💸 Is AI the Answer to Poverty?

In today’s email:

💸 Can AI help fight poverty? Discover how Togo’s innovative approach used AI to deliver cash to those in need and how it’s changing the game for poverty reduction.

🌐 DeepSeek is making waves in the AI world with a model that’s shaking up the industry—and costing way less than major players like GPT-4. Will this disrupt the market?

📖 Confused by AI jargon? We’ve got your back with a handy glossary that breaks down the key terms.

💸 Can AI help fight poverty?

In 2020, Togo decided to find out if AI can reduce poverty. During the pandemic, the West African country used AI to identify people in need and sent them $10 every two weeks via mobile payments. It might not sound like much, but it helped keep people from starving. Traditional poverty surveys are slow and expensive, so Togo’s Novissi program used satellite images and mobile data to get cash to the right people—fast.

This AI-driven approach has some serious upsides:

• Speed: No need for months of surveys. AI does it in a flash.

• Scale: It covers way more people than traditional methods.

• Accuracy (sort of): It finds patterns humans might miss.

But, of course, AI isn’t perfect. It can be biased and might overlook people who don’t leave a digital footprint. Still, old poverty-measuring methods aren’t great either, so why not experiment?

Meanwhile, researchers are teaching AI to spot poverty from space, analyzing satellite images to find clues like road quality and building sizes. The results? AI predictions match traditional surveys but with way less effort. While AI won’t single-handedly erase poverty, it might make fighting it much smarter.

🌐 Has DeepSeek Shaken Up the AI World?

DeepSeek, a Chinese AI startup, dropped its R1 model in January, sending shockwaves through the tech world. They claim to have built it for a fraction of the cost of major AI models like OpenAI’s GPT-4 and Google’s Gemini Ultra.

What Makes DeepSeek-R1 a Big Deal?

• Ridiculously Cheap Training Costs – Built for under $6M, compared to the $78M+ OpenAI reportedly spent on GPT-4.

• Fewer GPUs, More Efficiency – Used 2,000 Nvidia H800s, whereas Meta needed 16,000 H100s for LLaMA 3.

• Smart Architecture – Their “Mixture of Experts” (MoE) design activates only the necessary parameters, saving power and cost.

The Aftermath

The market freaked out—Nvidia’s stock tanked 17% in a single day, wiping out $600B in value. Meanwhile, DeepSeek’s app topped download charts but got banned in places like South Korea, Italy, and Australia over security concerns.

What It Means for the Future

Tech leaders are now reevaluating AI costs and strategy. Bain & Co suggests companies should:

1. Prepare for AI cost drops – AI is getting cheaper, fast.

2. Watch market shifts – Understand spending patterns.

3. Think beyond automation – AI isn’t just a cost-cutter; it’s a business game-changer.

So, is DeepSeek the future of AI or just a market disruptor?

📖 AI Glossary Cheat Sheet

AI can feel like a chaotic soup of jargon, and the folks at TechCrunch decided to throw us a lifeline—a handy glossary breaking down key AI terms. Here are the highlights:

• AI Agent: Think of this as your personal AI butler—it books your flights, files expenses, and maybe even writes code (but don’t expect it to do your laundry just yet).

• Chain of Thought: AI’s way of solving complex problems in steps—like how you’d break down a math problem rather than panic and guess.

• Deep Learning: AI that mimics the brain but without the existential crises. It’s what powers smart models, but it needs tons of data and computing power.

• Fine-Tuning: Taking a general AI model and training it further so it’s really good at something specific—like making a chatbot that actually understands your jokes.

• Large Language Models (LLMs): The brains behind ChatGPT, Google Gemini, and more. These massive networks predict text based on patterns from tons of training data.

• Neural Networks: The reason AI can do cool stuff—modeled after human brains but runs on GPUs (thanks, gaming industry!).

• Weights: Not the gym kind—these numbers help AI models decide what’s important, like how many bathrooms impact a house’s price.

Other cool AI stuff that is trending right now 🔥🔥

🤖Think you're good at spotting optical illusions? Turns out, AI sucks at it—so now these mind-bending images might become the next CAPTCHA test to keep bots out and humans in. - Read more

📚OpenAI dropped $50 million to digitize academic texts from top universities—because who needs copyright lawsuits when you can have a digital library revolution instead? - Read more

🕹️ AI is now speedrunning Super Mario Bros., but instead of saving the princess, it’s struggling with basic jumps—because thinking too hard might actually make it worse. - Read more

👨🏽‍⚕️ Get ready for "Dragon Copilot," Microsoft's new AI assistant that promises to tackle hospital bureaucracy and give doctors more time to focus on patients—coming to Germany this summer. - Read more

🤝vMIT and GlobalFoundries are teaming up to create super-efficient chips that will power everything from AI data centers to your smart devices. - Read more