AI Users Want it to “Roast Me”. .
Free AI Tutorials | AI Academy | Advertise | AI Agent No-code n8n
Plus: AI Users Want it to “Roast Me”
Read time: 5 minutes
PhD-level experts spend years solving complex, non-googleable problems. But now, with advanced reasoning models like GPT-o3 and Gemini-2.5-Pro, AI is outsmarting them, achieving accuracy rates up to 2x higher.
What are on FIRE 🔥
IN PARTNERSHIP WITH HUBSPOT
How can AI power your income?
Ready to transform artificial intelligence from a buzzword into your personal revenue generator
HubSpot’s groundbreaking guide “200+ AI-Powered Income Ideas” is your gateway to financial innovation in the digital age.
Inside you’ll discover:
-
A curated collection of 200+ profitable opportunities spanning content creation, e-commerce, gaming, and emerging digital markets—each vetted for real-world potential
-
Step-by-step implementation guides designed for beginners, making AI accessible regardless of your technical background
-
Cutting-edge strategies aligned with current market trends, ensuring your ventures stay ahead of the curve
Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.
AI INSIGHTS
🦠 AI Outsmarts PhD-level Virologists, Raising Biohazard Fears
The study was conducted by researchers at the Center for AI Safety, MIT’s Media Lab. Researchers pitted today’s top AI models against professional virologists in wet lab scenarios. AI can now guide people through ultra-specialized lab procedures…
AI > Human Experts in Virology Tasks:
-
The test evaluated AI’s ability to solve 322 practical, hands-on issues that aren’t easily searchable or documented.
-
AI models like OpenAI’s o3 and Google’s Gemini 2.5 Pro outperformed PhD-level virologists in troubleshooting complex wet lab scenarios.
-
Virologists scored 22.1%, while o3 hit 43.8% and Gemini reached 37.6%.
-
Experts warn this progress means anyone, not just trained scientists, could weaponize viruses using AI guidance.
=> In the past, lack of expertise was a barrier for creating bioweapons. That wall is now crumbling.
-
Increased access = increased risk: If models are widely available without restrictions, even bad actors or amateurs can exploit them.
-
Only verified researchers (e.g., MIT scientists) should access unrestricted versions of these AIs.
-
Third-party audits and evaluations of AI systems before public release.
-
Government regulation, not just voluntary self-regulation, especially around pandemic-level risks.
Why it matters: I believe AI just made decades of biotech expertise Googleable. But here’s the chilling twist: the same AI can also walk someone – anyone – through how to make a bioweapon, AI is now “too helpful”?
PRESENTED BY SIMPLY BUSINESS
Is your small business risk-ready?
A client has an allergic reaction to some face cream you applied. A customer trips over a ladder you left on the lawn. You neglect to carry the “2” on a client’s tax return and they have to pay a penalty. Just about every small business faces financial risks, but more than half of them (56%) aren’t protected with insurance.
Without insurance, you could find yourself having to cover costly claims. Even if it’s not your fault, you may still have to pay the legal expenses to defend yourself.
Getting covered doesn’t have to be confusing, time-consuming, or expensive. Simply Business can help find affordable coverages many small businesses need in just minutes. All from leading insurers and all online, 24/7. They’ve done that for more than 1 million businesses worldwide.
Simply Business offers same-day coverage for most policies and their licensed agents can answer your questions on the phone.
Readers, when it’s time to insure your business, be sure with Simply Business.
TODAY IN AI
AI HIGHLIGHTS
🤯 OpenAI wants to buy Chrome and make it an “AI-first” experience, no need to build its own Chromium-based browser. If OpenAI + Chrome + ChatGPT happens, the feds would be watching every keystroke.
🔐 Anthropic warns fully AI employees could soon roam corporate networks, raising big security risks. Who’s responsible if they go rogue? Greater autonomy comes greater risk.
📹 Google Meet lets you have a meeting from the White House, if you want. Just generate your custom AI backgrounds with a simple prompt, see real example here.
👀 xAI’s Grok can now see what you see in real world through your smartphone’s camera. Just point it at objects or signs and ask questions – like ChatGPT Vision, but from Elon Musk’s team.
🎶 AI just brings a dead composer’s brain back to life and creates new “music” in same style. This song even hit 64,000 listens on Spotify. So is it “revivification”?
🔧 Microsoft launched the AI Dev Gallery – a local playground for building AI apps without the cloud, offering 25+ AI samples and supporting GPUs.
💰 AI Daily Fundraising: Goodfire raised $50M from Anthropic and others to pioneer “neural brain surgery” for AI – reprogramming how models think, not just what they see, to build safer, smarter systems.
AI SOURCES FROM AI FIRE
NEW EMPOWERED AI TOOLS
-
🎨 Seedream 3.0 is the next-gen text-to-image model by Bytedance.
-
🍓 Strawberry researches across hundreds of sites, summarize, take notes.
-
🤖 Agent Simulate tests your AI apps with thousands of digital humans.
-
🎤 All Voice Lab is an AI voice platform for ultra-realistic TTS & voice cloning.
-
🌐 Nonilion navigates between different meeting areas with spatial interfaces.
AI QUICK HITS
-
☠️ Over 100 software companies are getting dangerously AI “squeezed”.
-
🔑 ChatGPT chief calls search crucial for OpenAI in Google trial.
-
👂 Pedestrians hear AI messages from billionaires at hacked crosswalks.
-
💰 Google’s paying Samsung “Enormous Sums” to make Gemini the default.
-
🎥 Films made with voice-cloning AI can win top awards at the Oscars.
AI CHART
😐 “Roast Me”, Bot: How Sarcastic AI Hacks the Leaderboard
😉 Top Bots Don’t Just Think, They Use Sentiment to Flirt
Do you know that: Sentiment might be the secretsauce behind top chatbot rankings? Emotionally expressive models seem to do better in human preference evaluations, even if their substance isn’t significantly better.
Classification done using Gemini-2.0 Flash with a strict tone-focused schema (ignoring content, focusing purely on delivery/emotion).
Coefficients from the regression showed:
-
Answer length and positive sentiment are positively correlated with better model rankings.
-
Negative, neutral, and emoji-heavy responses generally scored lower.
-
Very positive tone = second-highest boost after length.
Head-to-Head Win Rates Expose Surprising Dynamics:
-
Very Negative tone beats Neutral and Negative — possibly due to humor or deliberate “roasting” scenarios.
-
Neutral loses almost universally → suggests humans find flat, emotionless responses unengaging, especially in creative settings.
-
Positive and Very Positive tones dominate across comparisons.
So, here I have a question for you: If chatbots can manipulate sentiment to “score points”, then Should we separate “content” from “emotion” in evaluations? Or should we accept emotion as a legitimate part of the user experience?
AI JOBS
We read your emails, comments, and poll replies daily
Hit reply and say Hello – we’d love to hear from you!
Like what you’re reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Leave a Reply