius had approximately 600,000 accounts in its Earn program, and the accounts held a collective value of approximately $4.2 billion as of July 10, 2022, the filing noted. About $23 million of that value consisted of stablecoins. But all of that is now property of the estate, aka Celsius, the judge ruled.

SBF’s anticipated not guilty plea was a ‘smart play’ (TC+)

Former CEO of FTX Sam Bankman-Fried’s not guilty plea to several federal fraud charges was largely anticipated and something a few legal experts suggested was a tactical response. Bankman-Fried pleaded not guilty “because he had the absolute right to do so,” Anthony Sabino, a professor of law at The Peter J. Tobin College of Business at St. John’s University, said to TechCrunch. “And it was the smart play.”

Solana price spikes as newly launched dog coin BONK gains community hype

Last week, Solana (SOL) fell to its lowest level since February 2021. But its price has slightly risen this week and some are crediting it to interest from Solana community members in Bonk, a new meme token that airdropped about 50% of its 56 trillion token supply to users last week. The Shiba Inu–themed cryptocurrency has risen about 43.7% in the past 24 hours, according to CoinGecko data.

Investors say web3 and hype are in for 2023, high valuations are out — maybe? (TC+)

Some think 2023 will just be the start of a venture winter and overall economic recession, while others think we could see some stabilization as things head back to normal by midyear. But who is to say? To find out how investors are thinking about the year ahead and what they’re planning, we asked more than 35 investors to share their thoughts.

The latest pod

Chain Reaction’s first season ended in December, but we’ll be bringing in new content for Season 2 next week, stay tuned!

Subscribe to Chain Reaction on Apple Podcasts, Spotify or your favorite pod platform to keep up with the latest episodes, and please leave us a review if you like what you hear!

Follow the money

MSafe closed a $5 million round to develop multi-signature wallet solutions
Assure Wallet raised $1.4 million for its pre-Series A round
On-chain platform Ondefy raised $1 million to grow DeFi adoption
Web3-focused online education infrastructure DeSchool closed its seed round, hitting a $15 million valuation
Digital wallet Centbee raised $1 million to support Bitcoin remittances

This list was compiled with information from Messari as well as TechCrunch’s own reporting.

Crypto is ringing in the New Year with new lawsuits and new chaos by Jacquelyn Melinek originally published on TechCrunch

We didn’t make it to the in-person CES last year, but we did manage to get our hands on a few products from the comfort of our own homes. That list included Power1, an iPhone battery case with a built slot for AirPods charging. The product’s creator, AXS Technologies, has since done a rethink of the accessory, which debuts at this year’s show.

Like its predecessor, the Model 3 relies on the iPhone’s MagSafe feature, but instead of running the full length of the phone, it’s significantly smaller — though the the company says it adds up to 5x battery life for AirPods. There are two modes – one just charges the headphones, while the other charges both them and the phone at the same time. The pad itself has a magnet strong enough to stick to a phone in a MagSafe case.

“Power1 Model 3 is our most significant and advanced design yet giving users the ability to attach our charging system directly on iPhones or on most MagSafe cases” says CEO John Merenda. “Power1 is the first product to offer all-in-one on-device charging of AirPods and iPhone and creates a new category of charging systems”.

The product is compatible with AirPods Generations 2 and 3 and is set to start shipping in the spring, priced at $100 – the same as its predecessor.

The Power1 AirPods charging case gets a lot smaller by Brian Heater originally published on TechCrunch

Want to get a “text” from your dog when he’s hungry, wants to go outside, or wants to play? The dream of being able to communicate with your dog is coming closer to reality with the launch of FluentPet’s new app-connected talking button system. The new product is an update to the model first made popular by a dog named Bunny, whose 9 million+ social media followers landed the sheepadoodle on Forbes’ list of the top 50 social media creators of 2022.

On TikTok alone, Bunny has 8.2 million followers on the @whataboutbunny account.

At the Consumer Electronics Show in Las Vegas, FluentPet was demonstrating its new communication system for pets, the FluentPet Connect.

Like the first version, the new system involves programmable buttons that, when pressed, speak voice commands. The dog can press the buttons with its paw or nose, while hexagonal, multicolored tiles — which the company called “HexTiles” — hold the buttons in place and offer visual cues for the dogs as to the button’s location. The system can be expanded by snapping together more tiles as the dog’s vocabulary grows.

To use FluentPet, dog owners will record their voice speaking command words, like “water,” “outside,” “ball,” “play,” and others they want to teach their pet to know and understand. Dogs already know words, of course — as any pet owner will tell you — but what makes this system interesting is that the dog can press a button to communicate what it wants. The sound itself emits from a base station with a speaker attached.

Some dogs are smarter than others, and not all dogs will be able to achieve Bunny’s level of success with such a system, of course.

However, the company says that over 70% of dogs using the system will get two buttons within a month — a “play” button and an “outside” button, typically. Then the dog owner may then add on an “All Done” button to signal that a play session has ended. On average, dogs learn up to 9 words with the buttons, the company has found.

Still, the findings from the sales of the original FluentPet system have been eye-opening. Since starting shipping its first-gen product in June 2020, the company has reached over 100,000 households. It surprisingly found that dogs weren’t just able to learn individual words, but were also putting together word combinations to communicate — like “water” plus “bone” to indicate they wanted an ice cube.

“Because we’ve got all of these people that are kind of co-creating this new product category with us, they’re able to communicate to us interesting new things — new button combinations that their dogs seem to use. So this is why we think of it as ‘translation,” explains FluentPet CEO FluentPet CEO Leo Trottier.

Trottier’s background is in cognitive science and A.I., he says — noting he left the Ph.D. program at UC San Diego after doing his undergrad at the University of Toronto in order to build the company that became FluentPet. Originally called CleverPet, the goal had been to build something more in the dog games space. (The other original co-founders have since left).

The data (with users’ permission) is being shared with the Comparative Cognition Lab, and soon, with John Hopkins.

FluentPet’s other co-founder Alexis Devine, is Bunny’s owner — so it turns out, Bunny’s social media across Instagram and TikTok was actually a savvy marketing campaign for FluentPet’s systems.

Work on the updated model began in 2021. This version now introduces Wi-Fi connectivity and a connected mobile app. That means when the sound buttons are pressed by the dog, their message is also sent to the app. If you’re in a different part of the house and can’t hear the sound, the app will alert you.

“There’s app notifications — you can get a ‘text’ message from your dog,” said Trottier. (It’s not really being delivered over SMS, to be clear — it’s from the app.)

The FluentPet Connect system is accepting pre-orders now. The starter pack — aka FluentPet’s “Get Started” Kit — comes with 6 connected buttons, 3 HexTiles and the battery-powered base station with a higher-quality speaker for $159. Original system owners can swap out their old buttons and base station for the new ones, but continue using their own tiles.

The pre-Series A company has raised funding from Republic.com and angel investors; 3 million+ was raised as CleverPet while another 750K came after the pivot.

FluentPet’s talking button system lets you get a “text” from your dog by Sarah Perez originally published on TechCrunch

Come on in, the water is lovely. You wouldn’t have thought there was space for more players in the busy soup that’s the crowded sous-vide market, but Typhur begs to differ, thank you very much. At CES in Las Vegas, the company released an exquisitely designed $99 temperature sensor and announced its flagship product, the Sous Vide Station.

Now, sous-vide (which means ‘under vacuum’) isn’t a new technology. The idea of cooking at lower temperatures for longer, to ensure even heat distribution through the food was first described in 1799. For sous-vide, you vacuum-seal something you want to cook in a bag, and then cook it at a very specific temperature for a specific amount of time. If you have a way of accurately controlling the temperature of the water, and a timer, you’re good to go. A determined hacker with $40 to spare can make one themselves in 10 minutes, so it’s no surprise that Kickstarter is full of sous-vide projects, with various degrees of success, and a variable amount of sophistication.

Typhur skipped Kickstarter, raised $20 million, and is placing itself at the high end of the sophistication scale.

“Food is a science that requires precise timing and temperature to extract the best flavor,” said Frank Sun, Founder of Typhur Inc. “The process of cooking sous vide style has always been a well-kept secret in high-end restaurants. It’s a technique cherished for its ability to deliver a consistent and exact level of doneness in everything from savory steaks to poached eggs. With Typhur’s comprehensive Sous Vide Station, anyone can become an iron chef in their own home kitchen—even if the cooking is in a small space.”

Typhur’s answer to the challenge is admittedly extraordinarily fancy. It packs a 12.3-inch LCD screen, video-based guides and recipes, a powerful 1,750W heater and circulator, and a very cool design indeed. Among other things, the company created reusable vacuum bags and a hand-held air extraction pump that magnetically connects to the side of the machine, where it also wirelessly charges. It wouldn’t look out of place in the most high-design fancy home kitchens. At CES, the team insists that the pricing has not yet been set, but the Typhur website lists it with a $900 price tag, and a pre-sale price of $700. Pre-orders are opening ‘soon’.

The company also showed off the first product it is shipping: the Typhur InstaProbe, a $100 temperature probe to see if your food is done. You can buy less sophisticated versions for $15, but it can’t be denied that the InstaProbe is a fantastically sleekly designed version.

“We are thrilled to release what will be the fastest and most accurate food thermometer on the market” said Sun. “The speed and precision offered by the InstaProbe offers unparalleled control over your meats. Most food thermometers take 2-5 seconds to get a reading and are not completely accurate, The Typhur InstaProbeoffers highly accurate readings in less than a second thanks to our patentedtechnology.”

We have no way of verifying Typhur’s claims, so we’ll have to take their word for them, but it is undoubtedly one of the better-looking food thermometers we’ve seen, and there’s no shortage of people who optimize for that sort of thing, so we can absolutely see these showing up in high-end kitchen goods stores and at Silicon Valley barbecues in the near future. The thermometers are shipping now.

Typhur launches sous-vide cooker with 12-inch display by Haje Jan Kamps originally published on TechCrunch

Venture capitalists are in the business of predicting the next big thing, even if they get burned in the process. While everyone piled onto crypto in 2021 — and many remain bullish about its future despite multiple failures this year — 2022 saw the rise of generative AI.

But as is the case with any transformative new tech, hype is sure to accompany growing adoption, and generative AI has garnered so much attention and money that many VCs already feel the budding sector will be the next bubble.

TechCrunch recently surveyed more than 35 investors working in different geographies, investment stages and sectors about how they were feeling about next year. One of our questions sought their prediction of where the next bubble would be, and almost half of those surveyed mentioned generative AI.

While several investors said they were bullish on the new tech overall, they also admitted that the sector was likely to get lost in its own hype.

Don Butler, managing director at Thomvest Ventures, feels the bubble is already here. “We believe that the applicability of AI to so many use cases will lead to a very large number of startups being funded in the space, including at some eye-watering valuations, and so we think that the next big bubble is already expanding in this area,” he said.

This isn’t surprising, given how the sector has burgeoned recently as consumer tools like ChatGPT, DALL-E and Lensa AI soared in popularity both in and outside the tech community.

Whoops! Is generative AI already becoming a bubble? by Rebecca Szkutak originally published on TechCrunch