TimByte

7 ms·

151.
by TimByte 11mo ago
Tha's the most responsible form of sabotage I've ever heard
152.
by TimByte 11mo ago
That drive's probably earned a quiet retirement at this point
153.
by TimByte 11mo ago
So hard drives are getting better but just as importantly, we're also getting smarter about how we use and maintain them
154.
by TimByte 1y ago
You're right: even when one proposal gets stopped, it rarely kills the idea
155.
by TimByte 1y ago
Would be nice if this actually marked a turning point
156.
by TimByte 1y ago
This is a rare bit of good news on the privacy front
157.
by TimByte 1y ago
It takes real confidence (and restraint) to let strong engineers run with their ideas without feeling the need to course-correct every detail
158.
by TimByte 1y ago
There's a difference between defaulting to authority and resorting to it when needed. Consensus is great until it turns into paralysis
159.
by TimByte 1y ago
The shift from "expert with answers" to "facilitator of clarity" is something a lot of leads struggle with, especially when they get promoted because they were the smartest dev on the team
160.
by TimByte 1y ago
I wonder how much of that actually survives token filtering during training
161.
by TimByte 1y ago
One possibility is that alignment isn’t just about cultural distance in values, but also about textual representation in the training data
162.
by TimByte 1y ago
You're right to be cautious, especially given how many pop-psych books have aged like milk after the replication crisis came into full view. Henrich’s work feels a bit different to me. He’s not just stringing together catchy one-off st
163.
by TimByte 1y ago
I get the general point about institutional divergence, but thousand-year historical causal chains always make me a bit cautious
164.
by TimByte 1y ago
Sooo it's like training a robot to "think like a human," but all the reference humans are Silicon Valley product managers and undergrads from elite universities
165.
by TimByte 1y ago
But I think the key difference is the potential for harm and power imbalance involved
166.
by TimByte 1y ago
More rigor is always good, but there’s also value in studying messy real-world behavior as it happens
167.
by TimByte 1y ago
There's definitely a tradeoff in terms of experimental control, but the real-world insight seems worth it.
168.
by TimByte 1y ago
Silicon Valley's unofficial motto for the last 15 years
169.
by TimByte 1y ago
It's naive to think Chinese models have a free pass. Local censorship, language/data biases, and export restrictions cut both ways.
170.
by TimByte 1y ago
What's wild is that $1.5B sounds huge… until you compare it to the potential upside of owning the dominant AI model trained on everything
171.
by TimByte 1y ago
You can't grab pirated stuff and then hope fair use magically sanitizes it
172.
by TimByte 1y ago
It doesn't set precedent, but the message to other AI companies is clear: if you're going to bet your model on gray-area data, have a few billion handy for settlements
173.
by TimByte 1y ago
I'd love to see actual data on most-played maps across all FPS games
174.
by TimByte 1y ago
People just made stuff for the fun of it
175.
by TimByte 1y ago
The fact that people still remember those servers all these years later? That's legacy-level stuff
176.
by TimByte 1y ago
Totally agree on the point about toxicity. Back then, when you kept running into the same folks, there was some accountability
177.
by TimByte 1y ago
Wild to think one of the most influential shooters of all time started as a dorm room mod. And love that Valve didn't crush the mod but embraced it
178.
by TimByte 1y ago
Informal spaces let you skip the guardrails that academia imposes
179.
by TimByte 1y ago
It's not just a drive-by hit piece
180.
by TimByte 1y ago
I think the piece bends over backwards to keep the charitable frame because it's written by someone inside the community, but you're right that the touted "wins" feel a bit thin compared to the sheer scale of dysfunction

More ›