noosphr

It only has that depth because you prompt it towards the experts that understand that depth.

An example that happened 10 minutes ago: contracts in racket, it kept arguing that you can't use -> in a contract of a function with a rest argument. I had to mention ... explicitly that it wrote the code correctly.

So does llm inference. You're lucky if you hit 40% of the advertised flops.

All vibe coded projects are in the public domain by definition.

The question is if the massive copyright infringement that all the major labs engaged in means that all work derived from llms is under the copyright of the original owners of the work.

It is quite possible to keep track of what piece of training data changed the value of what weight to the point that the activations for any piece of code can be attributed back to the original copyright holder, including the random initial values which are public domain. I've done this on a toy project for gpt2.

If you have to ask then the answer is no.

Tens of millions in tokens for a code base which no one will understand in the end.

I've found that model performance starts substantially degrading after 10%.

Emperors new clothes.

Google wants to be the AI that everyone uses, not the plumbing.

Give it two years.

Watching language package managers reinvent everything distribution package managers have been doing since the 90s has been as fun as watching crypto people reinvent financial regulation.

Under any sane statistical model. If we assume that this is a naturally occurring once in century El Nino under an equal probability model we'd expect not to see it before we have 70 years of records.

This isn't climate, it's weather.

There are no proxies that record monthly water temperature. Only second order effects that are at best weakly correlated with water temperature.

I'm not sure how you can make that claim with only 29 years of data without making some pretty big assumptions about the underlying distribution.

This is why you need lisp.

First you write a programming language in which the language primatives are the business logic primatives, then you write a trivial program that describes what the business does in the new language.

If I may suggest people move away from harasses and start using Emacs. Gptel is amazing and the rest of Emacs plays incredibly well with llm development. Since the human is always in the loop token costs are also tiny compared to anything else.

Kids also need their own time too. Having a parent constantly interfering is just as bad as having one never showing interest. And you're not a playmate, you're a parent.

Children follow by example. If you're an addict your kids will be too.

Mindless cheerleading is what got us in the state we're in. More of it will not fix anything.

And in Singapore I can get 10Gbps home internet.

The future is not evenly distributed and the outliers matter little.

Electricity generation is almost always much, much cleaner than burning fuel in an internal combustion engine, and has been since long before EVs became available.

Depends entirely on your assumptions about where you draw the line for emissions from second and third order effects.

Saudi oil is so easy to get out of the ground that compared to coal it's practically green.

We still don't.

Until free/at cost car charging is provided at work when the sun shines you're just moving the place where combustion happens.

This may be a worth while trade in downtowns where the main form of car pollution is engine exhaust at an average of 20mph.

People doubt the timelines, not the claims. The timelines are pure fantasy and have been since the start.

Can't wait for the EU brain control bill.

We support our citizens right to free will so long as they don't think anything bad.

People are turning on immigrants because the pie has been shrinking for decades.

The EU wants to be the only one who spies on its citizens.

Go in a car from 1970 and try the seatbelts.

I can see why people didn't want them.

I too would rather not have a stiff blade like plastic meterial nearly cut my head off everytime the car breaks.

By comparison today we have luxurious silk strands that don't pinch anywhere.

That's the cost of pollution.

Every day we stray further into depravity and barbarism.

It was data. Alexnet worked because imagenet existed. It took less than a year between the dataset becoming public and the first version of Alexnet being trained.

Measuring software performance by lines of code is like measuring aircraft performance by weight.

I have no idea why everyone seems to have forgotten this simple fact over the last four years.

Computer science is to programming as physics is to plumbing.

If you want to be a plumber get an apprenticeship.