Zack
@DocZedd
Helping improve healthcare through the use of AI augmentation | Helping people understand how AI works | AI Researcher | Tweets are my own.
You might like
A big part future belongs to mid-career professionals who can identify the 20% of repetitive work consuming 80% of their time and build AI systems around it. Not to those waiting for perfect automation or paralyzed by job loss fears. AI doesn't need to replace your entire job.…
What are the biggest AI misconceptions in your field? They're not about the technology. They're about what role humans will play.
We've been obsessing over making AI think like humans. That's the wrong goal. The breakthrough isn't replication. it's complementarity. Stop building AI assistants that mimic human reasoning and start building tools that do what humans can't: process millions of data points…
Most problems in AI won’t be solved by more models — they’ll be solved by better thinking. Crisis care isn’t a dataset problem; it’s a human one. You can’t shortcut empathy with fine-tuning. Systems like this won’t improve through theory alone; they require real-world feedback,…
This paper brings out a drawback of medical LLMs. A general chatbot withdraws when people share more signs of suicide risk, which breaks basic crisis care. Only 14% of replies invited the person to keep talking, and this fell to 0.8% by the last turn in one tested sequence.…
Excited to release new repo: nanochat! (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single,…
You're not struggling with productivity. You're struggling with not having a project worth being productive for. If you were working on the right thing, you wouldn't be able to stop thinking, working, obsessing.
Chatbots are the lazy default in AI. Nobody wants to explain what they want when they could just do it. Uber didn’t build a chatbox, it built buttons powered by invisible ML. That’s the future: purpose-built models + intuitive UX that feel like magic, not work.
What is your personal tolerance for AI doctors? Would you trust a solely AI-driven medical interaction in today's environment?
Nobody wants to admit that: Hallucinations aren't a bug in AI - they're a mathematical certainty. OpenAI just proved that the way we train models makes perfect accuracy impossible, not just difficult.
5 Little-Known Deployment Realities That Could Be Undermining Your AI Project 1. Automating broken processes: AI amplifies what exists; broken workflows become expensive broken workflows at scale. 2. Missing human fallback mechanisms: Current AI limitations mean complex tasks…
"Evaluations are a scam"—until you need to pass a quality review board in law or healthcare. Then suddenly, you're scrambling to prove your system works. Here's what actually makes evals work: 1. Define "good" upfront. - Spend time deciding what quality looks like for YOUR…
United States Trends
- 1. Good Friday 44.5K posts
- 2. #FridayVibes 3,669 posts
- 3. #NXXT_Earnings N/A
- 4. Summer Walker 20.5K posts
- 5. #FanCashDropPromotion N/A
- 6. Wale 34.5K posts
- 7. #FinallyOverIt 6,640 posts
- 8. Happy Friyay 1,106 posts
- 9. #FridayMotivation 2,291 posts
- 10. Go Girl 25.3K posts
- 11. RED Friday 2,299 posts
- 12. Meek 6,101 posts
- 13. Bubba 8,333 posts
- 14. Robbed You 4,516 posts
- 15. Saylor 27.3K posts
- 16. 1-800 Heartbreak 1,771 posts
- 17. Monaleo 2,245 posts
- 18. Arrest Bill Clinton 6,393 posts
- 19. 5sos 23.6K posts
- 20. Bihar 338K posts
You might like
Something went wrong.
Something went wrong.