6. He learns through feedback. Not lectures. Agents, like toddlers, learn through guided experiences, not commands.
Raising a toddler is the most chaotic, real-time RL experiment I’ve ever lived through. Turns out: the best agent architectures are built w/ feedback and patience.
I’ve been building AI agents for months.
But it turns out… my 2-year-old is the best teacher I’ve had.
Here’s a few lessons from the world of AI, agents, and raising a toddler. 🧠🍼👇
(a thread)
5. The world model matters.
If I don’t update his understanding (like why we dont drink bath water), he operates on faulty assumptions. Same with agents: Bad context = Bad outputs.
@thesamparr Choosing the right partner is arguably the most important decision you’ll ever make—it shapes not just your family, but your life trajectory. Timing matters, but who you’re doing life with matters even more.
@thesamparr Totally agree that having kids sooner can be a huge blessing, but the timing isn’t just about careers or finances—it’s about finding the right partner. I feel lucky to have found mine, but I can’t imagine anything worse than having kids too early with the wrong person.
Excited to be headin’ to #PyDataNYC! On Nov 7, I’ll be diggin’ into ‘Why Your Machine Learning Model’s Probabilities Are Lying to You.’ It’s gonna be a ride, folks, so grab your ticket, and let’s dive deep. See y’all there!
@PyDataGlobal@NumFOCUS
New Post: What's In A Baby Name?🍼👶
Written for soon-to-be parents by soon-to-be parents (@sara_makenzie92 & myself).
https://t.co/IxZOxHBGBe
#Python#100DaysOfCode
@ramit My wife and I joined a boat club after she was diagnosed with cancer this past June. Firm believer that you can’t buy happiness but a boat can sure help numb the pain.