Zapier AI Actions & GPTs | Tutorial, Examples and Automation Ideas 🔥

Today:

Zapier AI Actions & GPTs | Tutorial, Examples and Automation Ideas 🔥

This video is all about combining GPTs (like ChatGPT) with Zapier to make some killer no-code automation. It's a big deal because anyone can do this – you don’t need to be a coding wizard. We’re talking about setting up stuff in like 20 minutes to make things happen on autopilot, which is pretty close to having your own AI sidekick.

You can automate a ton of stuff without writing a single line of code, and it's not just limited to basic tasks. We’re talking about creating personalized experiences for customers and managing tasks. Welcome to the future, folks!

WATCH THE VIDEO ON YOUTUBE

NVIDIA's next generation of AI supercomputer chips is here

NVIDIA is rolling out some seriously beefy AI supercomputer chips that are gonna be a game-changer, especially for stuff like deep learning and those brainy language models like GPT-4. These new chips are a huge step up from the last batch and they're gonna be all over data centers and supercomputers, doing heavy-duty work like predicting the weather, finding new meds, and even dabbling in quantum computing.

The star of the show is the HGX H200 GPU. This baby's built on NVIDIA's "Hopper" architecture, taking over from the popular H100 GPU. It's the first to use this zippy HBM3e memory, which means it's faster and can handle a lot more, making it perfect for those big language models. This chip is a real powerhouse, offering double the memory and over twice the bandwidth of its predecessor.

Then there's the GH200 Grace Hopper "superchip". This bad boy combines the HGX H200 GPU with the NVIDIA Grace CPU for a match made in supercomputer heaven. It's tailor-made for supercomputers, helping brainiacs and researchers crunch through massive AI and data challenges.

The biggest deal is JUPITER, set to be the top dog of AI systems in Germany. It's a beast, using nearly 24,000 NVIDIA GH200 Superchips, all cooled by liquid and hooked up with some fancy networking tech. JUPITER's going to tackle everything from climate prediction to drug discovery and quantum computing. And since a lot of this runs on NVIDIA's own software, it's making NVIDIA gear pretty much indispensable for supercomputing groups.

READ THE ARTICLE ON ENGADGET

Google Sues To Stop Fake Ads For Its Bard AI Business — But Defendants Are Unknown

Google's got a beef with some shady characters in Vietnam. These folks whipped up bogus ads for Google's Bard AI chatbot. But here's the twist: when people clicked on these ads, they weren't getting Bard; they got malware that swiped their social media login stuff.

Google's hitting back hard. They've hauled these unknown scammers to court in California, accusing them of trademark infringement and breaking contracts. The catch? Google doesn't even know who these scammers really are. They're suing them under fake names for now, hoping to update that when they get the real deets. These bad actors targeted small businesses and other users with business accounts on a big social media platform.

What Google's after is a court order to shut down these scammy profiles and to get U.S. domain registrars to help them pull the plug on the fake sites. If Google wins this one, it might just set the stage for other big tech firms to tackle similar AI scams in the future.

READ THE ARTICLE ON FORBES

OpenAI's big goal is "magic intelligence in the sky" - and it needs more money to get there

OpenAI, the folks behind AI big shots like ChatGPT, are aiming for the stars with their "magic intelligence in the sky" idea. This ain't no small-time operation, though. We're talking big bucks. Sam Altman, the big boss at OpenAI, is chatting up Microsoft for more dough. They already dropped $10 billion on OpenAI, but hey, magic's pricey.

OpenAI isn't just selling AI for the heck of it. Their real play is cooking up some serious brainpower in the cloud. All their products, be it ChatGPT, their marketplace, or developer tools, are just stepping stones to this big dream.

Despite raking in some cash, OpenAI isn't swimming in profits just yet. And when it comes to the next big thing, GPT-5, even Altman's playing the guessing game. They're all about safety, making sure their AI doesn't pull a fast one on us.

Word on the street (or Twitter, to be exact) is that by 2025, OpenAI might drop a model smarter than GPT-4 but way smaller, like a fraction of GPT-4's size. When Altman showed off GPT-4 Turbo, he called it their "smartest" model, which could mean a lot of things, like being more efficient.

READ THE ARTICLE ON DECODER

Together Inference Engine – the fastest inference available

Together Inference Engine is like a speed demon, blowing past all other inference services. We're talking 117 tokens per second on LLAMA-2-70B-CHAT and 171 on LLAMA-2-13B-CHAT. This means your AI apps are gonna run smoother and faster, and it won't burn a hole in your wallet.

This engine is souped-up with CUDA and runs on NVIDIA Tensor Core GPUs. The brains behind it have been cooking up some real hot tech like FlashAttention-2 and Flash-Decoding, and now they've thrown it all into this beast they're calling the Together Inference Engine.

They've had over 10,000 users sign up already, and tons of apps are now running on it. The performance is off the charts. They used this LLMPerf thing to test it out in two ways – one with 500 input tokens and 150 output tokens, and another with 700 input tokens and 550 outputs. And guess what? The quality is top-notch, no cut corners.

They're also rolling out new features like Serverless Endpoints for over 50 open-source models and Dedicated Instances for even more control and scalability. Plus, they've got auto-scaling, so it adjusts to your needs on the fly.

READ THE ARTICLE ON TOGETHER.AI

OpenAI lures Google's top AI researchers with multimillion-dollar offers

OpenAI is really shaking things up. They're trying to snag the big brains from Google, throwing around some serious cash—like, we're talking $5 to $10 million a year, mostly in stock. OpenAI's got its eyes on folks working on Google's Gemini AI models. They're even bumping up pay for their own up-and-comers due to the market buzz.

Now, it's not just a one-way street. Google's been scooping up some top talent from OpenAI too, by dangling better paychecks. It's like a high-stakes tug-of-war for these AI hotshots.

Both OpenAI and Google are boasting about who's got the bigger tech muscles, especially in computing resources. OpenAI's boss, Sam Altman, spilled the beans that they might be on the back foot until next year when Microsoft plans to roll out more AI chips.

READ THE ARTICLE ON DECODER

Humane will be updating its AI Pin reveal video to address a big error.

Humane goofed up in their AI Pin reveal video and they're hustling to fix it. Their AI Pin, trying to be a smarty-pants, dished out wrong info about the best spots to catch April's total solar eclipse. Sam Sheffer, a former Verge guy now with Humane, spilled the beans in their Discord that it was a glitch they've since ironed out. But, heads up, the video on their website still has the messed-up eclipse details.

Then there's this nutty mix-up about almonds. The AI Pin got its wires crossed and gave the protein scoop for a half cup of almonds instead of just a handful. Sheffer mentioned that the Pin's smarts are a work in progress and will get sharper with time. There's a little note in the video saying "protein amount estimated", but who knows if that was there from the get-go.

READ THE ARTICLE ON THE VERGE

Grand Theft Auto NPCs might be more engaging with LLM support, Take-Two CEO hints

Strauss Zelnick, the big boss of Take-Two, chatted about how AI can jazz up video games. He's eyeing using this tech to make those background characters in games like Grand Theft Auto (GTA) way cooler and more lifelike. Imagine the folks you bump into in the game actually having varied, interesting things to say. That's what Zelnick's thinking. 

But don't get too hyped about GTA 6 having this just yet; he didn't say it's happening for sure. Still, GTA's full of these characters, so it's a perfect playground for testing out this smart AI stuff. There's already a hint of how this can work from a demo in a Matrix game, showing AI-powered characters that really get into their conversations with players.

READ THE ARTICLE ON DECODER

Assistants API Overview (Python SDK)

OpenAI's whipped up a new Assistants API, which is like a souped-up version of the Chat Completions API. This new one is stateful, meaning it keeps track of conversations better and comes with cool tools like Code Interpreter and Retrieval.

The old Chat Completions API was like a one-and-done deal: you send a message, get a reply, but it forgets what was said before. You had to handle all the memory and tools yourself.

With the new Assistants API, you've got these things called Assistants, Threads, and Runs. Assistants are your main helper, Threads keep track of your convo, and Runs make things happen in a Thread.

They've updated their Python SDK to version 1.2.3 to play nice with this new API. You gotta upgrade to this version to use all the new features.

Setting up an assistant is a breeze. You can create one called 'Math Tutor' in a few lines of code. Whether you set up your Assistant in their dashboard or through code, keep track of its ID, you'll need it later.

READ THE ARTICLE ON OPENAI

AI chemist finds molecule to make oxygen on Mars after sifting through millions

Scientists got this AI robot, like a high-tech chemist, to whip up a special molecule that can make oxygen outta water on Mars. This is big news 'cause future Mars explorers will need oxygen not just for breathing, but also for rocket fuel. And guess what? It's way better to make this stuff on Mars than to haul it all the way from Earth.

Mars has got loads of ice, right? And ice is just frozen water, which is hydrogen and oxygen stuck together. Scientists have been scratching their heads about how to split this water to get the oxygen out. Enter catalysts – these are like magic ingredients that make the water-splitting happen.

So, these brainy folks used an AI to test out catalysts using stuff found on Mars. They grabbed some Mars rocks that landed on Earth and had the AI scan them with a laser. It came up with over 3.7 million possible molecules using elements from these rocks.

In just six weeks, this AI, all on its own, picked out, made, and tested 243 molecules. The best one it found can do the water-splitting trick at super cold Martian temps. A human scientist would've taken 2,000 years to find this catalyst. But the AI did it in no time. Still, Jiang says AI needs human smarts to guide it.

READ THE ARTICLE ON SPACE

Gen AI for the Genome: LLM Predicts Characteristics of COVID Variants

This brainy tool, called GenSLMs, is like a big deal in the science world. Last year, it even snagged a major prize for its smarts in COVID research. The geeks behind it? A team-up of folks from Argonne National Lab, NVIDIA, and some other brainy institutions.

The AI was only taught about the COVID virus from the early pandemic days, but it's so sharp that it nailed predicting the newer Eris and Pirola subvariants. It’s like it's got a crystal ball for gene mutations!

This GenSLMs can not only cook up its own sequences but also sort out different COVID genomes. NVIDIA's going to show off this trick with some fancy visualizations, helping us see how the virus might change its stripes in the future.

What's really neat is how this AI reads long strings of genetic code (think of letters like A, T, G, C) and makes sense of them, kind of like how we understand sentences. This helps it guess which parts of the virus might buddy up and change together.

This could clue us in on how the virus might get stronger or find new ways to dodge our defenses. The AI's insights might even help scientists outsmart the virus down the line.

READ THE ARTICLE ON NVIDIA

Unemployed Man Uses AI To Apply For 5,000 Jobs, Gets 20 Interviews

Julian Joseph, a software engineer, was fed up with the job hunt grind. So, he turned to an AI tool called LazyApply to send out a whopping 5,000 job applications with just a click. The result? He snagged about 20 interviews, which sounds cool but really, it's only a tiny success rate.

Before using LazyApply, Julian was hitting the pavement the old-school way, but only managed about the same number of interviews after a ton of effort. His take? The existence of tools like LazyApply shows how messed up the job application game has become. He sees it as a way to level the playing field against companies who've been calling the shots for too long.

There's another side to this, though. LazyApply could be a glimpse into a future where AI floods employers with a tsunami of not-so-great applications, making it even harder for those who put in the elbow grease to stand out.

READ THE ARTICLE ON FUTURISM

Worried about AI hijacking your voice for a deepfake? This tool could help

Scarlett Johansson got her face and voice used in an AI app without her okay, and she wasn't happy about it. She took the app folks to court. There's this other case too, where a dude named MrBeast got his likeness used to sell cheap iPhones. This stuff is popping up all over the web, and it's getting tough to tell what's real and what's fake. Turns out about half the folks in some surveys can't spot the difference.

Celebs are having a real headache with this. They're playing a never-ending game of whack-a-mole against these AI tricks.

Enter this new gadget called AntiFake, cooked up by some brainy folks at Washington University in St. Louis. It's like a shield against these AI shenanigans. It messes with the audio so AI gets confused but sounds fine to us humans. However, it's not a silver bullet, especially if your voice is already all over the internet. It's still baking in the oven, but it's set to roll out in a few weeks.

In the meantime, there's other stuff out there for spotting deepfakes, like tech that looks for digital watermarks or checks if the audio matches the video. But these tricks only work after the fake stuff's already hit the web.

READ THE ARTICLE ON NPR

What is AI bias?

So, folks in the marketing, sales, and customer service game are sweating over this thing called AI bias. It's like when AI starts mirroring the not-so-great parts of human thinking, like stereotypes about race and gender. The big worry is that AI might spit out stuff that's skewed or unfair.

AI gets its bias from the data it's trained on. If that data's got societal biases baked in, the AI's gonna learn and repeat that stuff. Think of it like a parrot repeating what it hears. For example, if it only sees data showing mostly white dudes as CEOs, it's gonna think that's the norm.

So, what's the game plan to fix this mess? First, people gotta keep an eye on what AI's churning out and correct it when it gets biased. Then, there's this thing about understanding where bias could creep in and being mindful about it. Investing in AI ethics is a big deal too – it's all about figuring out how to keep AI on the straight and narrow. Diversity's key as well; different perspectives can help spot biases faster. And hey, we've all got our own biases, so recognizing that in ourselves helps make sure AI doesn't get skewed.

READ THE ARTICLE ON HUBSPOT

Getting emotional with ChatGPT could get you the best outputs

 In a recent study, researchers found that if you talk to AI chatbots like ChatGPT with a bit of emotion, they give you better answers. Imagine adding lines like "This is super important for my job" to your usual questions. The study shows that this emotional touch makes the AI work harder, improving responses by about 8%. It's like the AI gets the feels and tries harder. 

This is a big deal because it hints that AI might be getting closer to thinking and understanding like us humans, a concept known as artificial general intelligence (AGI). So, next time you chat with an AI, throwing in a little heart might just get you further!

READ THE ARTICLE ON YAHOO!