ex-aws-dude

It is in the training data

People discuss weights on the internet

You can also have a model generate output and then just check what was activated

I don't think its that crazy that you could look at activations and tweak weights directly to get a certain outcome, but isn't the question whether you can do that without side effects or degrading/breaking other parts of the model?

If I want to do that in a generalizable way that preserves performance across all domains then am I not just back to training again?

I've been finding that too with these models

For example pointing them at a crash dump they manage to piece together and extract information I would not even try to extract by just looking at the raw memory

Anything where you are out in a forest

Pfft that is simply a special case of HackerRecommendationSiteGenerator.com

Ive noticed that too, I don’t know why people talk like we need to figure this out right now

Like we can even know what the future of math is this early

It’s not like you can just declare that up front, it’s something that happens naturally when people spend time using the tools

You don’t get it, what’s the actual profit gained to the company from that?

That’s the whole distinction

You’re not wrong, sometimes even a tweet by these people will be front page

Duh? If someone uses it to try hacking you its still 96.6% refusal

Are they even making 3M per month?

Isn't there a point where you're shipping too fast that the product hasn't even caught on enough yet to know what to build?

I looked at some of the games and it just looks like a Roblox clone

That is the tech bro circle of life

Invent Notion then keep adding features enterprise customers want until you're re-inventing JIRA/Confluence

Who left this stupid fence here?

It is kind of comical how things that previously would have been obviously true are now only statistically true

"Our product won't hack you 96.6% of the time"

I consider going online in general as going into the pig pen at this point

I’ll still go in to see what’s happening but don’t forget where you are and don’t take it too seriously

"The Grid. A digital frontier."

Not enough people are talking about the 1st point

Learning that it’s all just computation has some uncomfortable philosophical implications, perhaps religious too

Why would you be circumspect if the gap kept getting closer?

Doesn't that indicate its becoming _more_ true?

I guess if you view it like that, the model is the consultant itself

I don't think its always true though, for example the "super apps" in asia seem to do well

https://en.wikipedia.org/wiki/Super_app

Yeah I think efficiency is still a counterargument to the model doing everything.

If I need to process a million items it still wouldn't make sense to feed them directly to the model and burn a crazy amount of tokens

Of course the model can produce code to process them on the fly but if its a repeated task why would I want that to be re-created every time, especially if I've already tested/validated it

I mean it sounds stupid but I think its a fair question if you can spend 15 million and solve a millennium prize problem, what happens if you take the same resources and just aim the cannon at making money?

The thing is where is the line between a programming language and a game

What if your puzzle game is based around programming robots? How complex do you allow the language?

What I don’t get is how do so many companies make the exact same mistake?

It’s like corporations all get the same memo and do the exact same things in lockstep

Is that even a thing? I can't think of any case where company sold off game IP other than in acquisitions

You could not be more wrong

They were purchased decades ago

I don’t know if it’s like that everywhere but in LA it would not go on highway

I love taking Waymo but you have to compare the travel times.

Sometimes it will be like 45 minutes to go somewhere 15 min away because it won’t take any highways.

You're looking for an answer but the truth is no one really knows because we don't know how good the models will get.

Feels like the start of an OpenAI repeat