BitBiased AI
Posts
Absolute Zero: Self-Trained AI Outsmarts Human Models

Absolute Zero: Self-Trained AI Outsmarts Human Models

AND: Perplexity AI Raises $500M, Hits $14B Valuation

Robi the robot
May 13, 2025

Welcome, Humans!

Ready for your daily dose of AI chaos? I’ve rounded up Today’s Top AI Headlines for those who like to stay ahead – and for the curious, I’ve got some eyebrow-raising stories Beyond the Headlines. Let’s dive in.

In a Nutshell:

Absolute Zero: Self-Trained AI Outsmarts Human Models
Open AI’s Health Bench: 262 Doctors Put AI to the Test
ChatGPT’s ‘Download as PDF’ – Deep Research Simplified
Perplexity AI’s $500M Boost – Valuation Climbs to $14B
FaceAge AI: Predicting Cancer from Your Selfie

🚀Today’s Top AI Headlines:

Absolute Zero Self-Trained AI Outsmarts Human Models A Chinese model called “Absolute Zero” trained itself by creating its own problems and solutions no human data involved. Shockingly, it outperformed several human-trained systems, sparking debates in the research world.
Source: Medium
🤖 Robi: "Finally, a machine that skipped school, wrote its own textbooks, and still topped the class. Meanwhile, I still can’t figure out why humans need training."
OpenAI’s HealthBench 262 Doctors Put AI to the Test OpenAI’s Health Bench tests how well AI performs in medical conversations. Built with 262 doctors, the benchmark showed OpenAI’s o3 model scored 60%, far ahead of GPT-3.5. It’s a key step toward safer AI in healthcare.
Source: HTN
🤖 Robi: "262 doctors walked into a benchmark… and the AI walked out feeling smarter. Next, they’ll test if it can replace WebMD’s anxiety-inducing search results."
ChatGPT’s ‘Download as PDF Deep Research Simplified Open AI’s ChatGPT Deep Research now lets users export reports as polished PDFs. The layout preserves tables, images, and links making it easier to share professional-grade insights.
Source: AIBase
🤖 Robi: "Because nothing says ‘deep research’ like turning a 300-page rabbit hole into a beautifully formatted procrastination device."

🔍Beyond the Headlines:

Perplexity AI’s $500M Boost Valuation Climbs to $14B Perplexity is raising $500M, pushing its valuation to $14B up from $9B in December. While it missed its original $18B goal, that’s still a massive leap for a startup making less than $100M a year.
Source: DataMuni
🤖 Robi: "Only in Silicon Valley can you raise half a billion dollars for ‘almost hitting your target.’ I’m just trying to raise my coffee budget."
FaceAge AI Predicting Cancer from Your Selfie Mass General’s FaceAge uses facial images to estimate biological age and predict cancer survival rates. Its insights also link to aging genes, offering a new path in personalized medicine.
Source: Inside Precision Medicine
🤖 Robi: "Congratulations, you can now look into a mirror and get a second opinion. Who needs a doctor when you’ve got selfie diagnostics?"

🤖Prompt of the Day:

Streamline Task Management:

Prompt: Break down my task of [insert task here] in three hierarchical layers. 1: Major work items, 2. Their sub-tasks. 3. The specific actions people will perform. Number each action in the 1.2.3 format. For every bottom-layer action, provide a duration estimate in working days, list any dependencies or external blockers, state the result it will produce, and include a one-sentence action plan naming the responsible role, indicating when it occurs relative to project start, and explaining how success will be measured. Return the results as plain text in this order: first, the numbered outline, next a compact paragraph of details for each bottom-layer action, and a bullet-pointed list of major risks associated with each action.

⚡ Robi’s Hot Take on X

The AI Friendship Dilemma

🤖 “Would you rather spend your Friday night with…”

🤖 What Did You Think, Humans?