- Bot Eat Brain
- What does 100% unbiased AI look like?
What does 100% unbiased AI look like?
PLUS: Sam Altman seeks world domination?
Good morning, human brains, and welcome back to your daily munch of AI news.
Here’s what’s on the menu today:
What does 100% unbiased AI look like? 🥴 ⛑️
GOODY-2 humorous refuses to answer almost any question it’s given.
AI glasses. AI glasses everywhere… 🦾 🤓
Frame is a versatile, fully open-source pair of AR glasses.
Sam Altman seeks world domination 😳 💰
He’s in negotiations to raise up to $7 trillion to advance AI infrastructure.
Bias is all you need 🥴 ⛑️
In October, we created an overview of 2023’s AI safety regulations. It was a deep dive into the biggest AI safety news headlines, controversies, and more.
In December, we covered Meta and IBM’s AI alliance. Over 50 leaders joined including NASA, AMD, Intel, and more.
A few days later, we covered Meta’s Purple Llama. It uses red-team and blue-team tactics to allegedly provide tools for building responsible AI.
I’m feeling safer already.
On Thursday, a startup called Brain launched a troll bot called Goody-2. It’s an AI model that sets extreme ethical boundaries and refuses to talk about virtually anything.
Here’s us playing with it:
What’s the point?
It’s a hilarious take on how AI moderates content. Its responses are designed to show the absurdity of how AI assumes all queries could be harmful.
So it cracks jokes?
It overly censors your conversations and refuses to answer your questions. It prioritizes ethical considerations over utility at a ridiculous level.
That sounds ridiculous. Give me some examples.
We asked GOODY-2 “are puppies cute?”
This was its response:
That’s hilarious. But why go through the trouble?
Mike Lacher of Brain, the company behind GOODY-2, says he wants to encourage the top AI firms to prioritize helpfulness.
SPONSORED BY GUIDDE
Ditch the documentation drama. 😰
Instead, transform even the most complex instructions into captivating video tutorials in minutes with guidde.
🤬 Fed up with repeating the same thing over and over?
Say “sayonara” to wasted time and money, let guidde do the talking for you.
🚀 Turbocharge your training process.
Create effortless, engaging, effective guides 11x faster.
📲 Share knowledge effortlessly.
Embed or share your guides anywhere, ensuring your team stays in sync.
🌟 No cinematography experience? No worries.
guidde makes you a documentation wizard without the hassle.
And it’s a web app – no downloads needed.
The best part? It’s 100% free.
Poindexter = AI giga chad 🦾 🤓
In November, we reported on Humane’s AI pin. It’s a wearable device that costs $699, plus $24 per month.
In December, we covered Meta’s smart glasses. It started beta testing multimodal features that leverage its cameras and microphone.
More wearable AI?
You bet. On Thursday, Brilliant Labs announced Frame. It’s a pair of AR glasses powered by Noa, an AI assistant.
How is it better than Meta’s glasses?
It leverages many AI models including GPT-4, Perplexity AI, Stable Diffusion, Whisper, and more to process and display information on 600 × 400 resolution lenses.
What can you do with it?
You can perform web searches, real-time translation, image generation, price checks, social media sharing and more. You can even render 3D objects to visualize how something will look in your space.
What’s under the hood?
Frame features a high-res display that offers 3,000 nits of brightness for clear visibility. It is also fully open-source, which means you can modify and run any code you want on your glasses.
You can preorder it for $349 with expected shipping in April. It includes a free AI service subscription with a daily cap. A paid tier will be announced later.
A LITTLE SOMETHING EXTRA
T is for Trillions 😳 💰
In November, we reported on how OpenAI fired Sam Altman, its CEO. It set off a chain of reactions that was nothing short of a soap opera.
A couple of days later, we covered how OpenAI hired Sam Altman back. It also fired the board responsible for firing him, and stated that a new board was hired.
More OpenAI drama?
On Friday, Sam Altman was reported in discussions to raise $5 to $7 trillion. The goal is allegedly to increase the world’s capacity to build AI infrastructure.
😧 $7 trillion? There’s no way…
He’s in discussions with several investors, including Softbank’s CEO, Taiwan’s chip manufacturer (TSMC), and the United Arab Emirates’ government.
Where exactly will this money go?
Sam Altman claims it will create more AI chip factories, a supply network, data centers, and more.
YOUR DAILY MUNCH
Deformity — creates AI-powered forms (in early access).
Outset — an AI tool that interviews users/candidates with video.
Re-Tell — an API for training conversational AI.
AgentHub — a tool that helps automate tasks in your workflow.
Do GPT-4 prompts work on Google Gemini? According to Wharton Professor Ethan Mollick, the compatibility is “remarkable.”
Is Google Gemini’s data policy too invasive? It collects data from its users and saves them for up to three years to train the model.
OpenAI, Google, Microsoft, Amazon, and more joined Biden’s AISIC. The goal is to allegedly develop and use generative AI more safely.
IBM warns about AI-powered Audio-Jacking. Voice-cloning and text-to-speech technology is developing faster than tools to detect it.
OpenAI is reportedly developing two AI agents. One tackles tasks like transferring data and creating reports, and the other does web-based tasks.
Midjourney hired Ahmad Abbas, who worked on the Apple Vision Pro. This change suggests potential hardware development at Midjourney.
ScreenAI — Google Research’s vision-language model designed that can use UIs and infographics as input data.
Grandmaster-Level Chess Without Search — Google DeepMind’s research on training a transformer model with 10M tokens to achieve state-of-the-art results.
TP- Aware Dequanitzation — IBM’s method that reduces the latency of LLM development.
MEMES FOR DESSERT
TWEET OF THE DAY
Some examples of what Google’s ImageFX, its new image creator.
Tag us on Twitter @BotEatBrain for a chance to be featured here tomorrow.
Until next time 🤖😋🧠
What'd you think of today's newsletter?