What's up? You're reading Inclined AI. This is your friendly reminder to water your plants and give them the proper amount of sunlight.

Here's what we heard through the grapevine:

• OpenAI wants to open up a market for models

• Discover the secret AI factories that power development

• Project Tailwind is Google’s new, big-time AI-first notebook

• ElevenLabs raises fresh funding

Be a sponsor of our newsletter.

OPENAI CONTINUES TO COME UP WITH WAYS TO MAKE MONEY

You hear the phrase, “OpenAI wants to make an AI app store,” and your stomach flutters. It’s safe to assume they’ll capitalize on the rising demand for useful plugins and give developers more space to explain their products.

Nope. The new “app store” OpenAI is dabbling in is for AI models like GPT-4.

That’s less exciting than a plugin store…

…But fits their agenda. When Microsoft put billions of dollars into OpenAI, they expected profits. That’s the goal here.

Don’t get confused; the mission became money when OpenAI opened up a for-profit arm. You and I are worth ~$20 a month. Large corporations are worth a sh*t ton more and are easier to please. 

Hence the reasoning behind their new model marketplace. While nothing is concrete, the rumor shows OpenAI is shifting staffing toward serving enterprise clients.

This new storefront would allow people to access models with specific functions, like an AI tutor built by Khan Academy that trained on their lesson plans.

You wouldn’t pay for these Large Langauge Models because ChatGPT fills that need for you. But companies like models with a narrow focus that fill a gap they’re looking for, and that is how OpenAI makes more cash.

So don’t get too excited by this rumor, but watch for updates from OpenAI.

All signals show they’re moving away from consumer features and pushing to serve enterprises. That’s not great.

If that were the focus, we wouldn’t get ChatGPT, but I don’t think they’ll move away from the crowd soon. We provide them with data that makes their models better.

That’s not going away.

THE HUMANS BEHIND THE MACHINE

Data is the oil that powers neural networks and pushes AI forward. Well-labeled data is the difference between a failure and ChatGPT, so how does that happen?

It turns out humans are behind all this work, but the picture painted by a recent article in The Verge is anything but rosy.

Millions of people annotate data for AI projects that conceal their purpose and push for the lowest possible cost. The ecosystem relies on cheap labor to label data and keep their mouths shut.

It’s a rough reality we need to wrestle with today.

These people don’t consider themselves workers. They’ve coined the term “taskers" to fit their feelings on the matter because the startups cut up projects into siloed subtasks. That structure keeps anyone from knowing what they’re working on for these companies.

A job without a clear purpose is mentally draining.

All of that so these data farms can collect enough training data for the lowest cost to hand back to their customers, Google, OpenAI, and Anthropic.

The more significant issue is the rapid development pace.

Earning $1-$3 an hour in Kenya is not the worst job in the world, but that pay was never consistent, and now it’s non-existent. As AI develops, the need for more detailed data grows.

Tasker jobs are moving out of these regions and finding cheaper work in Southeast Asia. Many companies, which make millions of dollars from these work centers, did so without warning.

It’s parachute production that hurts more than it helps. We need humans to help AI models learn, but we can’t keep cutting corners like this.

I recommend reading the entire piece, but we’d need a full 30-minute newsletter to review all the details.

Quick Nuggets

☀️ There’s slight optimism in the EU after new AI regulations pass

🎞️ Vimeo script generator will make marketing filler more abundant

🦜 Parrot raises an $11m Series A to push AI-powered transcription forward

💨 Project Tailwind is Google’s highly anticipated AI-first notebook that’s set to release

🤔 WIRED, a magazine, thinks that Midjourney’s magazine is soulless. Yep, not a conflict of interest at all…

🍎 STEM teachers can get help grading using AI for data analysis

💸 ElevenLabs raises $19m and announces a new detection tool

🎶 Machine learning tries to uncover the secrets behind hit songs

☎️ Portland launches AI to handle non-emergency calls for their emergency call service

💧 Oops, ChatGPT and Bard leak Windows 11 and Windows 10 pro keys

🔥 Fresh Products

• Factiverse - find factual mistakes in AI-generated content (link)

• Jarside AI - generates SEO-optimized articles in a minute (link)

• Zeno - collaborate & chat with your knowledge base (link)

• Noted Returns - never miss a return window again (link)

• The Last Pay Wall - AI news summaries (link)

• TwinSync - Video Face Replacement Tool for Discord (link)

• MonAi - AI expense tracker (link)

• Typefully - effortlessly publish on Twitter & LinkedIn (link)

Good Content, Cyberpunk Disney

Dang, Lilo & Stitch have seen better days, huh?

_{That’s it for today. I hope you enjoyed the latest edition of inclined.ai - Davis.}