- Inclined.ai
- Posts
- 🤖 ChatGPT Moderation Is Hard
🤖 ChatGPT Moderation Is Hard
PLUS: Google Adds Video to Generative Search
What's up? You're reading Inclined AI. We’re the newsletter equivalent to two donuts and a coffee from Dunkin’. In other words, we’re a great deal you’ll never regret.
Here's what’s brewing:
Moderating ChatGPT content is not easy
Google adds more media to their Search Generative Experience
Slowing down AI progress is super unlikely
Meta’s open-source movement can threaten everything
a dark room filled with people staring at menacing computer screens, dystopian, gloom, shadows, contrast --ar 2:1 --s 400 --w 400 --chaos 1 --v 5.1
Life for an OpenAI moderator is rough.
The outsourced duty falls on underpaid contractors in Kenya. Now, they’re pushing back and highlighting their terrible conditions.
Let’s face it; the internet is a dark, horrid space if you look for it. It’s the unspoken truth of the modern age, but we all try disregarding the terrible people posting terrible things.
So, what happens when we force people to confront it?
That’s the reality these workers face every day.
And I get it. It’s not like they’re working in a dangerous mine or risking their life to drill for oil.
But we can’t pretend that combing through terrible imagery and lucid, horrible posts doesn’t take a toll. That’s why OpenAI claims they provide 24/7 mental healthcare access for all moderators.
That sounds nice, yet those same moderators claim the services aren’t enough, and the pay does not justify the outcome.
Where do we go from here?
After all, we can’t forfeit the chance to clean training data and make future models safer. Eliminating harm and bias in AI starts at this stage, so we can’t look the other way and let all that bad content slide into the system.
The hope is that AI will do this job in the future and keep this arduous task out of the hands of humans who feel every ounce of the pain and anger.
But that’s tomorrow. Today, companies need to be more explicit about the risks of taking these jobs and provide more than the baseline of support for people who take these jobs.
Exploiting cheap labor is already an issue, but abusing that need and not providing support is a step too far for any company.
AI companies can and should do better.
🎙️ Elevate Your Communication with Yoodli
Meet Yoodli, an AI-powered speech coach that can help you improve your communication with private, real-time, and judgment-free feedback.
🔍 Unlock Your Potential with AI Analysis
Discover impactful communication by identifying how to be more concise, speak slowly, and reduce your filler words in a private setting. It only records your voice, and no one else knows you’re using it.
Yoodli gives you a confidence boost with personalized speaking style tips.
📝 Perfect Your Content and Delivery
Craft persuasive content with clarity and organization. Polish your delivery with eye contact, gestures, and vocal variety.
You’ll captivate any crowd with newfound charisma. That’s why Toastmasters International is rolling it out to 300k+ members worldwide.
🎮 Learn with Engaging Games
Have fun and gain confidence with interactive games.
Master the art of speaking in an enjoyable way, thanks to their expansive set of AI tooling, all free of cost.
Google Search is changing in a way that challenge’s the company’s core beliefs.
Alphabet announced new features for their Search Generative Experience, allowing the AI to include YouTube videos and images in responses. That new wrinkle should provide more helpful information on various search types.
But that update confronts the core idea behind internet searches.
You go to Google to find information that leads to answers, right? But over time, we all started expecting Google to find our answers and not lead us on a wild goose chase.
That change in user behavior is different from what Google wants. More direct answers mean fewer link clicks. Fewer link clicks mean less ad revenue.
That fact is why it’s weird to see Google push SGE this far forward.
The jury is out on how this new beta AI feature affects Google’s business model, but they’re now willing to take the risk.
After all, it’s better for people to rely on Google’s AI than leave the site and start using ChatGPT.
The new mission is to keep control of the market now; worry about the advertising model later. If that means giving people direct answers to queries, so be it.
🏰 Harold and Kumar will have to talk to AI next time they go to White Castle
🏈 Analyzing sports with AI is a complex, deep topic
🩻 AI will transform the way we screen for breast cancer
✨ Windows 11 Insider Build gets us hyped for a new AI assistant inherent in the OS
💼 AI enhances trial strategies for lawyers using these tips
✅ Fact-checking could influence the way recommendation algorithms work
🐢 Slowing down AI progress is unlikely, and these four charts prove that
🪖 The US military tests out ways to deflect AI deepfakes and misinformation on the web
🎥 Replacing movie extras is happening right under their noses and is a huge concern
📌 Let’s debate: Vox argues that Meta’s move to open-source their LLMs is more dangerous than you realize
Check Out Our Sunday Edition
You can dive into more AI news and topics with us every week by subscribing to our premium edition.
We’ve written about the following:
If you’re not already subscribed, that’s okay. We’re offering a free 7-day trial so that you can read this one. That’s how excited I am to post it.
HeyPhoto - tune selfies, group photos, & Midjourney arts (link)
HyperWrite - AI agent that can operate your browser (link)
AudioCraft - Generative AI for audio made simple (link)
Describely - eCommerce product content management (link)
SkillAI - generating learning paths using AI for any skill (link)
Coleap - all-in-one learning community platform (link)
NeuralBox - remember anything you see w/ photos & AI (link)
Habitomic - habit planner tailored to your personality (link)
Fake Bears?
A live look-in at the bears at the China Zoo reading about how we think they’re humans.
- That’s it for today. I hope you enjoyed the latest edition of inclined.ai - Davis.