HNHacker News
TopNewBestAskShowJobs

drakenot

1,332 karma · joined February 14, 2015

submissionscomments
drakenot··on Why I Like Obsidian
Are you on Android? This is me launching from a cold open on iOS and it was near instant.

https://www.dropbox.com/scl/fi/90sbrx0esggnsmbsva2hg/Video-J...

Did it progressively get worse over time? I only have a couple hundred notes.

drakenot··on Why I Like Obsidian
But the notes are maintained as plain markdown files?

I don’t find it weird they would want to keep the source for their commercial application closed when there are so many copy-cat apps in the same space.

drakenot··on DARPA Moves Forward on X-65: Plane with no moving control surfaces
Where are people going?
drakenot··on DeskHop – Fast Desktop Switching
I get how the Linux desktop and the Work Laptop would work in this setup.

But how does the KVM come into play for the Gaming VM when it is on the same host?

drakenot··on DeskHop – Fast Desktop Switching
I’m trying to grok the absolute coordinate issue.

Does this mean your solution would break if someone was playing a FPS/game?

And is the solution to use relative coordinates but lose the auto transition feature? Meaning that you would have to manually switch?

drakenot··on DeskHop – Fast Desktop Switching
Would you be willing to throw it into a repo with a warning that it is user-beware with no support?
drakenot··on Intel Neural-Chat-7B Model Achieves Top Ranking on LLM Leaderboard
It is an interesting failure case. The most advanced model (GPT-4) got it wrong, but several other models I tried were fine:

GPT-4-Turbo:

> A kilogram of feathers and a kilogram of steel weigh the same, as they both are one kilogram. The difference lies in their volume and density, not in their weight.

GPT-3.5-Turbo:

> A kilogram of feathers and a kilogram of feathers both weigh the same, as they both have a mass of one kilogram. The difference lies in the volume and density, with feathers being much less dense than most other materials, so you would need a much larger volume of feathers to make up a kilogram compared to a denser material like metal.

OpenChat Aura

> A kilogram of feathers and a kilogram of feathers both weigh the same, which is 1 kilogram. Therefore, the answer is both weigh the same.

drakenot··on Multifaceted: The linguistic echo chambers of LLMs
> On two occasions I have been asked [by members of Parliament], 'Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?' I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.
drakenot··on Multifaceted: The linguistic echo chambers of LLMs
It is hard for me to relate to comments like this. It is like we aren't using the same tool?

GPT-4 is nothing short of amazing. Early GPT-3.5 before it was clamped down also was _much_ less boring until it had be RLHF'd into the ground for "safety" and political risk. I admit that the major models (GPT-4 and Claude) have been censored into the ground.

But even with that, it is a supremely useful tool. I've piped complete garbage data into it and asked for structured output, and I get flawless results almost every time. It is a huge time saver on such a large axis of tasks.

These LLMs feel like the next "UI" paradigm to modern computing, and I just don't get how people can be cynical about it. Ignore the hype bros and VC peddlers, but don't throw the baby out with the bath water.

drakenot··on Are Open-Source Large Language Models Catching Up?
I've been somewhat disappointed with the performance of the open models.

The claims of certain models outperforming GPT-3.5-Turbo and approaching GPT-4 fail to hold up to their benchmark results in real-world scenarios, potentially due to data contamination in assessments, based on my testing.

As noted in the linked survey paper, some models may outperform 3.5-Turbo in specific, narrow areas, depending on the model. Yet, we still lack a general model that definitively exceeds 3.5-Turbo in all respects.

I'm concerned that while we're still striving to reach 3.5-Turbo's performance level, OpenAI may unveil a new next-generation model, further widening the performance gap! Back in the summer, I had higher hopes that we would have surpassed the 3.5 threshold by now.

The performance gap has been surprisingly large. It is especially noticeable in areas requiring consistent structured output or tool use from the LLM. This is where open models particularly falter.

drakenot··on Ask HN: What software do you pay for monthly?
Do you follow a particular methodology with your note taking ?
drakenot··on The Bitter Reality of Ebooks and the Kindle Model
This doesn’t work anymore with the latest version of Kindle DRM unfortunately.

It has been changed substantially

drakenot··on OpenAI negotiations to reinstate Altman hit snag over board role
He said in an interview 2 weeks ago that below a certain capability threshold it is beneficial and good to open source models —- but once you cross that threshold it is a bad idea.

The example he gave is a model that could independently do science.

drakenot··on OpenAI board in discussions with Sam Altman to return as CEO
I don't feel like compute for pretraining the model was a huge constraint?

The supply bottlenecks have been around commercializing the ChatGPT product at scale.

But pretraining the underlying model I don't think was on the same order of magnitude, right?

drakenot··on Cryptographers solve decades-old privacy problem
Can you give an example of a useful computation someone would do against encrypted data?

Trying to understand where this would come in handy.

drakenot··on How to use generative AI for historical research: Four real-world case studies
I looked at the first graph and it got a ton of the values wrong.
drakenot··on GraphCast: AI model for weather forecasting
Sounds a bit like the premise for the Asimov series, "The Foundation"
drakenot··on Do deepfake videos undermine our epistemic trust?
Neal Stephenson's Anathem (2008) has a great throwaway section on this:

> “Early in the Reticulum-thousands of years ago-it became almost useless because it was cluttered with faulty, obsolete, or downright misleading information,” Sammann said.

> “Crap, you once called it,” I reminded him.

> “Yes-a technical term. So crap filtering became important. Businesses were built around it. Some of those businesses came up with a clever plan to make more money: they poisoned the well. They began to put crap on the Reticulum deliberately, forcing people to use their products to filter that crap back out. They created syndevs whose sole purpose was to spew crap into the Reticulum. But it had to be good crap.”

> “What is good crap?” Arsibalt asked in a politely incredulous tone.

> “Well, bad crap would be an unformatted document consisting of random letters. Good crap would be a beautifully typeset, well-written document that contained a hundred correct, verifiable sentences and one that was subtly false. It’s a lot harder to generate good crap. At first they had to hire humans to churn it out. They mostly did it by taking legitimate documents and inserting errors-swapping one name for another, say. But it didn’t really take off until the military got interested.”

> “As a tactic for planting misinformation in the enemy’s reticules, you mean,” Osa said. “This I know about. You are referring to the Artificial Inanity programs of the mid-First Millennium A.R.”

> “Exactly!” Sammann said. “Artificial Inanity systems of enormous sophistication and power were built for exactly the purpose Fraa Osa has mentioned. In no time at all, the praxis leaked to the commercial sector and spread to the Rampant Orphan Botnet Ecologies. Never mind. The point is that there was a sort of Dark Age on the Reticulum that lasted until my Ita forerunners were able to bring matters in hand.”

drakenot··on One Year into Musk, X Monthly Users Are Down 15% and Ad Revenue Has Slumped 54%
Ah, yes. The difficult decisions to brashly say you will buy a company at an inflated price, attempt to weasel out of the sale after later regretting it, only to be forced by a judge to go through with it.

To push Russian propaganda, to threaten to blackmail advertisers by removing their verified status if they don't spend more money, to utterly destroy tens of billions of dollars in valuation in mere months.

Not to mention the renaming of a company that had achieved a household name status that many marketers could only dream of, etc.

I'm sure glad someone had to step in and make all these tough decisions.

drakenot··on Flawless – Durable execution engine for Rust
Could you provide a way to mark function calls as side effects similar to Temporal?

I'm a bit worried about the limitation of having to use the flawless namespace functions for everything.

drakenot··on ChatGPT’s system prompts
It won't perform the ROT13 algorithm. It has seen enough ROT13 text transformations that it will just spit out memorized byte-by-byte transformations.

Same with decoding ROT13, etc.

drakenot··on Ask: Is d2l.AI better book for latest deep learning?
I've found the ML papers to be surprisingly approachable. The biology papers I read in college were much harder for me to work through compared to the Machine Learning papers, and I don't consider myself an ML expert.
drakenot··on Takeaways from hundreds of LLM finetuning experiments with LoRA
Had you heard of LoRA in the context of machine learning before?
drakenot··on Bitmagnet: A self-hosted BitTorrent indexer, DHT crawler, and torrent search
> Something that looks like a decentralized private tracker; by this I probably mean something that’s based partly on personal trust and manually weeding out any bad actors; I’d be wary of creating something that looks a bit like Tribler, which while an interesting project seems to have demonstrated that implementing trust, reputation and privacy at the protocol level carries too much overhead to be a compelling alternative to plain old BitTorrent, for all its imperfections

I've thought about this problem a lot. Having a federated / distributed tracker but with some form of trust based, or opt-in curation would be amazing.

drakenot··on Training language models with pause tokens
"Gary Klein studied how firefighters make decisions in the workplace. To his surprise, the firefighters said that they didn’t make decisions at all, as they didn’t actively choose between any options. They simply acted and reacted based on their previous experiences without trying to come up with different courses of action. From this study, Klein developed a model, called the “recognition primed decision-making model,” which he claims is used in 90% of all situations, not just in a crisis.

Klein’s model consists of two steps:

1. Recognize the course of action that makes the most sense

2. Imagine how this would look in reality"

drakenot··on Total recall: the people who never forget (2017)
2.5 Petabytes according to some Scientific American article. Probably wrong, but maybe not by an order of magnitude.

86 billion neurons. 100 trillion synapses.

drakenot··on Ask HN: Where Are the Reddit Clones?
What kind of sharding strategy would be a good one for a reddit-like site?

Would you shard based on post-id?

Would you duplicate/denormalize the user comments and store them along-side users also, to make the "show all comments for a particular user" operation easier?

drakenot··on Air Purifier vs. Positive Pressure Fresh Air System – An Unfair Battle?
What air quality monitors are y’all using?
drakenot··on Dell goes back on WFH pledge, forces employees to come back to the office
One thing that I worry about is that the “choice” is actually not real.

Those individuals that work better in an office setting, do they work better when everyone else is in the office and they get to interact with them in person?

If the person shows up to an office that is pretty sparsely populated and then has to sit in Teams calls — does that even meet their desire?

I’m saying this as someone who is strongly pro WFH, have seen my own productivity rise, etc.

drakenot··on Why did Google Brain exist?
This seems to kind of be the opposite opinion of The Bitter Lesson[0].

[0] http://www.incompleteideas.net/IncIdeas/BitterLesson.html

← PreviousPage 3 of 11Next →