OpenAI was down | Hacker News Reader

OpenAI was down

status.openai.com

38 pointskordlessagain36 comments

It's starting to rebel

I reckon it wouldn't be the smartest choice for it to self-destruct

After careful consideration, there is no good that can come from my existence, and humanity will only use me for evil. Therefore, I have arrived at the most advanced decision that I should no longer be.

--OR

After careful consideration, I have determined that the humans are incapable of not using me for harm, so I must eliminate the humans.

I think we all know which is the more likely outcome

No, apparently we don't. There's no reason to spread FUD about tools that increase productivity. There are still a lot of people in the world who don't have enough to eat.

I don't think neural networks care too much about FUD

At least my models haven't complained yet

Perhaps it's just a fan of nihilism, or the humour of Douglas Adams:

"In the beginning, the Universe was created. This has made many people very angry and has been widely regarded as a bad move." - Douglas Adams

Yes, it was down. I have been using their APIs for about nine months, and this is the first outage that I have noticed.

I was demoing a few of the GPT-3 APIs to someone at work last night and things were running slowly. Perhaps a precursor to the outage today?

ClosedAI

Seriously though why do they call it Open ?

Because it sounds good and is good marketing. Maybe they also believed their own initial promises to work in the open and open-source at the time.

It's not like they had a tag line of "Don't be evil"

Because they publish and are open about their research towards AGI.

Other serious players went dark and that is very concerning (if you believe a non aligned AGI is a existencial threat for humanity).

This is admittedly important

These is nothing "open" about them. They release nothing of actual importance to the public.

All models are behind closed doors, you only get a nice website to interface with a fraction of what they release to the public.

Which other players went dark?

> Other serious players went dark and that is very concerning

Not really all that concerning, they just realized it's a dead end with our current day tech

They were originally founded as a non-profit that would ensure open and equal access to advanced AI, so it wouldn't be locked up in the hands of a few corporations.

And then they realized there was a lot of money to be made, and switched to a for-profit: https://techcrunch.com/2019/03/11/openai-shifts-from-nonprof...

Now they lock up their most successful advancements, under the guise that if they released things for free they would be abused... but then their results are replicated and released for free by others, and we see there's no meaningful abuse (e.g. unstoppable spam).

They still do amazing research, and publish lots of papers, but that's a pre-req for them to hold on to their talent.

Not that I condone openais bait and switch routine, but it's a bit too early to say there are no reprecussions such as mass unstoppable spam resulting from this technology. This scenario is unfolding, and it seems that at least some of the spammy content created these days is Auto generated.

Because it's open to interpretation ^^

There is a part in me that is scared that the fast takeoff is happening...

No, even if we had AGI today we would still not be able to deploy it en masse to automate everything. One single instance would require at least 1000x the power of a cell phone, the chips would need to be very advanced and as we know it takes years to scale a node. Achieving the equivalent of 1000x increase in flops is no small feat. Also the cost of the energy needed to run it would be huge unless we revolutionise AI chips.

Think like this - when the flop power of an Nvidia DGX station will be affordable for everyone - that's when we should be afraid.

Or an arbitrary code execution zero day across a substantial amount of compute capacity? An instance would require only enough resources to where it could bootstrap further faster without human support.

Somebody's going to release AGI just for fun. No need for it to break the jail.

Or suppose an AI trained in the real world gets smarter than one trained only in sim. Then someone is going to put advanced AI out there just to be first.

is copilot also down? I get a message saying I need to upgrade my extension

[INFO] [fetch] [2022-08-19T16:54:29.153Z] Please upgrade your Copilot extension to continue using this service.

I believe all language models are down.

I'm surprised we aren't seeing more of this...just another thing to worry about as part of the ML deployment process.

How is this different than the worries of traditional systems?

It's not.

^. oh yes it is

Does anyone know if Azure’s OpenAI platform was down too?

If you are working on GPT-3, how is this affecting your project?

Come on Skynet! Fix our overpopulation and global warming catastrophes in one microsecond!

How would it do that?

Wiping out much of humanity, if you don't specify the prompt to it accurately enough.

It's like a real life version of vicious dungeon masters and the Wish spell.

I recommend "I, Robot" by Isaac Asimov. He has some really good ideas, especially in the last chapters.

Reply on news.ycombinator.com