Hoy, kumusta? Welcome back to Kaisipan!
From curing diseases to your banking app, AI is already here and the Philippines is just catching up.
Here’s what happened in AI this week:
AlphaFold's spin-off just raised the bar again.
PIDS: AI is already inside your e-wallet.
Advocates warn AI is targeting Filipino children online.
DOST will launch the Philippines' first national AI hub.
Google, Anthropic, and OpenAI all dropped major updates.
How AI Solved the Problem That Stumped Scientists for 60 Years
Let's talk about the good side of AI for a change. Hindi ito bagong balita, but it's a story worth revisiting, especially in its direct impact on health and medicine.
Proteins are the biological machines behind everything your body does. Their function depends entirely on their 3D shape. And figuring out that shape? For decades, scientists could only do it by crystallizing a protein and blasting it with X-rays. Isolating a protein requires years of work, sometimes an entire PhD. Over six decades, the entire global scientific community mapped about 150,000 proteins combined.
DeepMind cracked it with AlphaFold. Their AI learned from evolutionary patterns across species: which amino acids tend to stay bonded as proteins evolve, which mutations tend to travel in pairs, and what those clues reveal about a protein's final structure.
The proof came at CASP, the global protein-folding competition running since 1994. For decades, no model cleared the gold standard score of 90. AlphaFold 2 didn't just pass it, but its predictions are nearly indistinguishable from lab experiments. The scientific community was stunned. CASP founder John Moult called it a solution and that wasn't an overstatement.
Created using ChatGPT 5.2
Then DeepMind did something remarkable: they released it all. Every predicted structure is freely available to researchers worldwide. Sa loob ng ilang buwan, nag-upload sila ng mahigit 200 million protein structures, which is essentially every protein known to exist in nature. Research labs across the globe suddenly had decades' worth of data they never had before.
In 2024, AlphaFold's creators won the Nobel Prize in Chemistry. Hindi lang sila na-recognize — they changed what science is capable of.
A July 2025 research review by Fang et al. documented just how far AlphaFold 3 has gone since then:
Drug design: AF3's full-atom docking accuracy hit 76.4% (1.8 times better than the next best tool), and it successfully mapped over 14,000 drug-protein interactions previously too complex to model
Cancer research: Scientists used AF3 to identify new liver cancer targets and design nanoantibodies that successfully degrade tumor proteins and activate cancer-suppressing genes
Neurological diseases: AF3 modeled over 1,200 brain-related proteins and flagged multiple potential targets linked to Alzheimer's disease
Vaccine development: Researchers used AF3 to design and validate a peptide-based vaccine against the respiratory syncytial virus, verifying its structural stability before any lab experiment
Earlier this month, DeepMind's drug-discovery spin-off Isomorphic Labs unveiled a proprietary model scientists are already calling an “AlphaFold 4-level advance.” Researchers say it outperforms everything currently available at predicting drug-protein interactions but the company is keeping the technical details private, which is raising eyebrows across the research community.
AlphaFold is a green flag of AI. It advanced protein research a lot. And while it doesn’t have an immediate impact on our lives, nakatingin ako sa long-term. Siyempre, nasa research phase pa ang mga scientists natin. In the future, as AI becomes better and even smarter, AlphaFold will make a significant contribution to solving biology’s hardest problems and, eventually, makikita natin ito in the form of drugs and even cures for diseases that today’s medicine can’t cure.
Bes, Ano Yung…DeepMind?
DeepMind is Google's AI research lab, founded in London in 2010 by Demis Hassabis, a neuroscientist, chess prodigy, and 2024 Nobel Prize winner. Their mission is to use AI to solve the hardest problems in human history. They don't build apps or consumer products. Instead, they go after foundational science like defeating world champions in Go, cracking the protein folding problem, and solving unsolved math challenges. The results don't always appear on your phone, but they almost always reshape an entire scientific field. Isa sila sa pinaka-serious na AI research organizations sa mundo ngayon, and they're just getting started.
PIDS Study: Filipinos Are Interacting With AI in Banking Without Knowing It
Think about this: the next time you transfer money using GCash, check your bank balance online, or apply for a quick loan through an app, chances are AI is part of the process. Isn't that interesting?
A discussion paper from the Philippine Institute for Development Studies (PIDS) confirms what many of us haven't noticed. AI is already embedded in Philippine digital finance. Major banks and e-wallet providers are quietly running AI for fraud detection, credit scoring, customer service chatbots, and transaction monitoring.
The surprising part? Most Filipino users don't recognize these as AI-powered services. The study found that people with higher digital engagement are the most likely to be interacting with AI, whether they know it or not. Kung gumagamit ka ng online banking o e-wallet, naka-expose ka na.
But here's where it gets complicated. The Philippines scored 0.50 on the IMF's AI Preparedness Index, a middling score that masks a deeper problem. Mataas ang interes natin sa digital tech. Mahina pa ang pundasyon. The study calls this a “curiosity-capacity gap.” We're searching and engaging, but the infrastructure and institutional readiness aren't keeping up.
AI can genuinely help close financial gaps by improving credit access for the unbanked, enhancing fraud protection, and delivering more personalized services. Pero may kasamang panganib. Without proper data protection and consumer safeguards, the same systems could deepen inequality rather than reduce it — especially for Filipinos outside Metro Manila with weaker internet access.
The PIDS recommends to the National Financial Inclusion Strategy, in conjunction with the National AI Strategy Roadmap, to strengthen data protection laws and to begin measuring AI literacy in national surveys. Kasi hindi mo ma-improve ang hindi mo sinusukat.
Mga Ibang Ganap
Child rights advocates warned lawmakers and tech companies that AI-generated deepfakes are increasingly weaponized to abuse, extort, and exploit Filipino children online, ahead of this year's Safer Internet Day.
DOST will launch NAICRI on February 26, the Philippines' first centralized AI research hub consolidating computing infrastructure, government AI tools, and a Filipino-language search interface called iTANONG.
Google released Gemini 3.1 Pro across consumer and developer platforms, doubling reasoning performance over its predecessor with a 77.1% score on ARC-AGI-2, a benchmark for solving entirely new logic patterns.
Anthropic launched Claude Sonnet 4.6 as the new default model on Claude and Claude Cowork, bringing Opus-level performance to Free and Pro users at the same price point.
Sam Altman said OpenAI is on track to deploy an intern-level AI research assistant by September 2026 and a fully autonomous AI researcher by 2028, capable of independently leading large research projects.
Palaisipan
Is this AI or not?
Prompt Tip!
Struggling with the generated results of AI? There might be a fix for that.
Researchers at Google found that simply repeating your query back-to-back (no separator in between) wins on 47 out of 70 benchmark tests, with zero losses. Here's why it works: AI models like ChatGPT, Claude, and Gemini read your prompt left-to-right. Early words can't "see" what comes later. Repeating the prompt gives the model a second pass at your full request before it generates any response, like rereading a question before answering it.
How to do it: Paste your prompt. Then paste it again right after it gives a result. That's it.
Best for long, context-heavy prompts where details matter. One caveat, though, is that if you're using extended thinking or reasoning mode, the gains are smaller since the model already re-reads your prompt on its own.
That’s all for today!
PALAISIPAN Answer: The image is AI
Thank you for reading today’s newsletter!
Do you have suggestions? I’d love to hear your thoughts.