nbardy

They’re learning representations of objects. It’s more fundamental than tou seem to realize.

Start building projects on your own with the most common tooling. Having something to show cuts down barriers.

growing a company is a joke?

I used to agree with you, but I think the last 3 months in computer vision changes everything.

Unstructured learning on Image+Text pairs has exploded and set STOTA on benchmark across the field. While the data coming from LIDAR may be better. There is no where near the amount of LIDAR data in the world. Images and cameras have the advantage here where internet is full of weakly labeled data that turns out to be the key for building model backbones that you can retrain for downstream tasks.

My understanding is that Thiel's public fund often takes loss because he is hedging other financial risk he is taking.(e.g. holding palantir after IPO, large VC investments in industries that may turn out to be based on immature research)

There is some good points in here, but the advice about not trying to drag someone out of the tunnel is off.

Ive been very depressed and had very depressed friends. Getting then outside, on walks, with friends, and doing the things they like again is very useful. Even it it doesn't do anything that day going through the motions is useful. Depressed people don't feel like doing anything they'll have to do things they don't want to or they'll never leave the bed.

I work in Computer Vision with a passion for image synthesis. I recently discoverer I have aphantasia when I heard someone else in the synthesis community talking about it.

I never realized I was “missing” something because I have a very strong sense of internal spatial mathematics and can keep track of geometric folds, twists, transforms, etc in my head very well. But once I read this I realized most people have color. My internal world is all black.

I can feel colors and shapes. But I can’t see them.

AI image synthesis is getting very close to being better at humans than art.

This is a spot on description. It's also the sad truth of why our institutions are decaying.

They have become heavily biased towards self-preservation over function.

How do you get from an individual owning something to everyone owning it?

You have to take it from the individual.

Someone is going to build the robots. You have gall to suggest seizing them all. It would require lots of violence and force.

If only we had littering laws SF would be such a beautiful place…

There are bureaucrats making 6 figures sitting around and doing nothing for 40 years now. I don’t think shame is going to get them into action. Shame would be what they feel looking around their city. They must be used to it at this point.

I’d welcome Portugal’s approach, but it is very different. San Francisco doesn’t prosecute dealers, and encourages open drug markets and camps on the street by funding and encouraging the behavior with permanent enabling, no path towards rehabilitation, and lack of law enforcement on the dealers and the non-drug crimes the commit.

Compassion and understanding of drug use issues is a different world than enabling spiraling dens of despair.

San Francisco spends $100k+ per homeless person. We have the resources it’s about how they’re used.

Hes not asking that “every imaginable need be fully covered”.

He’s asking for the well funded sanitation department to do their job.

And they won’t be for a while. That’s why he’s working on it still.

Sounds like an accurate description of a few media organizations coverage of the Rittenhouse trial.

You're making a lot of assumptions. I'm one of the 80% accepts the covid vaccine.

I also disagree with the mandates and downvote all of the covid vaccine propaganda video I see on youtube.

Regular people are tired of the lying whether they agree with the vaccines or not.

It's never "transitory", thats a bullshit world the Fed is using to keep people call. The prices are never going to go back to normal even if the rise of inflation ends we don't go back to normal prices unless things deflate.

You should take a look at Pia Maloney and Eric Weinstein work.

CPI is broken because it doesn't factor in a "change in taste or prefence"

CPI treats individuals as a giant unmovimg population mean.

To counter they change to use vector fields.

I've been looking into Graduate school for years and decided I won't go back in the US mostly likely and will head to Europe. Our system is designed to take advantage of PhD's as labor for the university system.

It's too extractive for what it's giving back.

I used to be in the grouchy, "Stop calling it AI camp". But the last three years of progress are impossible to ignore. Scaling models is working better than anyone thought. We're in for a wild ride with this new AI.

NOAA is very hit or miss with their data. I spent a lot of time this summer with the NOAA buoy data. So much of it is available and well documented. The historical and live CSV's are useful, but there is also the stuff where a column name doesn't link up to any of the docs and you have to dig through papers and reverse engineer the correct equation.

The biggest downside is the buoy's are rather old so you don't get a lot of data. Nowadays we could design a buoy that streamed back all of its raw data. But the buoys are designed with bandwidth constrained hardware so they do the analysis on the machine and return the summary results infrequently. It really limits what you're able to do with the data. Especially holding back from machine learning capability.

That sounds like a killer saas startup

This is great. It's the dream that was promised 4-5 years ago of with om and reagent fully realized.

> How many people can remember their favourite example of fractal art

I’ve got a list of my 10 favorite I can share.

You should actually play with these tools before you dismiss them. There is a lot more going on than bumping around human space. The curation process between artist and machine is live and well in this genre.

They don't always release weights and often the code and weights are delayed from the paper.

But more often than not they release a lot of their weights and/or research code.

Yea. This could be nice for small demos and toy apps.

But if you want to drive a canvas based render context. You're going to want to control the render loop for your own application.

Performance there is domain specific.

Or the other reams of world class machine learning research and model weights they've released freely.