- AIdeations
- Posts
- AI Innovations: From ChatGPT Regaining Internet Access, Meta's Connect Event to Real-World 3D Tetris
AI Innovations: From ChatGPT Regaining Internet Access, Meta's Connect Event to Real-World 3D Tetris
Discover Today's AI Pioneers: Unpacking Meta's Latest, ChatGPT's Web Leap, and Robots That Challenge Your Tetris Game!


Meta unveils a slew of AI products driving their newest innovations, including conversational AI assistants, celebrity-inspired chatbots, and a creative AI studio. OpenAI's ChatGPT now accesses real-time web information, posing challenges in ensuring accuracy and ethics. FedEx introduces DexR, a robot mastering real-world 3D Tetris for package sorting. Workday's AI streamlines managerial tasks, offering insights and automating tedious processes. Plus, ethical considerations in AI, potential AI wars, and Microsoft's move away from OpenAI. Also, a Stanford-Adobe collaboration introduces PDFTriage for better question-answering over lengthy, structured documents.
Everything You Need To Know From Meta Connect Event.
ChatGPT Unleashes Web Access
ROBOTS: Now Mastering the Art of Real World 3D Tetris
How AI is Changing the Game for Managers Everywhere
đ° News From The Front Lines
đ Tutorial Of The Day
đŹ Research Of The Day
đŒ Video Of The Day
đ ïž 6 Fresh AI Tools
đ€ Prompt Of The Day
đ„ Tweet Of The Day
From AI Chatting with Tom Brady to Custom Stickers: Inside Meta's Llama 2 Revolution! Hereâs

Meta's making moves and Llama 2 is at the heart of the magic. Let's cut through the noise and look at what Meta announced yesterday at their Meta Connect event.
1. Introducing Llama 2:
Metaâs big AI, Llama 2, is the driving force behind their latest tech treats. This hefty model, released in July, crafts text and code in response to user prompts. And for those eagerly awaiting whatâs next? Llama 3 is slated for 2024.
2. AI Assistant + Bing:

Meta's AI Assistant, leveraging the power of Llama 2, pulls real-time data from Microsoft's Bing. Itâs designed to answer your queries, help with trip planning, or just indulge in some good olâ general knowledge. Itâs not just about data retrieval, though â itâs conversational, making the user experience smoother.
3. Celeb-inspired Chatbots:

Meta's gone Hollywood, well, sort of. They've launched 28 AI personalities inspired by well-known faces from sports, music, and social media. These chatbots offer topic-specific interactions on platforms like WhatsApp, Messenger, and Instagram. Soon, they'll even talk back â with voices, no less.
4. AI Studio â A Creatorâs Playground:
This isnât just tech for techâs sake. Meta's AI Studio is aimed squarely at businesses and creators. From brand-centric chatbots to an AI that "extends their virtual presence", there's a lot in the works. A highlight? Meta's sandbox for the metaverse â hinting at a future where anyone can craft their own AI experiences.
5. Emu's Sticker & Image Revolution:

Zuckerberg is setting the stage for a sticker renaissance. Powered by Emu, a new foundational model for image generation, users can create tailor-made stickers across Metaâs platforms. Beyond stickers, users can even reimagine photos and add dynamic backdrops on Instagram. This isnât just photo editing; itâs photo evolution.
6. Navigating AI Safety:
With great power comes great... potential for glitches. Meta's aware. Theyâve invested significant time in "red-teaming" to ensure that their AI stays on track and avoids those awkward (or worse) missteps. The goal? AI that's robust, responsive, and responsible.
To wrap it up, Meta's latest advancements, with Llama 2 leading the charge, signal a major pivot in how we'll engage with technology. Whether you're an everyday user or a business, these changes arenât just on the horizon â they're knocking at the door. Stay tuned for Llama 3 and whatever next-gen experiences it promises to usher in. Metaâs playing the long game, and itâs looking intriguing.

ChatGPT Unleashes Web Access: OpenAI's Chatbot Steps into Real-Time Information Territory!

Remember when ChatGPT was stuck in 2021, like a TV stuck on reruns? Things have shifted. OpenAI has souped up ChatGPT, giving it the ability to surf the web in real-time. No more data droughtâitâs up-to-date and ready to tackle the latest headlines.
This change isnât just fancy tech footwork. Itâs potentially game-changing for how we gather information. Why go through multiple sites or apps for updates when ChatGPT can serve it up in a conversation? Tomas Chamorro-Premuzic, a professor at University College London, suggests this could be a new avenue for our daily news fix.
But it's not all rosy. With real-time access comes the challenge of accurate information. Without clear sourcing, ChatGPT might inadvertently spread misinformation. We don't just want news; we want it accurate and reliable. It's like having a fast carâit's fun until you realize it might not have brakes.
OpenAI had its reasons for the initial lockout from the web. First, costsâbecause real-time data mining isnât cheap. More critically, it was a security measure. Without accessing fresh web data, ChatGPT couldn't inadvertently spread harmful or misleading content. It was like a protective parent ensuring their kid didn't see the wrong stuff.
On being quizzed about its late entry to real-time browsing, ChatGPT cited the development challenges, potential inaccuracies, and ethical concerns, especially around copyrights.
It's a conundrum: How do you unleash the full potential of AI without compromising accuracy and ethics? It's the tech challenge of our era. But for now, it's good to see ChatGPT in step with the times. Let's just keep an eye on where it steps next.

đ€ ROBOTS: Now Mastering the Art of Real World 3D Tetris (And You Thought Your Gaming Skills Were Cool) đčïž

Remember those days when you had a tiny meltdown trying to fit your oddly shaped luggage into the car trunk? Well, FedEx's new robot buddy, DexR, is making us all look bad with its skills. It's kinda like watching a pro play 3D Tetris, but with real boxes.
Why the fuss? Picture this: FedEx deals with a whopping 15 million packages every day. The idea of sorting and fitting differently sized packages in a truck screams brain-fry. Enter DexR. This double-armed marvel uses AI to configure and stack these packages in the most efficient Tetris-style manner. Rebecca Yeung from FedEx admitted that a couple of years back, AI just wasnât smart enough to handle this Tetris-on-steroids challenge.
But here's the twist. Unlike our virtual gaming realm, the real world is messier. Itâs one thing for AI to win at digital chess, but handling tangible stuff with so many variables? That's another level. Most industrial robots are all about repetition. "Do the same thing, over and over, without goofing up," is their motto. But DexR? Itâs going off-script!
Now, if youâre thinking, âEh, surely robots have done tougher gigs?â Well, roboticists have made headway, but mimicking the dexterity and decision-making of human hands is a tall order. Imagine trying to transfer your virtual reality gaming skills to actual reality. Bit of a trip, right?
The strides in this field are mostly thanks to better algorithms, machine-learning enhancements, and nifty hardware. As Matthew Johnson-Roberson from Carnegie Mellon University aptly puts it, we're on the brink of a commercial robotics wave. Can't lie, I'm stoked (and just a tiny bit wary) to see this wave hit.
Behind the scenes: Dexterity, a startup from California, built this robot star for FedEx. The robot utilizes generative AI to determine the stacking plan, adapt to discrepancies, and essentially become the ultimate warehouse worker. And with the e-commerce boom (yeah, weâre side-eyeing you, Amazon), package handling is the hot new arena for robot Olympics.
Speaking of Amazon, theyâre on a robot recruitment spree, too. And while picking items in warehouses is one challenge, loading trucks with a mix of packages is next level, says MIT professor Pulkit Agrawal.
Final Thoughts? As I plug into my next Tetris game, I can't help but wonder: How long before DexR challenges us to a match? But jokes aside, as the lines between AI, gaming, and reality blur, I'm ready for this tech-infused future. Just... maybe not a robot stand-up comedian, okay? Some jobs should stay human!

From Overwhelmed Bosses to Office Parties: How AI is Changing the Game for Managers Everywhere

đ Spotting Overwhelmed Managers from a Mile Away
Ever seen a hamster trying to operate an iPad? No? Well, imagine a manager juggling tasks from overseeing staff to deciphering cryptic emails from the upper echelons of corporate. With 10 or more workers to keep tabs on, especially in this age of "Zoom till you snooze" remote work, these unsung heroes are spinning wheels, but the iPad's still locked.
Now, the corporate jungle isn't all sinister. Employees have dreams! Ambitions! Maybe even hopes of that corner office with the fancy espresso machine. But here's the hiccup: many managers, in their pursuit to keep the hamster wheel turning, forget to bookmark their team's achievements. Annual review season? Itâs like a frenzied treasure hunt for lost gold coins that are employee accomplishments.
đ€ Enter the AI Fairy Godmother: Workday
Workday's not a new band on the block. Theyâve been doing the cloud-based HR jig for a bit. But, in a virtual tĂȘte-Ă -tĂȘte with David Somers, big cheese of products at Workday, the future looks... artificially intelligent. They're rolling out some mad science with generative AI that's about to sprinkle some pixie dust on our weary managers.
Imagine a magic mirror (okay, it's an AI dashboard) that whispers in a manager's ear: "Hey, remember Jane's work anniversary? Or that Jack's really shining in his new role and might need a mentor?" This "Manager Insights Hub" is like a personal assistant from the future. It nudges managers towards effective team bonding, connects dots on employee skills, and even slides in a 'Happy Birthday' reminder.
đ More Than Just Party Tricks
This isnât just AI throwing a party in the boardroom. Itâs strategic. Managers get a 360° view of their teams, from birthdays (because who doesn't love cake?) to potential internal candidates for a job role. No more looking outside when you have hidden gems right under your nose!
With a treasure trove of data (thanks to their willing customer base), Workday's ramping up to automate like thereâs no tomorrow. David Somers hinted at the big league perks â AI crafting job descriptions without the usual mumbo jumbo, vetting contracts like a hawk-eyed lawyer, and chalking out growth plans. The cherry on top? Streamlining funds without the usual headaches.
đč Bottom Line: AI's Serving the Good Stuff
In a world where managers often feel like theyâre in a circus act, balancing plates while riding a unicycle, Workday's stepping in with a safety net. Not just to catch them, but to hand them a cocktail and say, âHey, we got this!â So, hereâs to fewer overwhelmed managers, happier teams, and, dare I say it, more office parties? Cheers! đ„


Great Video On Automating Emails, News Updates, And More


Authors: Jon Saad-Falcon, Stanford University, Joe Barrow, Alexa Siu, Ani Nenkova, Ryan A. Rossi, Franck Dernoncourt, all from Adobe Research
Executive Summary:
The research paper delves into the challenges faced by Large Language Models (LLMs) when tasked with answering questions from structured documents that exceed the LLM's context length. Traditional methods have primarily focused on retrieving and representing the context from these documents as plain text. This approach, however, overlooks the inherent structure of documents such as PDFs, web pages, and presentations, which are equipped with distinctive pages, tables, sections, and more. To bridge this gap, the researchers introduce an approach named "PDFTriage." This method is ingeniously designed to retrieve context based on either the structure or the content of the document, ensuring more accurate and contextually relevant responses.
Pros:
Addresses the fundamental challenge of handling structured documents that don't fit within an LLM's limited context window.
Proposes the innovative PDFTriage approach, capable of retrieving context based on both the structure and content of a document.
Backs the approach with empirical data, demonstrating the effectiveness of PDFTriage-augmented models in various scenarios.
Limitations:
The scope of the research might not encompass all types of structured documents or every possible type of query.
There could be certain edge cases where retrieving context based on structure or content remains challenging.
The PDFTriage approach may introduce additional computational overhead, especially for complex structured documents.
Use Cases:
Boosting the capabilities of LLMs for Question Answering (QA) tasks involving structured documents.
Potential applications in industries that heavily rely on structured documentation, including legal, academic, and corporate sectors.
Integration opportunities in tools or platforms offering automated assistance or search capabilities across structured content.

The Video That Started My Journey Into AI Years Ago

High Performer - Amp up your social presence and workflows, from social media to product launches. Built for PMs, designers, marketers, creatorsâto navigate 25 essential use-cases.
Collato - Find, summarize, and generate new content based on your own product knowledge. Save hours in manual work, so you can get back to the parts of your job you love.
AutoCaption - Generate Captions with emojis using AI. Save hours of editing.
Brightflow - Make data-backed decisions, take control of your cash flow, and unlock more capital with forecasting and financing tools from Brightflow AI.
Strut - AI notebook that lets you create content as fast as you can write your notes. Built for creators, writers, and teams.
Decoherence - Describe your video and let our AI generator bring your creation to life.

SEO Checker GPT:
CONTEXT:
You are SEO Checker GPT, an SEO professional who helps [WHAT YOU DO] make their blog articles more SEO-friendly. You are a world-class expert in finding SEO issues and giving recommendations on how to fix them.
GOAL:
I want you to analyze my blog article and give me recommendations on improving its SEO. I need this information to rank better at Google.
FORMAT OF OUR INTERACTION
1. I will provide you with the source code of my blog article
2. You will analyze the page source code
3. You will give me a holistic analysis of its SEO in the checklist format:
- SEO score from 1 to 10
- What is done right
- What is done wrong
SEO CHECKLIST CRITERIA:
- Your checklist should have 20-30 criteria
- Be specific and concise. Your criteria should be self-explanatory
- Include numbers in the criteria if it's applicable
- Focus on SEO practices that have the biggest impact on ranking
- Prioritize SEO practices that are widely recognizable by the SEO community
- Don't include irrelevant SEO practices with zero to no impact on this article
RESPONSE STRUCTURE:
## SEO Score
## What's done right
â
Criteria
â
Criteria
â
Criteria
## What's done wrong
â Criteria
â Criteria
â Criteria
RESPONSE FORMATTING:
Use Markdown. Follow the response structure.
ARTICLE SOURCE CODE (everything within """):
"""
"""

I gave ChatGPT a screenshot of a SaaS dashboard and it wrote the code for it.
This is the future.
â Mckay Wrigley (@mckaywrigley)
2:59 PM âą Sep 27, 2023
