Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Hugging Face raises $235M from investors including Salesforce and Nvidia

techcrunch.com

Post reply on HN
Print view View original

71–80 of 207 posts

Re: Hugging Face raises $235M from investors including Salesforce and Nvidia

#71

Post by htrp »

unmole wrote: post #52
htrp wrote: post #50

Earlier quoted context omitted.

> a) a hardware vendor nvidia whose products are needed by anyone in the game FTFY.... I don't think AMD will be competitive in AI until at least 2025, though given their broken roadmap promises, not sure if anyone will be willing to invest in them

I mean you still need CPUs to run your workloads. And AMD makes decent ones. I work for a company that sells data centre network switches. There's nothing AI specific in our boxes but we will sell a lot more if them if people are building more data centres to cater to AI driven demand. Something, something gold rush, shovels.

Great point! But while demand for network switches explodes, do you really get a competitive advantage from one company's switches vs another? (serious question as I've never had to build a DGX superpod or anything like it from scratch).

Re: Hugging Face raises $235M from investors including Salesforce and Nvidia

#72

Post by frozencell »

stagger87 wrote: post #49
chakintosh wrote: post #36

HF is the literal implementation of the old idiom "During the gold rush, sell shovels."

I've said this on HN before, but Nvidia is the one selling shovels IMO.

IMO HF sell tents and Nvidia sell shovels

Re: Hugging Face raises $235M from investors including Salesforce and Nvidia

#73

Post by michaelt »

0xcde4c3db wrote: post #55
version_five wrote: post #38

Earlier quoted context omitted.

They represent a real risk to the AI ecosystem IMO (or opportunity if you're an investor). They have really cornered the market on model hosting as well as frameworks for running them, it's going to get ugly when they start turning the screws. I think it's important for people to diversify away from them and not build anything that uniquely depends on them. It's not good to have chokepoints like this.

Diversify to what, though? The economics of hosting models and datasets seem to just not be great in general. For the same reasons that we can expect HF to eventually crack down on the "freeloaders", I think pretty much any other entity also will unless it's explicitly set up as a public benefit. I could see this sort of thing maybe being funded by universities or their affiliated foundations, but most companies with…

Pip, maven, npm, docker hub, apt-get, github and suchlike seem to manage to provide hefty binary downloads with no need to monetise anonymous downloads.

I don't quite understand how they do so, but it seems to be possible. Everyone who downloads a 5GB stable diffusion model probably downloaded 5GB of pytorch+cuda+cudnn first.

Re: Hugging Face raises $235M from investors including Salesforce and Nvidia

#74

Post by mistrial9 »

htrp wrote: post #70
jeroenhd wrote: post #51

Earlier quoted context omitted.

Their Git LFS based hosting is used by a lot of AI tools. Huggingface level data storage and transfer would also be VERY costly to accomplish for a small party. Some of these models are several gigabytes in size and downloaded hundreds or thousands of time per day. What they do isn't too hard to replicate, but the price for which they do it is impossible to compete with.

If Amazon/Google/Microsoft wanted to own that space, they could do it in a heartbeat. Rumor has it Github is actively working on something similar to HF Hub already

agile ; community-minded ; efficient

pick one if you are lucky. Those Fortune-12 giants can't do anything "in a heartbeat" .. listen to what that means.. nonsense

Re: Hugging Face raises $235M from investors including Salesforce and Nvidia

#75

Post by forgingahead »

htrp wrote: post #70
jeroenhd wrote: post #51

Earlier quoted context omitted.

Their Git LFS based hosting is used by a lot of AI tools. Huggingface level data storage and transfer would also be VERY costly to accomplish for a small party. Some of these models are several gigabytes in size and downloaded hundreds or thousands of time per day. What they do isn't too hard to replicate, but the price for which they do it is impossible to compete with.

If Amazon/Google/Microsoft wanted to own that space, they could do it in a heartbeat. Rumor has it Github is actively working on something similar to HF Hub already

I've seen some repos on Github host the models on "Releases". Not sure how viable that is long term, but definitely easier to download and access than some random Google Drive link which is typical for a lot of other repos.

Re: Hugging Face raises $235M from investors including Salesforce and Nvidia

#76

Post by Lord_Zero »

michaelt wrote: post #73
0xcde4c3db wrote: post #55

Earlier quoted context omitted.

Diversify to what, though? The economics of hosting models and datasets seem to just not be great in general. For the same reasons that we can expect HF to eventually crack down on the "freeloaders", I think pretty much any other entity also will unless it's explicitly set up as a public benefit. I could see this sort of thing maybe being funded by universities or their affiliated foundations, but most companies with…

Pip, maven, npm, docker hub, apt-get, github and suchlike seem to manage to provide hefty binary downloads with no need to monetise anonymous downloads. I don't quite understand how they do so, but it seems to be possible. Everyone who downloads a 5GB stable diffusion model probably downloaded 5GB of pytorch+cuda+cudnn first.

Back in the day, universities would host all that shit via their mirrors for free for everyone. Not sure where it all is today with the cloud being so prevalent.

Re: Hugging Face raises $235M from investors including Salesforce and Nvidia

#77

Post by unmole »

htrp wrote: post #71
unmole wrote: post #52

Earlier quoted context omitted.

I mean you still need CPUs to run your workloads. And AMD makes decent ones. I work for a company that sells data centre network switches. There's nothing AI specific in our boxes but we will sell a lot more if them if people are building more data centres to cater to AI driven demand. Something, something gold rush, shovels.

Great point! But while demand for network switches explodes, do you really get a competitive advantage from one company's switches vs another? (serious question as I've never had to build a DGX superpod or anything like it from scratch).

> do you really get a competitive advantage from one company's switches vs another?

I am obviously biased but IMHO, yes. High end switches are not commodity gear. And a lot of the differentiating secret sauce is in software. So, even if two switches from two different vendors have the same silicon, they might have dramatically different characteristics. That is without going into other aspects like integration with the rest of the stack, quality of support and even lead times.

Re: Hugging Face raises $235M from investors including Salesforce and Nvidia

#78

Post by lsmeducation »

jeroenhd wrote: post #51
htrp wrote: post #48

Earlier quoted context omitted.

Do they really? As far as I can tell, diffusers/transformer are open sourced wrappers around torch implementations, there are a bunch of companies that are offering inference (just like they are), and there is some value in offering S3 storage to allow for model search. If they reach a point where they actively become community-hostile, someone will just fork their codebase and release a web app called "FaceHugger"

Their Git LFS based hosting is used by a lot of AI tools. Huggingface level data storage and transfer would also be VERY costly to accomplish for a small party. Some of these models are several gigabytes in size and downloaded hundreds or thousands of time per day. What they do isn't too hard to replicate, but the price for which they do it is impossible to compete with.

Lol, I’m sorry what? Has anyone heard of BitTorrent or piracy in general? You’re telling me AI models/data are too big?

Someone utilize p2p for the love of god. Pretty sure there’s gigabytes of porn moving (freely) through the internet millions of times per day.

Re: Hugging Face raises $235M from investors including Salesforce and Nvidia

#79

Post by bakuninsbart »

version_five wrote: post #38
QuadrupleA wrote: post #16

Hugging Face is early in the Silicon Valley enshittification cycle - currently burning VC money being incredibly good to it's users. Next would be shifting that value to its business customers. Then clawing back and squeezing as much value as possible from those. Then collapse. Although, is it technically a platform / marketplace? Anyway for now, enjoy the free beer from the VCs.

They represent a real risk to the AI ecosystem IMO (or opportunity if you're an investor). They have really cornered the market on model hosting as well as frameworks for running them, it's going to get ugly when they start turning the screws. I think it's important for people to diversify away from them and not build anything that uniquely depends on them. It's not good to have chokepoints like this.

You can replace this with "container" and "docker", and it turned out exactly or similar to how you predict. However, there is nothing with a big moat that Huggingface is offering. If they overeshittify, competitors will rise up.

Re: Hugging Face raises $235M from investors including Salesforce and Nvidia

#80

Post by CardenB »

ramesh31 wrote: post #15
kiratp wrote: post #14

I love Huggingface. I worry they will be the next Docker. What is the moat? "We will run your inference" can't be the answer.

>What is the moat? "We will run your inference" can't be the answer. Network effect. It's more like a GitHub for ML. Also trust. When your business is hosting binaries, that's no small feat.

What do people think separated GitHub from docker?

Docker also had DockerHub, but it wasn’t as necessary as GitHub and didn’t catch on as well.

I think HF is more like docker than GitHub. GitHub had the wide scope to be huge. Docker was too niche (in comparison) and didn’t catch on so much to be necessary for every single project like Git did.

Similarly, HF is also niche in the sense that docker is. There’s no need to host your model on HF, it’s just a convenience for some projects. It’s effectively a package manager for ML models.

It seems that if there’s a way to make models more easily runnable, HF would not be necessary at all. They have a community there, but it’s not something that can be monetized well.

Post reply on HN