Are there any security risks from this? I'd love to give it a try but am worried about the risk of a black box running arbitrary code on my computer

Turns out it might actually be possible?

https://www.youtube.com/watch?v=B7MIJP90biM

That's awesome.

Do you mean like trap streets? Seems like a good idea for model makers

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

Yeah, something like that may be already happening and various actors building their cases as we speak.

I don't understand. ChatGPT cost an estimated 10s of millions ot train. ChatGPT 4.0 has much better performance than the next best model. Isn't that a moat?

Think of it as a time series. It cost 10s of millions to train but in 6 months gpt4 open source equivalents will cost 100$ to train. The best model is one that you can build on top of in a way that’s it’s not a black box (SD).

Where are the gpt4 open source equivalents going to come from?

Does anyone have insight into why the static fire test did not reveal stage 0 issues?

They did. But there was a leadership decision to go ahead anyways.

They didn't light all of the engines for the static test, I think.

If you don't mind rambling, I found it an enjoyable read. You may want to skim sections 1-2 or skip entirely to section 3.

Really cool! From my understanding, it looks like this is doing something like word embeddings and searching for nearby points in the embedding space.

Crazy idea: what if you used a dimensionality reduction like t-SNE instead of learning a vector representation? Would you expect similar results?

Thanks! Yup, this is basically custom music embeddings + nearest-neighbor vector search.

I personally found that vector representations performed significantly better than other approaches.

And the results will actually be a lot better once I ship a better model (the current one can definitely be improved upon).

Link without paywall: https://web.archive.org/web/20220804035851/https://www.scien...

Not sure I agree. A "short fat" could theoretically be replaced by a group of "short skinny", whereas a "tall skinny" could only be replaced by another "tall skinny".

As value is largely determined by scarcity (supply and demand), this would cause "short fat" to be of lower value.

So you're saying one engineer who's good at lots of things can be replaced by a dozen engineers who are only good at one thing?

It doesn't work that way because coordination costs are a big part of the cost.

If you count cards, the pit boss will ask you to leave.

Personally, not a fan of the casino metaphor.

+1, fully agreed. There are definitely people making stupid bets and losing (like they should, IMO), but the notion that markets are this black box beyond human comprehension is laughable. That's especially true when you filter out the noise and find companies with solid fundamentals.