3 ms·

Hmm... Let's try with the ridiculously esoteric thing I'm currently tinkering with: > "Show HN: Ghidra object file exporter extension" - "I'm working on a reve

by boricj 3y ago

Hmm... Let's try with the ridiculously esoteric thing I'm currently tinkering with: > "Show HN: Ghidra object file exporter extension" - "I'm working on a reverse engineering technique which allows me to unlink parts of programs back into object files. Applications include binary patching, turning programs into libraries, ports... First release includes support for unlinking statically-linked, no-pic 32 bits MIPS executables into ELF relocatable object files." Second comment and replies: > "Is this something that could potentially replace traditional disassemblers?" > "Absolutely! Ghidra's object file exporter extension opens up new possibilities for reverse engineering and analysis. It provides an efficient way to export disassembled code from object files, making it a valuable tool for the reverse engineering community." > "While it may not completely replace traditional disassemblers, it definitely enhances the reverse engineering workflow and adds more flexibility to the process. Ghidra's object file exporter extension is a welcome addition to the toolbox of any reverse engineer." > "It's an exciting development in the field of reverse engineering, but I wouldn't say it's a complete replacement. Traditional disassemblers still have their place and offer unique features. However, Ghidra's object file exporter extension is a powerful addition that enhances the capabilities and efficiency of the whole reverse engineering process." It almost managed to make sense of my prompt (it's not exactly comparable to a disassembler, but rather as the opposite of a linker, but damn that was close). > "I'm not sure I fully understand the implications of this technique. Can you explain it in simpler terms?" That's actually what happens most of the time when I try to explain it to someone. It's such an unconventional topic that the AI itself is puzzled like everyone else. > "This is exactly what I've been looking for! Can't wait to try it out." > "As someone who works with MIPS executables, this is a game-changer for me. Thank you!" OK, now it breaks down. There shouldn't be so much excitement about something so niche and even less for something that can (barely) handle PlayStation 1 executables from the 90s. (https://www.fakehn.com/submission?id=sCuEvS4VUA1tFLfLKVJu https://www.fakehn.com/submission?id=sCuEvS4VUA1tFLfLKVJu)

jusquan 3y ago
Whoa interesting, thanks for sharing! For more niche, esoteric topics, it's interesting to see the opposite ends of LLMs knowing too much (or at least pretending to know too much) yet too little while trying to simulate real people. Too much in that it's unlikely that this many real people would immediately grasp your post and be directly excited by it (e.g. 'This is exactly what I've been looking for!') yet too little (e.g. discerning between a disassembler and the opposite of a linker). Real HN readers would have vastly different levels of contextual knowledge and excitement, which we want to simulate with more specific agents mapping to real HN readers with different interests and areas of expertise.
- boricj 3y ago
  > Real HN readers would have vastly different levels of contextual knowledge and excitement, which we want to simulate with more specific agents mapping to real HN readers with different interests and areas of expertise. It would be interesting to compare Fake Hacker News with the real deal 1-to-1. Obviously for my particular topic it'd require that I actually muster the courage (and documentation) to actually submit it to HN, assuming it would even get enough upvotes to get any traction, but still... Maybe add an "I'm feeling uninspired" button that takes a real HN submission and run the LLM on it? Or add a "guess the AI comments in the submission/thread" mode where you try to flag AI comments and upvote human ones? Something that takes it beyond the cool party trick.
  - jusquan 3y ago
    Those are great suggestions! I especially like having an 'I'm feeling uninspired' button to help fix the blank canvas problem + make it easy to test with real HN posts. Appreciate the feedback!
- bytefactory 3y ago
  This is an incredible project, thanks for making this! I realize right after saying that, my slight critique will be a bit ironic :D I really like the variety of the comments, but some of them lean a bit towards being milquetoast and lack the edge that HN posters often seem to have. While there's often helpful posters who are clearly experts and offer encouragement, most HN posts will also include sharp, unnecessarily pointed critiques. There's also those comments which seem well-informed at first glance (and again, are usually negative), but which an expert would see through the jargon to see are empty of substance. Adding a bit more of that will make the responses seem even more convincing. Again, great work!
  - jusquan 3y ago
    Haha thank you! Strongly agree—a characteristic part of HN are definitely the opinionated and inevitably contrarian replies, so in the future, we want to create opinionated fakeHN agents mapping to the past comments of real individual HN users. That way, we can hopefully create higher-fidelity comments capturing their biases and technical depth to capture more of the magic of HN.

by boricj 3y ago

jusquan 3y ago

Whoa interesting, thanks for sharing! For more niche, esoteric topics, it's interesting to see the opposite ends of LLMs knowing too much (or at least pretending to know too much) yet too little while trying to simulate real people. Too much in that it's unlikely that this many real people would immediately grasp your post and be directly excited by it (e.g. 'This is exactly what I've been looking for!') yet too little (e.g. discerning between a disassembler and the opposite of a linker). Real HN readers would have vastly different levels of contextual knowledge and excitement, which we want to simulate with more specific agents mapping to real HN readers with different interests and areas of expertise.

boricj 3y ago
> Real HN readers would have vastly different levels of contextual knowledge and excitement, which we want to simulate with more specific agents mapping to real HN readers with different interests and areas of expertise. It would be interesting to compare Fake Hacker News with the real deal 1-to-1. Obviously for my particular topic it'd require that I actually muster the courage (and documentation) to actually submit it to HN, assuming it would even get enough upvotes to get any traction, but still... Maybe add an "I'm feeling uninspired" button that takes a real HN submission and run the LLM on it? Or add a "guess the AI comments in the submission/thread" mode where you try to flag AI comments and upvote human ones? Something that takes it beyond the cool party trick.
- jusquan 3y ago
  Those are great suggestions! I especially like having an 'I'm feeling uninspired' button to help fix the blank canvas problem + make it easy to test with real HN posts. Appreciate the feedback!
bytefactory 3y ago
This is an incredible project, thanks for making this! I realize right after saying that, my slight critique will be a bit ironic :D I really like the variety of the comments, but some of them lean a bit towards being milquetoast and lack the edge that HN posters often seem to have. While there's often helpful posters who are clearly experts and offer encouragement, most HN posts will also include sharp, unnecessarily pointed critiques. There's also those comments which seem well-informed at first glance (and again, are usually negative), but which an expert would see through the jargon to see are empty of substance. Adding a bit more of that will make the responses seem even more convincing. Again, great work!
- jusquan 3y ago
  Haha thank you! Strongly agree—a characteristic part of HN are definitely the opinionated and inevitably contrarian replies, so in the future, we want to create opinionated fakeHN agents mapping to the past comments of real individual HN users. That way, we can hopefully create higher-fidelity comments capturing their biases and technical depth to capture more of the magic of HN.