HNHacker News
TopNewBestAskShowJobs

omeze

1 karma · joined August 17, 2024

submissionscomments
omeze··on If you're interested in eye-tracking, I'm interested in funding you
ALS really doesn’t leave much strength for having a headset strapped to you…
omeze··on Finetuning of Falcon-7B LLM Using QLoRA on Mental Health Conversational Dataset
people are already doing this on Character.ai and other popular AI consumer products, whether the companies want them to or not. I think it's better to accept people want something here and make it good & transparent vs trying to avoid it... it's like alcohol, sometimes people want a thing badly enough they don't care if its bad for them
omeze··on How Hinge's algorithm decides who you date
I think the standouts are generally people who are highly liked/rated. Id see a lot of people there before & after I saw them in my normal feed. Roses just give an escape hatch for pay to play but they’ll show up in your feed normally if the recommendation system thinks so.
omeze··on We’re all just temporarily abled
Theres a great guide from Microsoft on this (mentioned in the reply to this HN comment https://news.ycombinator.com/item?id=30062178)

We’ve all had to open a door with our hands full and thanked the wheelchair button by the door, or tried to watch a movie with the flu and a stuffed head so we snapped on CC.

omeze··on Strained housing affordability is a ‘manufactured crisis’ created by bad zoning
if a house is such a good investment why arent these investors building more to mint millions ezpz?
omeze··on Ask HN: Is making a self taught transition to AI/ML related fields possible?
Not just researchers, but theyre for experienced MLOps engineers. If youve built distributed training or inference pipelines at big tech, thats the sort of skillset that requires ML knowledge but has little to do with traditional research.
omeze··on Selling open-source software
Both your counterpoints are true, I didnt want to sprinkle the word “marginal cost” everywhere (marginal devops cost of buy vs self host). And ofc I agree, not all saas offerings are the same, EKS is fairly meh as far as setup and time savings.
omeze··on Selling open-source software
The truth is that people who are fine operating and managing their own ElasticSearch/Postgres/Spark whatever cluster are not the people who will be buying the SaaS offering. Theres no point trying to convert those people, theyll convert themselves once they get tired of constantly upgrading and maintaining a non-differentiated piece of their business, or they’ll just be happy users who bring your tech stack to their next gig, where the buy vs self-host choice gets revisited.

The main problem is that many teams think they’re “that guy (or girl)” who can manage a complicated piece of infra as a small team and will be saving money. If you spend 10% of your time doing devops work, you spend 10% less time on your actual product (you might as well be binging netflix). A majority of these teams convert (IME) to a hosted offering once the right price and SLAs and DevExp are there. The only part of your sales pitch that matters to these teams is “this isnt worth your time to self host when we have a great offering thats cheaper and faster to get setup”. If you have a logo from a big co that has the talent to self host but are buying your product, then your work is done.

omeze··on Simple browser boat game using Three.js
this is cool, is it using a physics engine like ammo or cannon? I like the water shaders too!
omeze··on 200k Compromised OpenAI Credentials Available for Purchase on the Dark Web
That's my worry, too. I'm reminded of the history of Intel, which at its founding decided that it would not have a research org (drop the R from R&D) and just focus on scaling FETs, because the founders learned that trying to do both meant that one would ultimately suffer.
omeze··on 200k Compromised OpenAI Credentials Available for Purchase on the Dark Web
It's not requiring logins per se -- this is the second mishap I'm aware of related to their consumer product surface area (the first being users' chat data going to other users). Just seems like a distraction, if the goal is to do fundamental AI research.
omeze··on 200k Compromised OpenAI Credentials Available for Purchase on the Dark Web
Yes, hence the MS deal (they had trouble raising money as a pure research org). Their research can be turned into a product people pay for -- they've done it! My question is: why did the engineers at OpenAI have to do that? Why couldn't they "toss it over the fence" to MS's Azure team and be happy with a revenue share, for example.
omeze··on 200k Compromised OpenAI Credentials Available for Purchase on the Dark Web
I seriously wonder why, given the Microsoft deal, OpenAI didn't just cede all product development and production-izing of their models to MS while they focus on research.

I know smart people work there, some friends included, but OpenAI was never a product-led startup, they were a research org. Ignoring the closed vs open debate for a moment: for all the talk of focus, focus, focus from YC, it seems unfocused to ship their models from a home-rolled platform when they could just get MS to foot the engineering and infra work. Is any of their platform stuff a real asset to OpenAI proper? Have MS do the cloud stuff, just pass them product design.

just my 2c from the outside.

omeze··on Researchers create open-source platform for Neural Radiance Field development
is there any work being done to make nerfs have collision or mesh deformation?
omeze··on Ask HN: Small Language Models instead of LLM. Does it makes sense?
I think it depends on the use case. The big win IMO is cost savings for developers, tiny models can run on devices so theres no need for inference servers. The real challenge is coming up with a use case that a tiny model can perform. I suspect there’s probably a lot though, like rephrasing or gmail-like suggestions. Theyre tiny features, not full apps, but still seem valuable to me

Fyi me and a friend made a demo app of tinystories we just submitted: https://news.ycombinator.com/item?id=36960333 - from testing it out, anything with <500ms latency and 15 tok/sec feels real-time, and that makes a world of difference in UX

omeze··on Simple on-device iOS AI demo app of TinyStories-1M
yea, I was also surprised at the lack of apps. I guess we can take a shot at llama-7b, maybe a base model and a storytelling fine tune. we'll have to look into which compiler is best, I do want to tinker with the mlc toolchain. will post to HN when we get it working!
omeze··on Simple on-device iOS AI demo app of TinyStories-1M
Hey all,

After reading the TinyStories paper (using small language models of 1m to 33m parameter size to generate short stories), me and a friend decided to port it as a test to see what performance of tiny LMs look like on mobile devices. The stories it generates aren't great or anything, it's really just a toy tech demo to play with. We used HuggingFace's exporters library to do this.

You can try it out here: https://apps.apple.com/us/app/tinystories-on-device/id645149...

(we are not affiliated with MS research or the authors)

omeze··on Show HN: Openexus – Building blocks for the internet
I don't understand how you make blocks, all of the demos seem to "just work" but it doesnt really make sense how some values are overrode and others have a default that auto updates?
omeze··on Teach your LLM to answer with facts, not fiction
Yea… my opinion on this is that startups are attempting to force a market for chat bots, instead of accepting that LLM embeddings are best utilized as a search feature, not new product surface area
omeze··on ‘World of Warcraft’ players trick AI-scraping website into publishing nonsense
Yea, its tragic. When I was a kid I would often just go to GameFAQs because that was the only real source, and the guides would be surfaced in google. Its still around but Google gives me 20 crappy ad riddled sites that require a new page click, listicle-style, to get through a 5 minute area
omeze··on The open source learning curve for AI researchers
gitsense is def a sick tool, ty for sharing
omeze··on Meta releases Intermediate Graphics Library
The asset stuff is the bottleneck here isnt it? We need browsers that support cacheing assets that are gigs in size
omeze··on Introducing Superalignment
yes I also have that impression. If you consider the concrete objectives, this is a good announcement:

- they want to make benchmarking easier by using AI systems

- they want to automate red-teaming and safety-checking ("problematic behavior" i.e. cursing at customers)

- they want to automate the understanding of model outputs ("interpretability")

Notice how absolutely none of these things require "superintelligence" to exist to be useful? They're all just bog standard Good Things that you'd want for any class of automated system, i.e. a great customer service bot.

The superintelligence meme is tiring but we're getting cool things out of it I guess...

omeze··on Chinese Tech Terms Explained in English
This is interesting, always fun to see the same work in a different light and perspective. Btw Id say chendian is “retrospective”, which most tech companies do after projects but some do them regularly. I also worked at a trading firm that did retros for big trades
omeze··on GPT4 is 8 x 220B params = 1.7T params
Ancestor model = pre-trained model that used a large diverse corpus

Descendant models = models fine tuned from an ancestor on one particular domain, e.g. by partitioning your training data by subject or source

Democracy = some weighted mix of the descendant models is used to find the next token

omeze··on Show HN: Answer Overflow – Indexing Discord content into the web
It wasn't meant to be "private" in the e2e sense. It was just meant to be a modern reimagining of Ventrilo/TeamSpeak, which were group chat programs for gamers. The video call stuff came after (since it was strange to not have video if you were already voice chatting in the year 201X). They want to branch outside of servicing online gaming communities but that was their origin, and you can still see how that culture affects their product design today. A good way of understanding the product is to just ask "would online gaming communities use this?".
omeze··on Ask HN: What would an IDE built for the Apple Vision Pro look like?
This is very cool!

I was thinking of a VS code extension that lets me put files or directories “physically” on my desk. Itd be emulating the way that I read text books. Instead of keeping notes and bookmarks physically in the book I could drop sticky notes that represent files directly on my desk and show a preview by by just shifting my eye focus to the AR/VR note.

omeze··on Ask HN: What would an IDE built for the Apple Vision Pro look like?
This is flame-baity but honestly what Id also expect. No one is going to invest in developing IDEs for this thing to rival VS Code or whatever if Apple wants a stranglehold on its ecosystem. Maybe there will be extensions to existing editors since people can ise it as a display extension for their macs. We didn’t see this happen on windows with game dev and the quest though.
omeze··on First impressions: Yes, Apple Vision Pro works and yes, it’s good
The first iphone let you have a real web browser anywhere you had cell coverage. This lets you have a web browser floating at home. Its not the same (and I love VR)
omeze··on Apple Vision Pro Is Apple's New VR Headset
Im extremely skeptical that Apple will let this be used with VRC via PCVR. It would need to be a new build for the device with support for tracking. Plus those body trackers don’t suffer from occlusion.
← PreviousPage 3 of 8Next →