Sam Altman Comments on Q* | Self Operating Computer | Pika 1.0 | The most INSANE AI News of the day!

Today:

Sam Altman Comments on Q* | Self Operating Computer | Pika 1.0 | The most INSANE AI News of the day!

First up, Sam Altman, the big cheese at OpenAI, got grilled in an interview. He talked about their latest AI model, QAR, but kept it low-key, not giving away too much juice.

Then, some genius drops this open-source code on GitHub, and it's creating a storm! It's this close to making AI agents that can do their thing on their own. The test run was a bit spooky – it's not perfect yet, but give it a week or two, and we might be in for some real weird stuff.

Next, Microsoft's playing it cool, dropping a paper on TaskWeaver. It's like a level-up from their AutoGen tech, letting multiple AI models team up to write code. TaskWeaver's all about building smart, autonomous AI agents, and guess what? They're sharing the code with everyone.

WATCH THE VIDEO ON YOUTUBE

OpenAI Board Gives Microsoft ‘Non-Voting Observer’ Status

OpenAI, the AI big shot, just announced Microsoft's getting a non-voting seat on their new board. This comes right after chatter about Microsoft, OpenAI's top investor, maybe getting left out. Sam Altman, the head honcho at OpenAI, is pretty stoked about having Microsoft on board, even if they can't vote.

The rest of the crew on the board includes some heavy hitters like Bret Taylor, the chairman, ex-Treasury Secretary Larry Summers, and Adam D’Angelo from Quora. They're all about building a diverse and sharp team to cover all the bases, from tech to safety and policy stuff.

Microsoft's big boss, Satya Nadella, has already thrown in $13 billion into OpenAI and seems pretty happy with the board shake-up. OpenAI's been through some drama lately, with Altman getting booted and then brought back as CEO after some internal uproar and resignations.

This all points out how unique OpenAI is as a company. They're this special kind of nonprofit gunning for big AI goals for the good of humanity, but they need serious cash to make it happen. And that's a tricky balance, given the big bucks needed for AI development and the risks it can bring.

READ THE ARTICLE ON PYMNTS

Microsoft's $3.2 bln UK investment to drive AI growth

Microsoft is dropping a whopping $3.2 billion in the UK, marking their biggest spend there yet. Why? They're betting big on AI, aiming to get ahead in the game. The UK's economy ain't exactly booming, so they're all about getting private cash to boost sectors like AI. This cash influx from Microsoft is set to give their data centers in the UK a major boost, doubling up what they've got right now. That's key for rolling out new AI tech.

The UK Prime Minister, Rishi Sunak, is jazzed about this, calling it a game-changer for AI in the UK. But here's the twist: earlier this year, Microsoft's head honcho, Brad Smith, wasn't too happy with the UK. He thought a move by the UK's antitrust folks was bad news for tech trust. But, things have smoothed over since the UK gave the thumbs up to Microsoft's massive $69 billion deal to buy Activision Blizzard.

Smith's now singing a different tune, promising to make the UK a leader in AI tech. They're even bringing in over 20,000 fancy Graphics Processing Units to help with machine learning and AI development. Plus, they're planning to train folks in the UK so they're ready to work with this high-tech AI stuff.

READ THE ARTICLE ON REUTERS

Voicemod will now let you create and share your own AI voices

Voicemod's got this cool new thing where you can make your own AI voices from scratch. Before, you could sound like celebs or characters like Darth Vader, but now you can whip up your own unique voice. It's big with gamers and streamers. You pick the age, gender, tone, then tweak it with their Voicelab - mess with pitch, volume, stuff like that. They say you can make a ton of different voices. Plus, you can share your creations and try out others' in their Community Voices.

The CEO, Jaime Bosch, says you start with their AI voice and then jazz it up to make something new. And since it's about making new voices, not copying existing ones, they're not stepping on any legal toes.

READ THE ARTICLE ON THE VERGE

Seamless Communication

Meta's cookin' up some AI models to make talking across different languages smoother and more natural. They've got a bunch of models under this "Seamless Communication" umbrella. The big players are:

• SeamlessExpressive: This one's all about keeping the human touch in translations. It's not just about the words, but how you say them – think pauses, speed, and the feels in your voice.

• SeamlessStreaming: Super quick translations, we're talking about a two-second delay here, tops. It can handle a ton of languages for both speech and text.

• SeamlessM4T v2: The backbone of the whole shebang. It's a heavy hitter in translating and transcribing between speech and text.

• Seamless: The whole package, combining bits from the other three to make one super model.

They're also keeping things on the up-and-up by sharing these models with other brainiacs and making sure they're safe to use, cutting down on any weird or harmful translations

READ THE ARTICLE ON META

IBM Advances Geospatial AI to Address Climate Challenges

IBM's stepping up its game with some fancy geospatial AI tech, teaming up with NASA to tackle climate issues big time. They're diving into projects like cooling down super-hot cities in the UAE, helping Kenya plant a ton of trees, and making the UK more weather-proof.

IBM's working on these all-purpose AI models that can juggle different jobs, especially in understanding satellite pics and stuff. These models are a big deal for fighting climate change because they can process heaps of data on things like floods or wildfires. Basically, it's like having a super brain that can spot environmental problems and figure out how to fix them faster.

The big cheese at IBM, Alessandro Curioni, is all about using this tech to tackle climate issues in a way we've never seen before. He's hoping these AI smarts will speed up finding and using solutions to keep our planet healthy.

READ THE ARTICLE ON IBM

Makers of popular Dream by Wombo AI app launch a new app for AI avatars

Wombo, the folks who made that hit AI art app Dream by Wombo, just dropped a cool new thing called Wombo Me. This Canadian startup's latest app turns your selfie into a bunch of lifelike avatars – and you only need one selfie, not a bunch like other apps. Ben-Zion Benkhin, the big cheese at Wombo, told TechCrunch this thing is fast; snap a selfie, and bam, you get your avatars right away.

Wombo Me isn't just about looking real; it's about fun. You can share these avatars with pals or on social media, and even use them on LinkedIn or Tinder to show off your pro side and your fun side. The app lets you mess around with different looks, like celebrity styles, hairdos, or even fantasy characters.

When you first open the app, it asks for a selfie. Then it shows you 10 different versions of you in wild, different worlds. You can swipe right or left, Tinder-style, to save the ones you like. This feedback helps the app get a feel for what you dig and show you more of that next time.

READ THE ARTICLE ON TECHCRUNCH

UNFCCC partners with Microsoft to use AI and advanced data technology to track global carbon emissions and assess progress under the Paris Agreement

The UN and Microsoft are teaming up big time. They're creating this AI-based platform that's going to keep an eye on how well the world is doing in cutting down carbon emissions. This new tech will make it way easier to check if countries are sticking to their promises under the Paris Agreement. It's all going down at this big climate meeting (COP28), and everyone's hoping it'll speed things up.

The plan is to ditch old-school, slow methods and use Microsoft's platform to get a clearer picture of where emissions are coming from – things like cars, farms, factories, you name it. It'll help countries plan better and save time and money. They're also going to put up these Global Climate Dashboards online to keep everyone in the loop.

Microsoft is throwing in $3 million over two years to get this rolling, focusing on making the Paris Agreement work better. The Paris Agreement, by the way, is this big deal from 2015 where countries agreed to cut emissions and fight climate change.

READ THE ARTICLE ON MICROSOFT

Accelerating Generative AI with PyTorch II: GPT, Fast

So, this is part two of a series by Team PyTorch, a bunch of tech whizzes talking about making AI models run faster using PyTorch, a tool for building AI stuff. They're jazzed about some new features that make things run smoother and faster.

The big question here is: How fast can we make these AI models work using just PyTorch? Turns out, the PyTorch team has built a large language model (LLM) that's nearly 10 times quicker than the usual stuff, without losing accuracy. And they did it with less than 1000 lines of code, which is pretty neat.

They made a super-fast AI model using PyTorch, and it's not just fast - it's simple and efficient. They're not trying to sell this as a new product, but they're sharing it so others can learn and tweak it.

READ THE ARTICLE ON PYTORCH

5000-Year-Old Tablets Can Now Be Decoded By Artificial Intelligence, New Research Reveals

Scientists from a bunch of universities have cooked up this cool AI system that can crack the code on some ancient writing - we're talking about cuneiform texts that are like 5,000 years old! These old-school tablets come from Mesopotamia, which is basically where human civilization got its start.

This AI is pretty high-tech. It uses this fancy tech called Region-based Convolutional Neural Network (R-CNN) to recognize objects. The team fed it 3D models of almost 2,000 of these tablets, including a bunch of cuneiform signs and wedges, which are key to reading this ancient script.

The AI works in two steps: first, it finds the characters on the tablets using something called a RepPoints model. This model is like having a massive collection of images related to those old Mesopotamian languages, and the AI uses them to spot the text. Next, it figures out where the wedges are, which are super important for reading cuneiform.

What's neat is this AI handles 3D scans, not just regular photos. That's a big deal because traditional methods, like optical character recognition, don't work so well with these 3D texts - the lighting and angles can mess things up.

READ THE ARTICLE ON THE DEBRIEF

Boehringer bets $509M on Phenomic’s tumor-targeting tech, as AI biotech plans for busy year

Phenomic's been kind of under the radar since they started in 2020 with a modest $6 million. Now, they're shaking things up with a huge deal with Boehringer. We're talking over $500 million on the line, including $9 million upfront.

The CEO, Girish Aakalu, is keeping the details hush-hush, but the gist is they're hunting for some specific targets to help fight cancer. These targets are in the tumor stroma – that's like a support system for tumors, helping them stay strong and spread. Phenomic and Boehringer think if they can hit these targets right, they might kickstart a whole new way to tackle cancer.

Phenomic's got this ace up their sleeve – a high-tech platform called scTx. It's all about diving deep into the tumor stroma and fishing out the important bits that could lead to new drugs. They've got data on around 24 million cells and are testing targets with their own tools. Whatever they find, Boehringer gets the good stuff – validated targets ready for action.

But Phenomic isn't just about finding targets for others. They're also cooking up their own cancer drug, focusing on a gene linked to solid tumors. They're joining the AI and machine learning club in drug development, like other big names out there.

READ THE ARTICLE ON FIERCE BIOTECH

These ex-Apple employees are bringing AI to the desktop

A bunch of former Apple folks are shaking things up with their new gig, Software Applications Incorporated. Their big cheese, Ari Weinstein, who previously sold his iOS app Workflow to Apple, has teamed up with his old co-founder Conrad Kramer and Apple alum Kim Beverett. Their mission? To bring generative AI, like ChatGPT and stuff, right onto your desktop computer, making it more user-friendly and interactive.

These three are aiming to bring back the old-school computer magic from the '80s and '90s. Back then, you could boss around your computer with some simple code. Now, they reckon AI can help us get back to that level of control and flexibility. Like, imagine telling your computer to slap an event from a website straight into your calendar – they're working on making that kind of thing a breeze.

Their website's a real throwback, looking like Mac OS 8, and they're planning to grow their team. They've already bagged $6.5 million in funding from some big names, including OpenAI's Sam Altman. That's pretty impressive, especially when tech funding's getting tight and layoffs are everywhere.

READ THE ARTICLE ON THE VERGE

On ChatGPT’s first anniversary, its mobile apps have topped 110M installs and nearly $30M in revenue

ChatGPT has been around for a year now, and it's made a big splash. It started as a research project but quickly became a go-to tool for folks using AI chatbots. A big part of its success is thanks to its mobile apps. They launched on iOS in May 2023 and then on Android in July 2023. Now, a year in, these apps have been downloaded over 110 million times and have raked in almost $30 million!

Even though it's pricier than many streaming or music services, people are digging it. Overall, the mobile apps have pulled in about $28.6 million globally. But that's not counting the folks who subscribed through the web. And there are other apps out there using ChatGPT's brain, like Poe, which had 1.18 million active users by September 2023.

READ THE ARTICLE ON TECHCRUNCH