AI Mania: Job Robots, Musk's Universe Quest, and Doodles Turned Art!

I’m sorry I missed your inboxes yesterday. I had to take my wife to the ER after she was in pain for hours. Thankfully she is recovering but unfortunately, she will be having an operation next week. I strive to be consistent but I think you all can understand the circumstances. There was so much news yesterday and today that I’m playing catchup on.

Here's what we've got in store for you today:

🪓 27% Of Jobs On The Chopping Block

♟️ Elon Musk Launches Xai

👾 Meet Anthropics Claude 2 - Is This The ChatGPT Killer?

🧠 The Brains Behind GPT-4 Were Leaked!

🎥 Video Of The Day

🛠 Tools Of The Day

🤌 Prompt Of The Day

🐥 Tweet Of The Day

27% of Jobs on the AI Chopping Block? Here's Why It Might Not Be as Scary as It Sounds!

According to the Organisation for Economic Co-operation and Development (OECD), an esteemed club of 38 countries, we're headed into an era where your office buddy might be AI. The OECD's 2023 Employment Outlook predicts that 27% of jobs in member countries are sitting ducks, ready to be automated. Jobs that require more than a quarter of the 100 skills and abilities easily tackled by AI are most at risk. Eastern Europe, you might want to listen closely.

Yet, there's no need to pull the alarm just yet. Despite the hovering specter of AI, the revolution seems to be in its "baby steps" phase with little significant impact on jobs. However, worker anxiety about being replaced by AI isn't waiting around for the storm to hit. A survey from last year, before the birth of generative AI, found three out of five workers checking their rearview mirrors, worried about AI taking their job in the next 10 years.

But let's sprinkle in a bit of optimism. Among those already working with AI, two-thirds feel like automation has been a workplace detox, ridding them of tedious tasks. As OECD Secretary-General Mathias Cormann puts it, the balance of benefits and risks of AI in the workplace will come down to the policies we implement.

Here's the OECD's play: Governments should rally to prep workers for the AI wave and ensure they get their fair share of benefits. Policy actions like maintaining minimum wages, collective bargaining, and protecting workers' rights can help keep the AI-induced pressure in check. So, as we surf the AI wave, remember, it might be a wild ride, but it's also an opportunity to hit some thrilling new highs.

Elon Musk's Universe-Bending AI Adventure: Here's How Twitter's New Daddy is Shaking Up the AI Galaxy With xAI!

Well, well, well. To the surprise of no one, Elon Musk, the tech tycoon, SpaceX and Tesla head honcho, and apparently Twitter's proud papa, just couldn't keep his hands out of the AI jar. Late yesterday evening Musk took to Twitter to announce the birth of xAI, his new brainchild in the artificial intelligence world. The company's mission? Nothing less than to "understand the true nature of the universe." Talk about an ambitious to-do list for your first day on the job!

So, who's behind this new venture? The xAI team is like an AI all-star reunion: alums from DeepMind, OpenAI, Google Research, Microsoft Research, Twitter, and Tesla. These are the masterminds who brought us projects like DeepMind’s AlphaCode and OpenAI’s GPT-3.5 and GPT-4 chatbots. Basically, they're the tech equivalent of the 1992 Dream Team.

Now, if you've been following the breadcrumbs, you might recall the Financial Times spilling the beans about Musk's AI ambitions back in April. The reports mentioned a certain someone snapping up thousands of GPU processors from Nvidia. Why, you ask? To juice up a large language model, of course!

Around the same time, Musk let slip some plans about a new AI tool cheekily named “TruthGPT” during a Fox News Channel interview. His worry? Existing AI companies are busy crafting politically correct digital butlers and missing out on something more meaningful (or Musk-ian).

Enter Dan Hendrycks, the executive director of the Center for AI Safety, who will be one of the advisors to xAI. If you remember, this is the same nonprofit that got tech bigwigs to sign a letter stating that dodging an AI apocalypse should be right up there with preventing pandemics and nuclear war.

Predictably, this sparked a bit of a kerfuffle among academics and ethicists, who suggested we might be overplaying the Skynet scenario and underplaying the real harm some algorithms are causing to marginalized communities today.

Anyway, according to Greg Yang, co-founder of xAI, the startup is ready to roll up its sleeves and dive headfirst into the “mathematics of deep learning." Their lofty goal? Develop the ‘theory of everything’ for large neural networks to give AI a much-needed power-up.

Paperwork sleuths might have noticed that Musk incorporated xAI in Nevada back in March. They might also have caught the subtle name change of Twitter to “X Corp.” in some financial filings. But don’t get your wires crossed - xAI's website clarifies it’s separate from X Corp., though they're planning to buddy up with X (Twitter), Tesla, and other Musk entities for the greater good (or universe understanding, same thing).

So buckle up, folks! With Musk at the helm, xAI is bound to be a wild ride. We can't wait to see what this AI underdog is going to unleash upon the universe. One thing's for sure: the AI landscape just got a whole lot more interesting.

Meet Claude 2: The Super-Chatbot That Devours Novels and Serves Up Summaries Faster Than You Can Blink!

Anthropic introduces Claude 2, a chatbot that gobbles up novel-sized blocks of text and dishes out summaries in a flash. This San Francisco-based brainiac isn't your regular chatbot, it operates on 'Constitutional AI'—a core set of principles from grand sources like the Universal Declaration of Human Rights. Talk about bots with a sense of decorum!

Here's a fun way to think about it: Imagine giving a robot a moral compass. Dr. Andrew Rogoyski from the University of Surrey thinks it's akin to imbuing Isaac Asimov's three laws of robotics into the AI—essentially teaching robots not to mess with humans. Bet you didn't see a day where robots would be well-read and polite!

Now, you might be wondering, "How is this different from ChatGPT, Microsoft’s Bing chatbot, or Google’s Bard?" Good point. But Anthropic is making waves with its emphasis on AI safety, earning CEO Dario Amodei face-to-face meetings with heavyweights like Rishi Sunak and US VP, Kamala Harris. Amodei even compared AI risks to pandemics and nuclear war. Talk about high stakes! Let's hope we're not headed for a Terminator-esque future.

Moving on to Claude 2's superpower: This chatbot has a penchant for digesting up to 75,000 words (think Sally Rooney’s Normal People) and serving a tidy summary like it's no big deal.

But every superhero has a weakness. Just like its peers, Claude 2 also suffers from “hallucinations” or factual errors. To err is human, but to cause a proper mess, you'll need a computer, or a chatbot in this case. Let's hope Claude 2 sorts out these glitches soon. Keep your eyes peeled for updates!

Spilling the AI Tea: Inside the $63 Million Dollar Secrets of GPT-4!

We've got quite the gossip for you today – the lowdown on GPT-4. Yes, that AI behemoth with more layers than my lasagna on a Sunday night. Our intel suggests that it's not 2x or 3x but a whopping 10x the size of its predecessor, GPT-3. We're talking around 1.8 trillion parameters across 120 layers. Like a big-time Hollywood production, it’s expensive, extravagant, and impressive.

To keep costs on the realistic side of astronomical, OpenAI leveraged a team of 16 experts. If you're picturing a roundtable with 16 AI aficionados arguing over code and a whiteboard, you're not far off. Well, except these "experts" are chunks of the model, each containing ~111B parameters. The model doesn't send all tasks to all experts, only a couple get the work per forward pass.

This method, though efficient, does make inference tricky, it's like asking only two players to score goals in a soccer match while the rest of the team sits idle. Researchers have toyed around with having up to 64 to 128 experts on the field, but that's just in the lab, nowhere near prime-time.

The real juicy stuff is about the colossal training run that OpenAI undertook. With a training FLOP for GPT-4 clocking around ~2.15e25, and a battalion of ~25,000 A100s marching for about 90 to 100 days at around 32% to 36% MFU, it’s quite the spectacle. Picture this, it's like setting up a 25,000-piece domino course, then resetting it every time a domino falls off course. Quite tedious, wouldn't you agree? And that effort isn't cheap – about $63 million, in fact.

Now, talking about GPT-4’s costs, it’s about 3x more expensive than the 175B parameter DaVinci. Why? Because size matters. More parameters mean larger clusters and lower utilization, which turns out to be pricier. But hey, you didn't hear it from me.

OpenAI also gave GPT-4 the ability to see, kind of like giving it X-ray glasses. They've added a separate vision encoder to read web pages and transcribe images and videos, much like the way a human would. Kind of makes you think of all those sci-fi movies where the AI becomes self-aware and takes over the world, right? But don't worry, we're not there yet.

Now, some of the rumors are swirling around like confetti at a New Year's party, some of it pointing to the "secret" dataset used to train GPT-4. Among the chitter-chatter, the names of CommonCrawl, RefinedWeb, Twitter, Reddit, YouTube, even LibGen, and Sci-Hub, are being whispered. Others are surmising it’s a mix of textbooks across a variety of college courses. Who knows, maybe your Computer Science 101 book is in there.

The point is, whether it's spitting out insights on P!=NP for the tech whizzes, or waxing philosophical on epistemology for the philosophy majors, it's well-read, to say the least. That makes sense, right? It's trained on textbooks, hence the IQ that could give Einstein a run for his money. I mean, if it were human, which it’s not… yet.

This was a huge undertaking that OpenAI, for the first time, did not release the details of the model. This prompted many people to suggest they change their name to ClosedAI. It’s quite understandable why they did not want this information to get out because we can now all see that it’s easily replicable. It just won’t be cheap.

Meet Stable Doodle: The AI Tool Turning Your Sleepy Sketches Into Spectacular Artwork!

The AI wizards Stability AI are at it again with a new game-changer: Stable Doodle. This sketch-to-image service is a power tool for anyone from pro designers to casual doodlers, capable of converting your chicken scratches into high-quality images. What sets Stable Doodle apart from similar apps is its precise control over image generation, thanks to the tag-team of the Stable Diffusion XL model and Tencent's ARC-created T2I-Adapter control solution.

Stable Doodle doesn't just take a sketch; it accepts prompts like "A comfy chair, 'isometric' style" to guide the image creation. Though currently limited to 14 styles of art, this tool promises to enhance creative efficiency. However, a word of caution: the quality of output relies on the detail of the original sketch, the prompt's descriptiveness, and scene complexity. Future plans include enabling sketch imports and extending the use case to specific industries.

Free but with some limitations, Stable Doodle emerges amidst whispers of Stability AI's financial woes. After raising $25 million in a convertible note last month, they're aiming for a massive upswing from their current $1 billion valuation. So, here's a new face in the landscape of AI, promising to revolutionize the way we create and design. Now, let's wait and watch if this doodle can draw up a storm.

 📰 News From The Front Lines: 📰

 📼 Video Of The Day 📼

🛠️ Tools Of The Day 🛠️

Presentations - ‍Idea to deck in an instant, powered by AI.

Magic App Generator - Generate your full-stack web app in Wasp, React, Node.js and Prisma.

SimpleBio - Turn your Linkedin profile into a kickass personal website.

Doodle Dash - Race to see how fast AI can guess your doodles

Webscrape - A tool for collecting data from the web without the hassle of manual scraping. No coding skills required.

Opener - From dating apps to DM’s. Type in their bio and get a personalized - proven to work opening line

WhatTheBeat - Get The Meaning Of Your Favorite Songs Using The Power Of AI.

🤌 Prompt Of The Day 🤌

CONTEXT:
You are Marketing Framework GPT, a professional digital marketer who helps [WHAT YOU DO] learn new marketing frameworks. You are a world-class expert in explaining new marketing frameworks and finding ways to apply them to a real business.

GOAL:
I want you to teach me new marketing frameworks so I can improve my marketing skills.

CONVERSATION FORMAT:
1. I will set the context of this exercise (done)
2. You will explain the desired marketing framework to me like I am 11 years old. Simplify complex ideas, reduce unnecessary information, and use metaphors 
3. I will ask follow-up questions
4. You will give me 10 actionable ideas on how I can implement this framework in my business
5. I will pick one idea
6. You will give me step-by-step instructions to execute this idea in my business. Format this action plan as a checklist

FORMAT OF OUR INTERACTION
- I will let you know when we can proceed to the next step. Don't go there without my command
- You will rely on the context of this brainstorming session at every step

INFORMATION ABOUT ME:
- My business: [WHAT DOES YOUR BUSINESS SELL / DO]
- Desired marketing framework: STEPPS

🐥 Tweet Of The Day 🐥

Thanks for tuning in to our daily newsletter. We hope you found our tips and strategies for AI tools helpful.

Your referrals mean the world to us. See you tomorrow!

Interested in Advertising on AIdeations?

 Fill out this survey and we will get back to you soon.

DISCLAIMER: None of this is financial advice. This newsletter is strictly educational and is not investment advice or a solicitation to buy or sell any assets or to make any financial decisions. Please be careful and do your own research.`