OpenAI with SoftBank and Jony Ive to design the next big hardware platform, aiming to offer a more natural and intuitive user experience and mark a new era in the world of artificial intelligence
OpenAI Reportedly Taps SoftBank For $1 Billion ‘iPhone Of Artificial Intelligence’ Venture
OpenAI, the brainiacs behind ChatGPT, is teaming up with Jony Ive, the genius designer from Apple, and the big-shot Japanese tech firm, SoftBank. They’re cooking up something big, a $1 billion project! They’re aiming to create the next big thing after the smartphone—calling it the “iPhone of artificial intelligence.”
Jony’s company, LoveFrom, is all set to bring OpenAI’s first gadget to us regular folks. Jony, the maestro behind the iPhone and iPod, is brainstorming with Sam Altman, the head honcho of OpenAI. They’re dreaming up new designs, aiming for something that feels natural and easy for us to use, to interact with AI.
Masayoshi Son, the CEO of SoftBank, is possibly throwing more than $1 billion into this venture. They’re talking about creating a brand-new company, pulling in talents and tech from all three—OpenAI, LoveFrom, and SoftBank. The word is Son wants Arm, a chip designing company SoftBank mostly owns, to be a key player in this project.
AWS announces the general availability of Amazon Bedrock and powerful new offerings to accelerate generative AI innovation
Amazon’s cloud service, AWS, has rolled out a bunch of new tools to help companies use generative AI – a fancy term for tech that can generate content. They’re saying this AI could change how businesses work in a big way. Here are the juicy bits:
- Amazon Bedrock– It’s a tool that lets companies use massive machine learning models made by top AI companies. So, if a company wanted to use AI for stuff like searching or even making new drugs, Bedrock could be their ticket.
- Amazon Titan Embeddings– This tool takes words and turns them into numbers that can be fed to AIs. If you’ve got info that isn’t in the model’s training, Titan helps the model get the gist and spit out better answers. It can work with more than 25 languages and understands long chats.
- Meta’s Llama 2 – Another AI model that’s coming soon. It’s a beefed-up version of the old Llama and is especially good for conversations.
- Amazon CodeWhisperer – For the tech nerds, it helps with coding by suggesting lines of code. Soon, it’ll even tap into a company’s private code stash to make even better suggestions.
- Amazon QuickSight’s New Feature – For the number crunchers, this tool will take a simple question and turn it into fancy charts. Less manual work, more “Wow, look at that chart!”
- Free Training – AWS isn’t leaving anyone behind. They’ve got a free online course to get folks up to speed on using Amazon Bedrock.
AWS is busting out some new offerings that are all about making AI more accessible, secure, and customized to help businesses innovate and get ahead!
Adobe’s Photoshop on the web launch includes its popular desktop AI tools
Adobe’s Photoshop has hit the web! After nearly two years of testing, you can now use a simpler online version of the famous photo-editing app. This web version packs some popular tools from the desktop app, including Generative Fill and Expand tools. These are pretty cool because they let users make quick changes to pictures using descriptions in over 100 languages, all while keeping the lighting and perspective in check.
These features are powered by Adobe’s smart tech, called Firefly generative AI model. So, whether it’s for personal projects or big commercials, these features are good to go. Photoshop on the web has got a bunch of tools you’d find in the regular desktop version.
Adobe’s also rolling out a new feature, the Contextual Task Bar. It basically suggests your next moves in the editing process. And yes, they’re planning to throw in more features like the patch tool, pen tool, and smart object support soon. Users can invite other folks to work on projects together, even if they don’t have a Photoshop subscription. They can view and comment on the files.
Google is opening up its generative AI search experience to teenagers
Google’s letting teens ages 13-17 in the U.S. get in on its chat-style AI search, called the Search Generative Experience (SGE). It’s a cool way to ask questions more conversationally. Google’s also making sure the tool’s safe, adding safety nets against any nasty stuff, especially since the young folks really dig this feature. They’re even tossing in an “About this result” feature to shed light on how the AI came up with its answers.
But Google’s aware that sometimes the AI might slip up and give an answer that might sound off, so they’re working on making it smarter and safer. Plus, they’re trying out some new things like video and image support, local tips, and even experimenting with ads. So, all in all, it’s like Google Search getting a modern makeover for the chat era. Cool beans!
Zapier launches Canvas, an AI-powered flowchart tool
Zapier’s dropping a new tool called Canvas, letting folks draw out their business plans and processes. It’s got some AI magic to help turn those plans into automations using Zapier. It’s in early access now, so some users can get a sneak peek.
On top of that, Tables, another Zapier service that helps organize data, is now open to everyone.
Wade Foster, the big boss at Zapier, says people are building whole projects, even businesses, using Zapier. But, as these projects get bigger, they can become messy—too many parts and people. That’s where Canvas and Tables come in, helping to sort out the chaos.
Zapier is also rolling out some smaller updates, like a new editor and more control features, along with increasing their integrations to over 6,000.
Microsoft opens AI Co-Innovation Lab in San Francisco to empower Bay Area startups
Microsoft’s kickin’ it up a notch with a new AI Co-Innovation Lab in downtown San Francisco! It’s the fifth one they’ve opened and it’s all about giving startups and big companies a space to work with Microsoft’s AI experts to build and test their AI ideas.
The whole point? Get ideas from being just ideas to actual things you can see and touch, offering the tools and advice to get AI projects off the ground. Microsoft’s been backing AI and cybersecurity startups like Typeface and Hidden Layer through its venture fund.
One startup, Space and Time, hooked up with Microsoft in the lab and jazzed up their SQL Server by adding some natural language features, making complex queries easier to handle.
Meta filmed Mr Beast, Paris Hilton and 26 more to build celebrity AIs based on Llama 2
Hollywood actors are all in a tizzy about how AI’s gonna change the entertainment game and how they get paid when the AI looks just like them. Meta, Facebook’s big daddy, has moved past the drama and just dropped 28 AI versions of famous folks like Mr. Beast, Paris Hilton, Tom Brady, and Snoop Dog. They ain’t just moving pictures; they’re AI animations that can chat with you on Meta’s platforms. You wanna talk sports? Tom Brady’s AI, named Bru, is your guy. Fancy some dance tips? Charli’s AI, Coco, has your back. And Paris? She’s helping you solve mysteries.
These animated pals are built on some fancy tech called Llama 2. They’re made by filming the real celebs and then doing some computer magic to make them interactive. They’re text-based now, but word is they might get voices next year. And if you’re wondering how the real-life celebs get paid for these digital doppelgängers, Meta’s keeping mum.
IBM Tries To Ease Customers’ Qualms About Using Generative AI
IBM is stepping up its A.I. game to ease the worries of businesses interested in using generative A.I., which can crank out things like business reports with humanlike smoothness. A lot of companies are itching to use it but are sweating over legal issues, data secrecy, and whether the A.I. gives the right answers.
To settle their nerves, IBM is saying, “Use our A.I., and if any legal mess pops up, we’ve got your back.” This is a big deal since businesses are scared of getting sued for copyright stuff. IBM is also going to show everyone the data used to teach the A.I., which isn’t a common move. This could help IBM stand out in the crowded A.I. market, where everybody, including OpenAI and Google, is showing off their shiny new A.I. tools.
AI and science: what 1,600 researchers think
A bunch of science folks (1,600+) were asked about using Artificial Intelligence, or AI, in their research. A lot of them think these AI tools are going to be a big deal in science, like “central” big, according to a survey by Nature. More than half reckon these tools will soon become “very important” or downright “essential” for their work.
But scientists are also a bit worried about AI. About 69% of them think it can make us rely too much on recognizing patterns without understanding them, and 58% worry about AI perpetuating bias and discrimination. There’s also concern about more mistakes, fraud, and misinformation due to AI. Some are even saying it’s messing up the way science is done by making it too easy to produce research without thoroughly understanding it.
Another thing, while some scientists are embracing AI, there are a bunch who aren’t using it. Some feel they don’t need it, don’t find it useful, or they just don’t have the time to delve into it. Also, there’s talk about big companies dominating the AI space, and some researchers are raising eyebrows about this, especially when it comes to ownership of AI tools.
Non-engineers guide: Train a LLaMA 2 chatbot
Want your own chatbot without typing a single code line? This guide’s got you covered. By using tools like Spaces, AutoTrain, and ChatUI, you can whip up a chatbot like ChatGPT. Oh, and all these tools come from Hugging Face. The big idea? Making AI tech available to everyone.
Setting up AutoTrain:
- Go to Hugging Face’s “spaces”.
- Name your space. If you wanna share it, choose a license.
- Use Docker to get AutoTrain.
- Choose free CPU to run your app.
- Add your Hugging Face token (found in your profile settings) so your space connects to your account.
- Keep your space private, for now.
- Click “Create Space”. Wait a bit.
Training a Model with AutoTrain:
- Inside AutoTrain, pick the “LLM” tab for language model training.
- Choose the language model. For this example, it’s Llama 2 from Meta.
- Select the CPU or GPU you wanna use.
- Upload your data. It’s gotta be in CSV format.
- If you want, upload some “Validation Data” to test your model.
- There are fancy settings to tweak, but defaults work fine.
- Click “Start Training”. This may take a while, so maybe grab a coffee or two.
Using Your Model in ChatUI:
- Start a new space, but choose ChatUI this time.
- Pick the hardware to run your model.
- Add a database if you have one to store chat logs. If not, don’t sweat it.
- Add your model’s name under “Space variables”. Check your Hugging Face profile for that.
- There’s other settings to mess with, but sticking with defaults is chill.
- Click “Create”. Boom! Your own ChatGPT.
Lost? No worries. Hugging Face has a paid advice service to give you a hand.