pjfin123

I have a Dell Latitude 5490, which was cheap refurbished from Amazon, and it works great with Ubuntu.

Go to a thrift store and buy a dumb TV. I bought a 32" dumb TV at the Salvation Army for $60. It's not a flat screen but is only 2" thick. It plugs into HDMI for my laptop and Nintendo Switch, has a nice (dumb) remote for volume and brightness, and has worked wonderfully for 3 years now.

The TVs have gotten a lot better in the last 10-15 years so even basic ones work great. I would have been thrilled to have this TV as a kid.

> Each attempt to ban a book by one of these groups represents a direct attack on every person's constitutionally protected right to freely choose what books to read and what ideas to explore.

I don't think school library book banning is the biggest threat to free speech today; it's mass internet censorship.

They use all sorts of dark patterns to upsell you on their paid services. I still use them because TurboTax is generally easy to use and gets the job done.

On Android you can still install Brave or other browsers.

I'm pretty skeptical of "AI safety research" being a useful field of inquiry. I'm guessing the the problems AIs create among themselves will be similar to the ones Humans create. Things may start changing so fast it gets disorienting though.

My strategy is increasingly to not worry about optimization and instead write good semanticly correct code that the compiler can optimize.

Copyright needs to be largely abolished.

This is ridiculous; we created A is that can program themselves and people are worried about incidental copyright infringement.

They're leading the way with open source models, I wish them the best!

They're the ones that released Stable Diffusion (the ML model that can generate images from text prompts).

I don't think defined benefit pensions are a very good idea.

I'd like to see hardware more suited for running sparse ML models. You don't need branch prediction like on a CPU but you also don't need as much shared memory as a GPU.

Almost anything can be modeled as a translation of one sequence of character to another sequence of text characters.

Argos Fantasy TCG is a trading card game available in the public domain that is designed to be compatible with the MTG game. All of the cards are based on, and have equivalent abilities, to real MTG cards in the Legacy format. All copyrighted material material from the the original cards has been removed so these cards can be freely redistributed and printed allowing anyone to play regardless of financial ability. Game rules and mechanics are not covered by copyright law so these cards are allowed to have identical functional abilities to the cards they are based on. The art has been replaced with art that was already available in the public domain in the United States (and most other countries), and the description of the abilities has been rewritten (although the function remains the same). For card names such as "Hail of Arrows" that simply consist of dictionary words the name's from the original MTG cards is used since short phrases aren't protected by copyright in the U.S., however, when the card name is a based on a name of a potentially copyrighted character in the MTG universe like "Serra Angel" then the name is changed to a generic name like "Attacking Angel" even though the function is maintained.

All cards created using the open source image editing software GIMP, contributions welcome!

A RL agent with a large Transformer seems to me like a design that would build an AGI.

The base rate for exchanges getting hacked is definitely over 0.01%

America is effectively the only country left with free speech/expression.

Western governments had claimed for a long time to value free speech and democracy but once social media actually gave normal people the ability to compete with large media companies governments quickly changed their mind.

Yeah this is the opposite, they did impressively well with fewer parameters.

In general larger models and more data has been an effective strategy for getting better performance but getting the right ratio is also important: https://www.deepmind.com/publications/an-empirical-analysis-...

The most notable thing about this model is that they use fewer parameters (20 billion) than many of the other LLM which makes it less resource intensive to train and easier to run.

They also use an encoder-decoder architecture, which is common for machine translation, unlike most large language models which are decoder-only.

https://community.libretranslate.com/t/alexatm-a-20b-multili...

I've heard the talent denial to competitors reason but even for companies as big as FB and Google this seems unrealistic to me. Is hiring an expensive engineer really worth it to prevent one of many other potentially competing companies from hiring them? This makes sense for very specialized roles but for general engineering it seems like there would be more cost than benefit.

Based on this Python tutorial: https://gist.github.com/ymoslem/a414a0ead0d3e50f4d7ff7110b1d...

I took this class!

A difference with crypto is that they won't be receiving a taxpayer funded bailout for any excesses so there's less moral hazard in the long run.

I think Tesla (partial) self driving cars have the lowest accident rates.

Cyber security is a real concern though, imagine a mass hack terrorist attack where everyone's Teslas start acting unsafely simultaneously. If I had a self driving car I'd want a hardware kill switch.

This looks like a cool project, I had issues packaging Qt apps with Python for the Argos Translate GUI https://github.com/argosopentech/argos-translate-gui

Not ideal but I've used dual licensing under MIT and CCO.

Google really seems to be declining (outside of DeepMind).

Keynes advocated government borrowing and spending during a recession to cushion the decrease in aggregate demand. Critics of this deficit spending claim that the market should in the long run correct itself to which Keynes famously responded "In the long run we're all dead."

The U.S. government did this type of deficit spending in 2020-2021 printing and spending a large amount of money/debt which moved future spending to the present.

Clearly not all is bad but they don't have your best interests in mind and I would argue it's overall quite harmful.

It's pretty crazy that we allow ourselves to be bombarded with the psychological attacks that are advertising. Ads use pictures of food or sex appeal that manipulates our lizard brains or social attacks to convince us to buy products we don't need. If you pay close attention a large percentage is for negative sum products like unhealthy and addictive fast food.