Google has banned the training of deepfakes in Colab • Hacker News Board

Re: Google has banned the training of deepfakes in Colab

#31

Post by Datenstrom » Sat, May 28, 2022, 4:32 PM UTC

endisneigh wrote: post #20

fartcannon wrote: post #19

Earlier quoted context omitted.

Oh yeah, how?

Class action, violation of the Terms and Conditions. I'd honestly be shocked if Google/Microsoft are monitoring the clients of their public cloud and keeping the data to create their own products off of.

I'd be shocked if they weren't, it seems like a totally standard big corp playbook move. It would be the same exact thing Amazon was doing with their independent seller data[1].

[1]: https://www.wsj.com/articles/amazon-scooped-up-data-from-its...

Re: Google has banned the training of deepfakes in Colab

#32

Post by hugh-avherald » Sat, May 28, 2022, 4:35 PM UTC

nestorD wrote: post #4

I really do not like that because the ban feels fuzzy. If you are a researcher exploring those algorithms and using Colab because you have very limited GPU access of you own, you might now be worried that your very own topic of research will get you banned from the ressource you are using.

Your complaint appears to be "funding is a constraint on research".

Re: Google has banned the training of deepfakes in Colab

#33

Post by fartcannon » Sat, May 28, 2022, 4:38 PM UTC

endisneigh wrote: post #23

fartcannon wrote: post #21

Earlier quoted context omitted.

Prepare to be shocked.

proof?

The existence of copilot, for one.

Re: Google has banned the training of deepfakes in Colab

#34

Post by albert_e » Sat, May 28, 2022, 4:46 PM UTC

ClassyJacket wrote: post #24

version_five wrote: post #5

I have wondered in the past what kind of monitoring / data harvesting google is doing with colab. I don't use them for anything nontrivial or work related precisely because I worry about google stealing my stuff or that I would be violating agreements with my customers by exposing their data to google. This story reinforces that they are snooping. That aside, I don't think I find this too upsetting. They don't let pe…

Is there a good alternative to Colab you can just pay for, that's good for hobbyist individuals? Especially one that gives you shell access? I tried using PaperSpace for a while but it works maybe 1 in 3 times I tried to use it, they (temporarily) banned me for no reason, they send me "itemized" bills with empty line items, and their support was no help.

- Amazon Sagemaker Studio Lab (free, email signup, no AWS account or credit card required)

paid services on AWS:

- Amazon Sagemaker Studio

- Amazon Sagemaker Notebook Instances (think EC2 + jupyter + integration with AWS services)

Re: Google has banned the training of deepfakes in Colab

#35

Post by jakobov » Sat, May 28, 2022, 4:46 PM UTC

The regulation of ai begins

Re: Google has banned the training of deepfakes in Colab

#36

Post by endisneigh » Sat, May 28, 2022, 4:47 PM UTC

fartcannon wrote: post #33

endisneigh wrote: post #23

Earlier quoted context omitted.

proof?

The existence of copilot, for one.

that's not proof, lol. That uses publicly available code. are you suggesting it's trained on private code? not to mention I was talking about cloud workloads since you said everything

Re: Google has banned the training of deepfakes in Colab

#37

Post by endisneigh » Sat, May 28, 2022, 4:48 PM UTC

Datenstrom wrote: post #31

endisneigh wrote: post #20

Earlier quoted context omitted.

Class action, violation of the Terms and Conditions. I'd honestly be shocked if Google/Microsoft are monitoring the clients of their public cloud and keeping the data to create their own products off of.

I'd be shocked if they weren't, it seems like a totally standard big corp playbook move. It would be the same exact thing Amazon was doing with their independent seller data[1]. [1]: https://www.wsj.com/articles/amazon-scooped-up-data-from-its...

that's not the same thing, since it's public info.

Re: Google has banned the training of deepfakes in Colab

#38

Post by thegreatdukd » Sat, May 28, 2022, 4:52 PM UTC

One thing I don't like is that they also banned SSH-ing to Colab, which means I can no longer remote SSH to VS Code and use Copilot on the stuff I am working on.

Re: Google has banned the training of deepfakes in Colab

#39

Post by chroem- » Sat, May 28, 2022, 5:00 PM UTC

hugh-avherald wrote: post #32

nestorD wrote: post #4

I really do not like that because the ban feels fuzzy. If you are a researcher exploring those algorithms and using Colab because you have very limited GPU access of you own, you might now be worried that your very own topic of research will get you banned from the ressource you are using.

Your complaint appears to be "funding is a constraint on research".

By default, Google will not allow you to access GPU instances on regular GCP. The GPU quota is initialized to zero and you have to open a support ticket in order to request GPU access, justifying why your use case deserves access. Evidently, "personal research project" is not a valid use case, so that leaves Colab as the only other option in the Google ecosystem.

AWS has a similar quota policy, but they seem less strict about use cases.

Re: Google has banned the training of deepfakes in Colab

#40

Post by version_five » Sat, May 28, 2022, 5:02 PM UTC

ClassyJacket wrote: post #24

version_five wrote: post #5

I have wondered in the past what kind of monitoring / data harvesting google is doing with colab. I don't use them for anything nontrivial or work related precisely because I worry about google stealing my stuff or that I would be violating agreements with my customers by exposing their data to google. This story reinforces that they are snooping. That aside, I don't think I find this too upsetting. They don't let pe…

Is there a good alternative to Colab you can just pay for, that's good for hobbyist individuals? Especially one that gives you shell access? I tried using PaperSpace for a while but it works maybe 1 in 3 times I tried to use it, they (temporarily) banned me for no reason, they send me "itemized" bills with empty line items, and their support was no help.

What I settled on is I have a computer with a lower-end nvidia gpu that lets me debug code and do basic stuff but is impractical for training any large model or with lots of data (although since the incremental cost is only the electricity, you can often still get far by training overnight or when you're doing something else). And then I use AWS more deliberately once I've figured out exactly what I want to run. Otherwise, you're reserving a cloud GPU for all the setup and debugging.

If getting a gpu is not practical, I'd try one of the K80s or other cheaper GPUs on AWS for all your debugging and then port to the more expensive instances only if you need them. Doing this as a hobbyist will be relatively inexpensive. I think a lot of the cost incurred as a hobbyist comes from reserving a gpu when you don't actually need it.