Stability AI’s advanced model promises a leap forward in AI image generation, making Spielberg-esque hyper-realistic creations accessible to all.
Stability AI launches SDXL 0.9: A Leap Forward in AI Image Generation
Stability AI’s newest gizmo, SDXL 0.9, a nifty AI contraption that can conjure up snazzy images like a magician pulls rabbits out of a hat. This is the latest in the Stable Diffusion text-to-image family and boy, it’s a lot better than the beta we got in April.
Now you can take it for a spin on ClipDrop, and they’re promising API access soon. Research types can poke at the innards starting mid-July when they aim to hit 1.0. Despite running on your everyday, run-of-the-mill GPU, SDXL 0.9 opens up a whole new world of AI image creation that would make even Spielberg raise an eyebrow. Think hyper-realistic creations for movies, music, and more.
SDXL 0.9 works with two CLIP models, including a big’un named OpenCLIP ViT-G/14. This beefs up the ability to create detailed, high-res images.On the tech side, you just need a modern GPU, Windows 10 or 11, or Linux, and some specific graphics cards. No rocket science degree required.
ElevenLabs Launches New Generative Voice AI Products
AI startup ElevenLabs, which went from the drawing board to over a million registered users faster than a New York minute, is making a splash in the tech world. After their January 2023 beta launch, they’ve been creating audio content with AI so convincing, you’d swear it’s a real person talking to you.
They raised $19M in a Series A round, with a team of tech big shots like Nat Friedman, Daniel Gross, and Andreessen Horowitz leading the pack. Not too shabby for a company that didn’t exist until 2022.
ElevenLabs’ AI voice can spin an entire audiobook quicker than you can say “I forgot my library card.” It’s already got the nod from a broad spectrum of industries, like publishers, game developers, and even a radio channel. Heck, even the visually impaired are on board with this handy way to access written content.
The newly minted $19M will be used to push their research on Voice AI further and branch out into new sectors. ElevenLabs’ CEO, Mati Staniszewski, said that this journey has only just begun, but with the bigwigs they’ve got on their side, they’re off to a flying start.
The company also introduced ‘Projects’, a new feature that gives users a deeper level of control over their AI-generated audio content. Think Google Docs, but for audio creation. On top of that, they’re in the process of testing an AI dubbing tool that’s slated for release later this year.
In their quest for transparency, ElevenLabs is also releasing an ‘AI Speech Classifier’. This handy-dandy tool lets users upload audio and identify if it’s AI-generated from their platform. This is their way of ensuring a safe generative media landscape.
LinkedIn Reveals AI Image Detection Research That Catches Fake Profiles
LinkedIn’s been playing Sherlock with some new AI tech, managing to spot fake profile photos like a hawk on a mouse. Apparently, it’s got a whopping 99% success rate with only a 1% chance of crying wolf. Now, you might be asking, “Why do folks even bother with fake profiles?” Well, some folks reckon it adds a bit of shine to their websites, making them seem more trustworthy.
Google, our all-knowing overlord, has been nudging folks to display E-E-A-T on their content. Some believe a solid LinkedIn profile helps with that. Well, LinkedIn’s new eye-in-the-sky isn’t taking kindly to these shenanigans.
And, let me tell ya, fake profiles aren’t a drop in the bucket. In the first half of 2022, LinkedIn had to give the boot to 21 million of ’em. With the rise of AI and its picture-generating smarts, fakes are becoming trickier to spot.
LinkedIn, however, is one step ahead. Their new detector looks for “artifacts,” which are dead giveaways of a fake AI image. They’ve figured out that these AI images share similar patterns, especially around the eyes and nose – unlike real photos, which are as varied as a box of chocolates.
US-based generative AI job postings up 20% in May, Indeed data show
Job postings for roles related to generative AI saw a rise faster than a runaway jackrabbit in the US last May, spiking around 20%, according to job search giant, Indeed.
To put it in plain numbers, 204 in a million job listings were hankering for AI whizzes, a figure that’s more than double the count in 2021. And we reckon we’ve got OpenAI’s ChatGPT to thank for sparking this shindig.
Now, who’s bagging these gigs? Well, data scientist roles got a piece of the AI job pie, with a slice of 5%, but software engineers, machine learning engineers, and data engineers were also jostling for elbow room at the table.
Nick Bunker, who knows his beans as the director of economic research at Indeed, reckons there’s been a real buzz around AI jobs, thanks in no small part to the splash made by ChatGPT.
Funny enough, this AI job surge is playing out in a broader tech job market, with big guns like Meta Platforms and Amazon tightening their belts amidst uncertain times. Overall, tech jobs are down by a pretty sobering 43.6% from last year.
It’s a dog-eat-dog world out there, with AI job availability struggling to keep up with the swarm of eager beavers on the hunt for them. In fact, searches for generative AI jobs zoomed up from virtually zilch to 147 per million total jobs searched in May.
And who’s hiring, you ask? The line-up reads like the who’s who of tech: Meta Platforms, Apple, Tiktok, Pinterest, and Amazon. If that’s not a golden ticket, I don’t know what is.
Microsoft wants your company to feed its private data into ChatGPT
Microsoft has a shiny new tool for companies where they can spill their company secrets into Azure’s OpenAI Service. Why, you ask? Well, it’s all to make its chatbot, ChatGPT, smarter and more helpful for them. Now this chatbot is kinda like Microsoft’s golden child, with the big M throwing buckets of money into its development.
Now, some people are getting their knickers in a twist over privacy issues and the sheer amount of data this chatbot swallows. But Microsoft, like a dog with a bone, has gone in the opposite direction. Their bigwig, Andy Beatman, is hollering that this was something their customers were begging for.
So here’s how it works: Azure, being the smart aleck it is, fetches the data it needs from the company to help answer the worker’s query. There’s no need for the company to roll up its sleeves and fine-tune its own AI models. Instead, it just hands over its data, sits back, and voilà! The AI dishes up tailored answers, not just run-of-the-mill responses.
Microsoft isn’t the first to pull this trick. Nvidia already started letting organizations feed their own data into their Large Language Models. The result? A chatbot that’s hip to the company’s specific lingo and can provide more useful answers.
The new Azure OpenAI Service is said to grease the wheels for tasks like software development, indexing, and HR stuff. It can pull insights from data anywhere, even off premises, and works seamlessly thanks to Microsoft’s APIs and SDK. And as long as you’re okay with your firm’s data being kept by default, you can start chitchatting with the OpenAI models and make smarter business decisions.