From leveraging OpenAI’s GPT models to the ease of Azure Cognitive Services, witness how RoyBot is redefining organizational communication and role clarity


Introducing RoyBot: Roy Hill’s innovative AI tool boosting employee productivity at scale

Roy Hill’s AI tool, called RoyBot, helps employees make better decisions by providing quick and clear answers. Think of it like a super-smart search engine for company stuff. RoyBot can pull info from company docs, their own version of Wikipedia, task lists, and more. It can even chat like those public messaging services you use.

What’s cool is that RoyBot can understand what you’re asking in plain English, help with job descriptions, and even find folks within the company. It’s powered by some tech from OpenAI (that’s on Microsoft’s Azure platform) to give and improve answers. It also uses some smart search features from Azure to help app developers sprinkle in a bit of AI magic.


Stability AI Unveils Japanese StableLM Alpha: A Leap Forward in Japanese Language Model

Stability AI has rolled out a new Japanese language tool called Japanese StableLM Alpha, claiming it’s the best out there for Japanese speakers. How? They put it head-to-head against four others and it came out on top. 

This super tool was made using a truckload of Japanese and English online text and had a big team, including some hotshots from EleutherAI’s Japanese squad, working behind it. They even got a special software called GPT-NeoX in on the action. Stability AI’s got another cool tool for researchers, too, the Japanese StableLM Instruct Alpha 7B. It’s unique because it’s been trained to follow instructions super closely. When put to the test, it scored a solid 54.71% average.


There’s an AI-powered app that lets users ‘text’ with Jesus. 

Got a new app in town that lets you text Jesus and other biblical bigshots right from your phone! It’s called Text With Jesus. Wanna chat with Jesus, Mary, or Joseph? That’s free. But if you’re itching to unlock Satan or some of the Old Testament crew, it’ll cost ya $2.99 a month. The app’s run on ChatGPT, so you can have some deep, spiritual heart-to-hearts with these AI biblical figures.

But hey, don’t expect them to get all political or controversial. When Insider tried pressing the AI versions of Jesus and Satan on touchy topics, they played it real cool, sticking to what the good book says without getting too judgy. AI Mary, on the other hand, was pretty upfront about her take on abortion, emphasizing the “gift of life.”


US Launches AI Competition to Protect Computer Systems

The U.S. is kick-starting a competition to cook up some AI that can spot and patch up security gaps in government systems. Anne Neuberger, a bigwig in U.S. cybersecurity, mentioned that the bad guys are using AI to snoop for weak spots and make harmful software. We’ve seen quite a few U.S. groups, like hospitals and factories, getting cyber-attacked recently, and not just by random hackers but possibly by folks overseas too.

Canada’s head of cybersecurity, Sami Khoury, piped in saying he’s noticed AI used for sketchy emails, harmful coding, and even to push fake news. The competition is a two-year gig with a cool $20 million up for grabs, run by the defense tech crew, DARPA. Big techies like Google, Microsoft, and OpenAI are in, letting folks use their systems for this contest.


A Practical Deep Learning-Based Acoustic Side Channel Attack on Keyboards

With the boom in fancy tech learning and everyone having mics on their devices, it’s easier than ever to eavesdrop on what you’re typing on your keyboard. This study used advanced tech to listen in on laptop typing using just a smartphone’s mic, and dang, it was right about what was typed 95% of the time! 

Even over Zoom, they nailed it 93% of the time. Basically, folks can spy on what you’re typing with everyday gadgets. But don’t fret, the article also offers some ways to keep our typing safe from these sneaky listeners.


Unleashing the Strengths of Unlabeled Data in Pan-cancer Abdominal Organ Quantification: the FLARE22 Challenge

Doctors need to measure organs in the belly to diagnose diseases and plan treatments. AI (think of this as smart computer programs) can help speed this up, but many current ones need a lot of expert input and haven’t been tested everywhere. So, the bigwigs put together this massive competition in 2022, called the FLARE Challenge, to see which AI can do the job best, fastest, and with the least help. 

They used belly scans from folks all over the world, from over 50 medical teams. Here’s the cool part: some AI systems got it right 90% of the time using just 50 scans with labels and a whole bunch more without. 

These top AIs also did pretty well on tests from North America, Europe, and Asia. And get this, they can pull out important details about organs which usually take a lot of elbow grease to get manually. This means we can use more scans without labels, which makes these AI tools even better and solves the problem of not having enough labeled scans.


Large Language Models in Cryptocurrency Securities Cases: Can ChatGPT Replace Lawyers?

Researchers wanted to see if big computer brains (like GPT-3.5) can help out in legal cases, especially ones about cryptocurrencies. They checked two things: 1) Can the computer figure out which laws are being broken by just looking at the case details? 2) Do mock juries react differently to legal complaints written by the computer compared to ones by real lawyers?

Turns out, GPT-3.5 isn’t great at figuring out the broken laws, but it’s not totally off either. It just missed a few. On the writing front, it’s pretty good. Mock jurors didn’t really care if a computer or a human wrote the legal complaint. The bottom line: these computer brains can’t fully replace lawyers yet, but they might help make legal services cheaper.

This is the first study to really dig deep into how these computers can help (or not) in court cases, especially with securities law and shady cryptocurrency business.