Top all-time · #7735
Clement Neo
Personal blog · clementneo.comaiprogramming
Clement Neo's personal blog at clementneo.com covers ai and programming. Since 2023, 10 different Hacker News readers have linked it 12 times, most often “We Found an Neuron in GPT-2”. The last post was in 2023, but the archive is still widely read.
Too few sampled comments (11) to say how often HN recommends it.?
Why HN links it
HN readers link Clement Neo's 2023 post "We Found an Neuron in GPT-2", which examines a single GPT-2 neuron. Commenters use it as evidence that language model internals can be studied.
Summarised by Claude from a sample of 11 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Blog
Most-linked posts?
- We Found an Neuron in GPT-210 people · 12 mentions · first linked 2023-02-16
What Hacker News says?
This post is interesting: https://clementneo.com/posts/2023/02/11/we-found-an-neuron
Comment linking /posts/2023/02/11/we-found-an-neuron
on “Better and Faster Large Language Models via Multi-Token Prediction”Submitted We Found an Neuron in GPT-2 as a story · 3 points
HN discussionI'm pretty sure any system built via linear regression or similar is an opaque box even to the most experienced researchers. For example: https://clementneo.com/posts/2023/02/11/we-found-an-neuron These are massive functions with billions of parameters that evolved over millions of computing years.
Comment linking /posts/2023/02/11/we-found-an-neuron
on “Introducing ChatGPT and Whisper APIs”"Its", not "It's". A helpful tip: whenever you need to decide whether to use one or the other, ask yourself whether you want to say "it is"; if so, use "it's"; otherwise, use "its". Interestingly, ChatGPT knows whether to use "it's" or "its" at what seems like a superhuman level, but we don't even really know how it knows whether to use "a" vs. "an": https://clementneo.com/posts/2023/02/11/we-found-an-neuron
Comment linking /posts/2023/02/11/we-found-an-neuron
on “The False Promise of Chomskyism”I've been working in systems neuroscience for a few years (something of a combination lab tech/student, so full disclosure, not an actual expert). Based on my experience with model organisms (flies & rats, primarily), it is actually pretty amazing how analogous the techniques and goals used in this sort of research are to those we use in systems neuroscience. At a very basic level, the primary task of correlating ne…
Comment linking /posts/2023/02/11/we-found-an-neuron
on “Language models can explain neurons in language models”he's likely talking bout the internals. sure we know how to train them but nobody knows what the models learn exactly. how those billions of parameters shape the output on inference. a few months ago, just this year some researchers discovered what might be the neuron that largely decides when to use an in gpt-2. yes 2. that's what he means. https://clementneo.com/posts/2023/02/11/we-found-an-neuron
Comment linking /posts/2023/02/11/we-found-an-neuron
on “Pausing AI Developments Isn't Enough. We Need to Shut It All Down”Prompt engineering is only getting better. Also, > You can't search these weights with command-f Sometimes you can, https://clementneo.com/posts/2023/02/11/we-found-an-neuron
Comment linking /posts/2023/02/11/we-found-an-neuron
on “Show HN: IngestAI – NoCode ChatGPT-bot creator from your knowledge base in Slack”Don’t let the mathematical terms fool you, these are just fan theories. For a real investigation you need debug access. For a good example: https://clementneo.com/posts/2023/02/11/we-found-an-neuron For image recognition, machine learning researchers eventually figured out the neural networks are paying attention mostly to textures. Hopefully we will have a better understanding of what language models are doing some…
Comment linking /posts/2023/02/11/we-found-an-neuron
on “The Waluigi Effect: an explanation of bizarre semiotic effects in LLMs”What about a system emerging from another system? https://clementneo.com/posts/2023/02/11/we-found-an-neuron
Comment linking /posts/2023/02/11/we-found-an-neuron
on “'I Worked on Google's AI. My Fears Are Coming True'”>Or, OpenAI added a calculator function. Furthermore, I would add: No lol. The test doesn't have to be mental arithmetic. and accuracy mistakes creep in at large numbers for multiplication. That's not how calculators work i'm sure you know >ChatGPT hasn't "learned" anything outside of what exists in it's model. I'm sorry...and you have? >If it is an AI-based response, it's still derived from token-based inferencing.…
Comment linking /posts/2023/02/11/we-found-an-neuron
on “What GPT-4 Does Is Less Like “Figuring Out” and More Like “Already Knowing””The architecture is understood, but the specifics about how they calculate which words are high-probability are mostly a mystery. Here’s a good blog post though: We Found An Neuron in GPT-2 https://clementneo.com/posts/2023/02/11/we-found-an-neuron If anyone knows of any other research like this, I’d love to read it.
Comment linking /posts/2023/02/11/we-found-an-neuron
on “Understanding ChatGPT”Our consciousness is biochemical statistics. It's a weak argument to dismiss AI as "not like us". https://clementneo.com/posts/2023/02/11/we-found-an-neuron
Comment linking /posts/2023/02/11/we-found-an-neuron
on “On ChatGPT”