H2O.ai, a provider of open-source AI platforms, announced today two new vision-language models designed to improve document analysis and optical character recognition (OCR) tasks.

The models, named H2OVL Mississippi-2B and H2OVL-Mississippi-0.8B, show competitive performance against much larger models from major tech companies, potentially offering a more efficient solution for businesses dealing with document-heavy workflows.

David vs. Goliath: How H2O.ai’s tiny models are outsmarting tech giants

The H2OVL Mississippi-0.8B model, with only 800 million parameters, surpassed all other models, including those with billions more parameters, on the OCRBench Text Recognition task. Meanwhile, the 2-billion parameter H2OVL Mississippi-2B model demonstrated strong general performance across a range of vision-language benchmarks.

“We’ve designed H2OVL Mississippi models to be a high-performance yet cost-effective solution, bringing AI-powered OCR, visual understanding, and Document AI to businesses,” Sri Ambati, CEO and Founder of H2O.ai said in an exclusive interview with VentureBeat. “By combining advanced multimodal AI with efficiency, H2OVL Mississippi delivers precise, scalable Document AI solutions across a range of industries.”

The release of these models marks a significant step in H2O.ai’s strategy to make AI technology more accessible. By making the models freely available on Hugging Face, a popular platform for sharing machine learning models, H2O.ai is allowing developers and businesses to modify and adapt the models for specific document AI needs.

Efficiency meets effectiveness: A new approach to document processing

Ambati highlighted the economic advantages of smaller, specialized models. “Our approach to generative pre-trained transformers stems from our deep investment in Document AI, where we collaborate with customers to extract meaning from enterprise documents,” he said. “These models can run anywhere, on a small footprint, efficiently and sustainably, allowing fine-tuning on domain-specific images and documents at a fraction of the cost.”

The announcement comes as businesses seek more efficient ways to process and extract information from large volumes of documents. Traditional OCR and document analysis methods often struggle with poor-quality scans, challenging handwriting, or heavily modified documents. H2O.ai’s new models aim to address these issues while offering a more resource-efficient alternative to larger language models that may be excessive for specific document-related tasks.

Industry analysts note that H2O.ai’s approach could disrupt the current landscape dominated by tech giants. By focusing on smaller, more specialized models, H2O.ai may be able to capture a significant portion of the enterprise market that values efficiency and cost-effectiveness.

Open source and enterprise-ready: H2O.ai’s strategy for AI adoption

“At H2O.ai, making AI accessible isn’t just an idea. It’s a movement,” Ambati told VentureBeat. “By releasing a series of small foundational models that can be easily fine-tuned to specific tasks, we are expanding the possibilities for creating and using AI.”

H2O.ai has raised $256 million from investors including Commonwealth Bank, Nvidia, Goldman Sachs, and Wells Fargo. The company’s open-source approach and focus on practical, enterprise-ready AI solutions have helped it build a community of over 20,000 organizations and more than half of the Fortune 500 companies as customers.

As businesses continue to grapple with digital transformation and the need to extract value from unstructured data, H2O.ai’s new vision-language models could provide a compelling option for those looking to implement document AI solutions without the computational overhead of larger models. The true test will be in real-world applications, but H2O.ai’s demonstration of competitive performance with much smaller models suggests a promising direction for the future of enterprise AI.

Amy Wu, founder of the AI-based mental health app Manifest, has a bold prediction for the next wave of tech.

“Separately from the AI trend, I think so many people are seeing this loneliness epidemic that’s happening with Gen Z,” she said. “There is no doubt in my mind that there will be unicorns that emerge from those categories to address the loneliness epidemic.”

Manifest isn’t quite a unicorn yet — it’s only in its seed stage, having just raised $3.4 million from a16z Speedrun and a number of other investors. But Wu sees her company as part of a new crop of products trying to mitigate a rise in loneliness.

Wu is in her late twenties, right on the cusp of the murky boundary between millennials and Gen Z, but she understands the struggles of the younger generation. A report from Cigna found that three out of five adults report that they sometimes or always feel lonely; that number is even higher among respondents aged 18-22, at 73%. Manifest is the app she wishes she had when she was an undergraduate at Stanford, navigating a competitive, intimidating environment while living on her own for the first time.

“I really felt like the real world punched me in the face,” Wu told TechCrunch. “I feel like school teaches you all these things around, here’s how to get a job at Facebook, or Google, or Microsoft, or Goldman Sachs, but it doesn’t teach you how to go build your own emotional toolkit.”

When you open the Manifest app, you’ll see a pastel gradient orb in the center of the screen. You can hold the button to talk, or tap it to type, in response to a number of prompts: “What’s on your mind?,” “What are you worried about?,” or “What would be useful for us to talk about?”

Then, the app’s AI will mirror your language and turn it into an affirmation, which you can turn into a personalized audio meditation.

For example, if you tell the app that you’re finding it hard to be proud of yourself after running a 5K because you got last place in your age group (totally not pulling from personal experience…!), it will spit out a couple of affirmations, like, “I strive to appreciate my progress, no matter how small,” or, “I trust that my commitment to this process will lead to growth in both my physical and mental health.”

Maybe those words of AI-generated wisdom help. Maybe they don’t. But Manifest isn’t meant to be an end-all-be-all mental health solution or a replacement for actual mental health treatment. Instead, Manifest is designed to be something that you can use for a few minutes every day to feel just a little bit more grounded.

“We are a wellness app that’s really kind of designed to meet Gen Z where they’re already at,” Wu said. “The real core thesis behind Manifest was like, can we make these bite-sized interactions with wellness super easy and super delightful, where it doesn’t feel like a chore to go do Manifest?”

In a time when young people are overwhelmed by the constant noise of social media, it may seem counterintuitive to use technology — let alone something that can feel as impersonal and amorphous as AI — to address loneliness. But Wu thinks that if Gen Z is already sucked into their phones, then wellness needs to happen there, too.

“Gen Z is hanging out way less in person,” she said. “So it’s like, what do you give a generation that we’ve already done this to? Like, the idea that you tell that person to go outside and hang with their friend is an astronomical leap for them, so how do you go and give them something where they’re already at?”

Manifest launched in stealth this summer, and so far, users have generated 18.7 million “manifestations” in the app.

As with any app of its nature, Manifest has to navigate the ethical challenges around making a consumer mental health product with no medical backing. Wu said that there are safeguards embedded in Manifest’s AI, such as redirecting users to a suicide hotline if they mention self-harm. There are some topics like this that Manifest will decline to engage with.

From a risk standpoint, this could be a smart move for Manifest — it’s dangerous to leverage an experimental AI as a tool to help with something as serious as preventing self-harm. But other startups battling loneliness, like chatbot company Nomi AI, take a different approach. When Nomi AI users open up about thoughts of self-harm, the AI companions won’t halt the conversation — instead, they will try to de-escalate the situation by talking the user through their feelings.

Alex Cardinell, the founder of Nomi AI, argues that just stopping a conversation and providing a suicide hotline number could be alienating to someone who’s struggling for connection.

“I want to make those users feel heard in whatever their dark moment is, because that’s how you get someone to open up, how you get someone to reconsider their way of thinking,” Cardinell told TechCrunch in a recent conversation. “I really want to look at what’s aligned with the user, rather than what’s aligned with the strictest attorney’s loss mitigation strategy.”

Wu doesn’t think that Manifest, or any consumer app, is where people should go if they are in a situation where they need legitimate medical help. But young people are turning to these tools when seeking real medical care isn’t accessible. So, if Wu is right about the impending unicorn startups that will combat the loneliness epidemic, those companies — and Manifest — will need to tread thoughtfully.

TechCrunch has an AI-focused newsletter! Sign up here to get it in your inbox every Wednesday.

Amazon wants Prime Video to be the place you watch coverage of election night. The company announced today that it will be streaming a live election night special on Prime Video hosted by former NBC Nightly News anchor Brian Williams. The show will kick off on November 5th at 5PM ET.

“Brian Williams will be joined live by guests including prominent contributors across news and traditional media, representing a range of backgrounds and perspectives, to share real-time poll results and commentary while also referencing third-party news sources across all political affiliations,” Amazon writes in a blog post. The company adds that it will be an “informative, accessible and non-partisan presentation.” Amazon will announce the guests in “the coming weeks.”

The show and Williams’ involvement were rumored to be in the works last month.

This election night special is just the latest live programming from Amazon on Prime Video. The company has aired live Thursday Night Football NFL games since 2022, and it will be getting a lot of NBA games starting in 2025.

Apple’s celebrating ten years of Apple Pay today, and it’s clear that the Cupertino-based giant is pretty pleased with its payment offerings. I’ll be the first to admit that Apple Pay is still handy and convenient ten years down the road; most of the time, rather than reaching for a physical card, I simply double-tap the power button on my iPhone and pay.

Plenty more card partners have arrived in the years since the original launch, and more retailers have been adopting the standard. Alongside teasing future updates, Apple is rolling out two fresh ways to pay using Apple Pay.

NASA’s Perseverance rover is in the middle of a months-long journey up the rim of Jezero Crater on Mars, and on Thursday it beamed back a status update.

The vehicle started the climb in August in what’s considered to be the most ambitious and arduous phase of Perseverance’s mission since arriving at the red planet in early 2021.

“My journey to the rim of Jezero Crater has been a challenging one,” the rover said in a post on its social media account overseen by NASA’s Jet Propulsion Laboratory, which is conducting the Mars mission. “As you can see in this image from my rear Hazcam, I’m dealing with some steep and slippery terrain. But thanks to my team and autonomous navigation system, I’m avoiding any big hazards as I slowly make my way up.”

The image, below, shows one of the rover’s six wheels and the tracks left by them as Perseverance performs its climb.

My journey to the rim of Jezero Crater has been a challenging one.

As you can see in this image from my rear Hazcam, I’m dealing with some steep and slippery terrain. But thanks to my team and autonomous navigation system, I'm avoiding any big hazards as I slowly make my way up. pic.twitter.com/zlZRqeTPWM

Advertisement

— NASA's Perseverance Mars Rover (@NASAPersevere) October 17, 2024

In an earlier post, Perseverance said it has a few spots to explore on its way to the top, adding that the JPL team was most excited about an area over the crest and outside the crater called Witch Hazel Hill, which includes the most ancient martian crust that the rover is likely to encounter on its journey.

To get there, Perseverance is tackling slopes of up to 23 degrees and which rise around 1,000 feet (305 meters).

Perseverance has spent pretty much all of the last three-and-a-half years exploring the floor of Jezero Crater, as well as the site of an ancient river delta. NASA selected this area for the mission as it was once a huge lake that may have contained microbial life. Perseverance has been collecting cores of rock from this location for later analysis that could reveal if life ever existed on the faraway planet.

Social apps can be very useful for staying in touch and sharing moments with your contacts. However, they can also be dangerous, especially for teenagers, as they facilitate direct contact with bad actors. Meta has announced a series of safety features aimed at teens on Instagram. The changes aim to address cases of sextortion.

Sextortion is a form of blackmail under the threat of sharing intimate photos with your contacts or posting them publicly. The perpetrators can demand large sums of money to not publish such content or even other things. Teenagers are currently the most vulnerable to this type of attack due to the nature of social networks.

Instagram’s new safety features against sextortion on teen accounts

Now, Meta has announced some changes that seek to even prevent “potentially scammy” profiles from contacting more potential victims via Instagram. Firstly, users’ main inboxes will hide or even completely block message requests from such accounts. In addition, accounts belonging to teenagers (up to 16 years old) will not be able to exchange messages with anyone they do not already follow.

Instagram will also display a warning to teens when they receive a message from a suspicious account. Accounts tagged as “potentially scammy” by Instagram will not be able to access the list of followers or people tagged in photos of teen accounts. This will prevent them from being able to access a list of new potential victims.

There are no more specific details on how Instagram determines that an account is “potentially scammy.” However, some factors the system considers are the age of the account, contacts in common with a particular profile, or the country of origin of the account.

Security shields integrated into DMs too

If bad actors manage to bypass those safety features, Instagram still has built-in tools on DMs for teen accounts. The temporary messaging feature will prevent screenshots or screen recordings for photos or videos. These types of images will also not be available on Instagram web. There is also a tool that detects photos with nudity in DMs and automatically blurs them.

Meta has recently been making additional efforts to ensure the safety of teenagers on the platform. The company is in the crosshairs of US authorities for potential harm to young people’s mental health. So, it must try to gain the trust of the public and state representatives.