🖼️ ChatGPT Adds "Smell Effect" to Images

Free AI Tutorials | AI Academy | Advertise | AI Agent No-code n8n

Plus: AI can "Language-ify" & Talk to Living Cells

Read time: 5 minutes

From AI that sniffs out billions of scam ads before they reach your eyes, to models that can "language-ify" living cells like they're chatting back — and now, images that can literally trigger your sense of smell... What happened?

What are on FIRE 🔥

• 🤯 Google ‘Cell-to-Sentence’ Model Can ‘Talk’ to Human Cells

• 🌟 AI Highlights

• 📚 AI Sources From AI Fire

• 🏅 AI Tools

• ⚡ 5 AI Quick Hits

• 📊 AI Chart

• 💼 4 AI Jobs

IN PARTNERSHIP WITH CLOAKED

See what Info about you is Exposed and Take Back Control with Cloaked

Tech companies & data brokers profit off of your personal information. This makes you vulnerable to spam, fraud, & phishing attempts. Cloaked searches your phone number across 120+ data brokers to find & remove your personal information. Get a free scan now and take back control of your privacy!

AI INSIGHTS

💬 Google ‘Cell-to-Sentence’ Model Can ‘Talk’ to Human Cells

What if you could ask a single cell how it’s feeling? What if you hand it a question like “Hey buddy, how would you react if I hit you with this cancer drug?”, and the cell could actually answer?

No, this isn’t sci-fi. This is the future Google is building with something called C2S-Scale model (short for Cell2Sentence-Scale) that lets researchers ask cells specific questions, and receive answers in plain English.

First, the problem here is a single cell spits out thousands of numbers representing which genes are turned on and off. Numbers, numbers, numbers. Now, C2S-Scale could “language-ify” all that data:

→ It turns gene expression data into plain English “cell sentences” — literally arranging gene names in order of activity, like a priority list.

• Conversational biology: You can ask the model natural language questions about cell behavior and it answers.

• Data interpretation: It can generate summaries of biological data, suggest cell types, and highlight potential experimental insights.

• Predictive modeling: Given a cell’s current state and a treatment, the model can predict future gene expression — like a virtual simulation of the cell's reaction.

• Scaling laws: Like other LLMs, larger models show clear and predictable performance improvements in both predictions and generation tasks.

You can now “talk” to your biological data. It’s biology made chat-friendly. It’s proof that “language as the interface” is the future.

Why it matters: Do you know what’s the bigger play here? Google and Yale are betting on something much larger: If you can teach machines the language of life, you can teach them to help cure, design, and predict life too. And honestly? I think they’re right.

PRESENTED BY GENEVA TOURISM

The Science City You Didn’t Know You Needed to Visit

Geneva, Switzerland is famous for luxury watches and international summits, but it’s also a hub of scientific innovation. At CERN, you’ll find the world’s largest particle accelerator and the cutting-edge CERN Science Gateway.

Designed by renowned architect Renzo Piano, the Science Gateway offers hands-on exhibits, immersive VR, and live demos that bring the mysteries of the universe to life. Whether exploring particle physics or experiencing the Big Bang, it’s an unforgettable journey for visitors of all ages.

Next time you're in Geneva, go beyond the scenic lake and charming Old Town—dive into the future at CERN’s Science Gateway.

TODAY IN AI

AI HIGHLIGHTS

💡 Claude’s best practices for “agentic coding”, there’s a term called “ultrathink” - a cheat code to get Claude to spend its 32K context window thinking before answering.

🖼️ ChatGPT is sparking a web storm now by creating images with a "smell effect", activating a real sensation of smell.

⚡ Grok 3 Mini offers 4 variants to choose between different levels of speed and reasoning capacity. It beats some of the world’s best reasoning models on graduate-level benchmarks while being up to 20x cheaper.

⚖️ Google is in hot water as a DOJ trial begins, aiming to break up Google’s search monopoly. This could open up space for new, more competitive AI products, particularly generative AI tools.

🤖 Anthropic just analyzed 700,000 Claude conversations and published latest paper on Model Alignment and found its AI has a moral code of its own.

🔍 ChatGPT is booming in Europe with 37.5 million searches daily, thanks to AI's rise and OpenAI's partnership with Microsoft. It's also making $1 billion next year!

💰 AI Daily Fundraising: 01VC has secured a $69 million fund to invest in AI and robotics. This supports global growth, emphasizing AI's rising importance in China's tech scene and international expansion efforts.

AI SOURCES FROM AI FIRE

NEW EMPOWERED AI TOOLS

1. 🎬 Wan2.1-FLF2V - Alibaba’s open-source first and last frame video model.

2. 📸 Google Whisk 2.0 turns a still image into an 8-second animated clip.

3. 👀 EyesOff is an app that alerts you when someone’s peeking at your screen.

4. ⚡ RapidMCP convert your REST API into MCP Servers in minutes!

5. 📚 Bookaroozie is an AI Kindle alternative that works with PDFs, EPUBs.

AI QUICK HITS

1. 🧠 GPT-o3 and o4-mini hallucinate significantly more than older models.

2. 🌐 OpenAI merges memory into search feature while scanning the web.

3. 👨‍💻 MIT’s ‘Sequential Monte Carlo’ approach lets small AI models outperform larger ones.

4. ⚠️ AWS faces backlash over “arbitrary” limits on Anthropic's AI.

5. 🔻 Google AI Overviews are extremely hurting click-through rates.

AI CHART

🚨 9.1 Billions of Bad Google Ads: AI is Cleaning Up the Internet

Ever wonder how many scammy, shady, or straight-up dangerous ads try to sneak into your Google search results and YouTube videos every year? Spoiler: a lot. But Google’s latest Ads Safety Report says AI is doing some serious heavy lifting in the fight against bad ads.

What’s going on? In 2024, Google used AI to detect and remove over 5.1 billion rule-breaking ads. Yeah, billion. With a "B".

What kinds of ads are we talking about?

• Ads that abused Google's ad system.

• Ads hijacking trademarked content.

• Personalization gone wrong (or downright creepy).

And the mission behind it is pretty clear: Less junk. More trust. One of the biggest growing problems was AI deepfakes being used to impersonate celebrities or brands. Google doesn’t rely on one AI model, it uses an army of them to scan and detect.

So Google “restricted” more than 9 billion ads under some sensitive areas. And it’s not just advertisers - publishers, too. Google also took action on 1.3 billion publisher pages. Basically, if it’s the dark side of the web.

AI JOBS

• Google: Customer Engineer, Cloud AI, Google Cloud

• Upwork: Video Creator -- AI Talking Head (with aesthetic and captivating edits/cut scenes)

• FurtherAI: Software/AI Engineer (New Grad)

• Walmart: Principal, UX Designer - AI Assistants

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