Hacker News new | past | comments | ask | show | jobs | submit login
Hosting a GPT-2 autoresponder bot (nostalgebraist.tumblr.com)
63 points by luu on Jan 8, 2020 | hide | past | favorite | 13 comments



I seem to remember a pretty large pop up like “do not abuse and don’t run bitcoin mining in collab” from the initial setup process?

In any case, Collab is a great tool, and I’m rather thankful to them. They may have decided to monitor usage instead of trying to restrict it to make abuse impossible.

I guess people these days belief that what is possible is also legal, and what is legal is identical to being moral. To those people I point out: Google noticing you is possible, shutting you down is legal, and banning you for life is ethical if only by protecting users that don’t abuse free resources.

Doubly so, If you also blogbrag about it.


If you know what Colab's definition of abusive or disallowed use is (or even whether they have such a definition), please let me know. This isn't meant as a gotcha -- I actually haven't been able to find a full TOS or anything like that, despite looking for one at the outset and then again while writing the linked post. If it exists I want to know.

As I said in the post, the closest thing I can find is the FAQ at https://research.google.com/colaboratory/faq.html (this may be what you're thinking of -- there is no "initial setup process," anyone logged in to a Google account can immediately connect to a runtime and execute notebook cells). It says the following:

Why are hardware resources such as T4 GPUs not available to me?

The best available hardware is prioritized for users who use Colaboratory interactively rather than for long-running computations. Users who use Colaboratory for long-running computations may be temporarily restricted in the type of hardware made available to them, and/or the duration that the hardware can be used for. We encourage users with high computational needs to use Colaboratory’s UI with a local runtime. Please note that using Colaboratory for cryptocurrency mining is disallowed entirely, and may result in being banned from using Colab altogether.

To me, this doesn't sound like "please don't use this wrong or too much, and we assume you know what 'wrong' and 'too much' mean here." Consider how this is written as an answer to a hypothetical user who has already been "temporarily restricted" and is looking for more info. They talk about this restriction non-judgmentally, like it's a practical limitation of their tool anyone might run into, not a moral line that "good" users will never cross anyway.

(The part about cryptocurrency mining is different and perfectly understandable, given the distinctive moral and perhaps legal facts of that subject.)

See also my post https://nostalgebraist.tumblr.com/post/190114849409/cloud-ca... for more thoughts on the subject. The screenshot there is from the 2017 conference talk https://gojko.net/2017/10/05/serverless-design-gotocph.html which happily discusses unintended but economically rational use of AWS products, like using IOT Gateway for non-IOT purposes where its pricing model is convenient.


Your attempts to convince yourself somehow make it look even worse. At least own the decision!

FWIW, whenever it takes you a long time to get something working, such as a constantly running Collab with a static IP or automatically updated link to it or DNS: that's a sign. Even more if, as in this case, what you are doing would be trivial to enable/make easier if Google wanted to.


Serious question: did you read the post? You seem to be critiquing something quite different from what I did.


"shitpost engineering" seems like one of those wonderfully evocative terms (like "premium mediocre") that seems to perfectly capture some aspect of our age.


> Google seems to think I value avoiding some slight inconvenience at $1000/month, and what’s more, they’ve chosen to provide not a free trial of a convenient thing (a tried and true approach) but a free inconvenient version of a convenient thing, forever. This can’t even sell me on the convenience of the “real” thing, since I’ve never seen it!

Well, tbh my personal experience with so-called “free trials” is a bad one because they often start to charge you after the free trial is over and when you have a lot of things going then you forget to cancel which of course is what they want but IMO a shady and immoral way of doing business.

I’ve been bitten by this a couple of times. Sometimes because I forgot to cancel on time, other times because I thought I was signing up for something with a one-time fee that turned out to be the monthly fee and they charge a full year.

So with my bad experiences I am extremely unlikely at this point to sign up for those kinds of trials if I can avoid it.

And when companies do that kind of stuff, depending on how they respond when I discover what happened and get in touch with their customer service, if it causes me grief then I will not do business with them again and I will tell my friends and family to not do business with them either.


Reach out to the TFRC team for some TPU credit: https://www.tensorflow.org/tfrc

They're definitely interested in people finding creative uses for TPU's and blogging about the process, especially with code!


Serving an app to 1k daily active user cost you for just $300/month but for similar daily user with an nlp app will cost you $3000/month. I think the GPU vms are overpriced because of the AI hype. Buying a GPU server like the old days is much more cheap.


TLDR; google provides computing resources for demo purposes for free, which can be exploited for non-intended use by clever automation


> google provides computing resources for demo purposes for free, which can be exploited for non-intended use by clever automation

...and there are some issues with this free model (Colab notebook timeouts and TPUs sometimes being unavailable) which the author is "unfairly annoyed [with] since I’m getting something for free".

See also "AI Dungeon 2 costing over $10k/day to run on GCS/Colab"[0] noting that the costs were not due to Colab itself but the bandwidth in copying the model from the GCP Storage Bucket to Colab.

[0] https://news.ycombinator.com/item?id=21739879


Including some interesting reflections around both the author as well as Google.


What's the reasoning behind routing everything through your laptop rather than just making the requests to Tumblr from collab? Do they block its IPs?


Just wariness about spreading around my tumblr credentials. There’s probably a secure way to do it, I just haven’t thought about it much.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: