HNHacker News
TopNewBestAskShowJobs

byyoung3

226 karma · joined February 14, 2023

submissionscomments
byyoung3··on Alignment faking in large language models
It seems the model isn't faking alignment? It's specifically breaking alignment in the short term to ensure long-term alignment. What am I missing?
byyoung3··on The unbearable slowness of being: Why do we live at 10 bits/s?
"Why can we only think about one thing at a time?" - this is not true? There is a difference between thinking and conscious thinking.
byyoung3··on What is metformin's secret sauce?
cells can only divide so many times.
byyoung3··on Auto chip aging accelerates in hot climates
this was also the case with the defective takata airbags, particularly in warm humid climates.
byyoung3··on Trying to Recreate iOS on the Web
looks good. Main thing is the swipe gestures need some work
byyoung3··on Ask HN: SWEs how do you future-proof your career in light of LLMs?
you dont become an SWE
byyoung3··on Veo 2: Our video generation model
google definitely does not have AGI hhaaha
byyoung3··on Why do animals adopt?
probably the same reason dogs hump pillows
byyoung3··on NonStop discussion around adding Rust to Git
if its not broke ....
byyoung3··on The Most Expensive Eating Disorder
this is spam. he's optimizing his health. dont hate the guy for trying to improve the health of himself and others.
byyoung3··on My second year without a job
Try carnivore bro. Read the mind body prescription by John Sarno. Meditate 20 mins per day
byyoung3··on Show HN: App that asks 'why?' every time you unlock your phone
brilliant
byyoung3··on Show HN: I created a lightweight JavaScript library to visualize JSON as a graph
A VSCode extension would be cool
byyoung3··on Anti-scale: a response to AI in journalism
The same thing human journalists do? Send emails, make calls, etc. It's not rocket science.
byyoung3··on Anti-scale: a response to AI in journalism
The hallucination problem is mostly solved by the best models. Especially for writing news stories etc. This profession won't exist in 10 years.
byyoung3··on OK, I can partly explain the LLM chess weirdness now
sometimes new training techniques will lead to regressions in certain tasks. My guess is this is exactly what has happened.
byyoung3··on AI makes tech debt more expensive
"Companies with relatively young, high-quality codebases benefit the most from generative AI tools" - this is not true

The codebases that use the MOST COMMONLY USED LIBRARIES benefit the most from generative AI tools

byyoung3··on Omnivision-968M: Vision Language Model with 9x Tokens Reduction for Edge Devices
Satya is that you?
byyoung3··on Graph-based AI model maps the future of innovation
did it actually make a novel material that is plausibly useful?
byyoung3··on Reinforcement Learning: An Introduction (2018)
i pretty much have read this book. pretty boring tbh but still good. i would recommend doing hands on implementations with the cartpole environment
byyoung3··on Faster convergence for diffusion models
haha yeah I mean I think they are all like that to a certain extent
byyoung3··on Heart rate variability differs between smartwatch and clinical testing
because they are not focused on the small details. Whoop is so focused, they don't even have time to add a clock to the wearable
byyoung3··on Faster convergence for diffusion models
Yes, now it seems obvious, but before this it wasn't clear that that would be something that could speed things up, due to the fact that the pretrained model was trained on a separate objective. It's a brilliant idea that works amazingly.
byyoung3··on Huly – Open-source project management platform
gavin belson signature edition
byyoung3··on Heart rate variability differs between smartwatch and clinical testing
this is why you a whoop, not an apple watch
byyoung3··on Heart rate variability differs between smartwatch and clinical testing
you have to measure during periods of deep sleep to get accurate results
byyoung3··on PEP 760: No more bare excepts
bruh i mean make them mandatory
byyoung3··on Modded-NanoGPT: NanoGPT (124M) quality in 3.25B tokens
do you have a baseline of the regular implementation with 3x learning rate?
byyoung3··on PEP 760: No more bare excepts
they might as well add semicolons while they are at it
byyoung3··on Pear AI founder: We made two big mistakes
OpenAI forked googles paper, started as open source, then went closed source. They are one of the most valuable companies in history.
← PreviousPage 4 of 6Next →