return_to_monke

an argument for it is that it couples content and styling together, instead of them being in different layers.

Sometimes, they are, and should be, separate. But in other cases, styling is related to the content, or part of it.

that would sadly hurt more, I think.

Journals are eager to publish new studies as facts (did you know that eating sugar will make you lose weight???) (ignoring that all those who were studied were pro byciclers burning off 3000 calories a day)

this will just make it worse, because a bit larger amount of bad papers will get publushed

also: https://youtu.be/USKD3vPD6ZA

they are probably banned from doing any business in eg. Russia

how else would one know they aren't sponsored by Big HDD?

The nix language, albeit for a package manager, is really similar. You get json-like "attrsets" and can have variables and functions. Then of course you can export it to different formats as well.

https://nixos.org

There is also the Dhall and Nickel languages providing features alike

The title encompasses my thoughts about the LLM perfectly. It is amazing to see that a very weird concept (randomising data, testing it, and randomising the best ones at predicting the next tokens again) would work but it seems to do.

Of course, this is not intelligence. these chatbots should come with a "HEY THIS IS NOT INTELLIGENT AND VERY NOT A HUMAN" warning sticker.

Yet, I wonder were we are in the hype cycle. If have hopes if this will somehow go on to improve.

Maybe by AI companies fine-tuning on initial prompt -> last response [what the user accomplished, by prompting gpt incrementally] pairs?,

I think we could come like 50-60% close to what a human mind can accomplish.

last thought; from my own experience, toddlers sometimes hallucinate / string random words together, too.

Tick tick has a free version. Reflect does not.

The internet has turned people into a mindless mob, also including myself.

Most people tend to take the same side their favourite celebrities do, and then argue for that to the extremes.

I think your first statement is incorrect. Lora seems to be a method to fine-tune and optimize the weights of models like Alpaca. It is not a different dataset.

This reduces model sizes and therefore also compute costs.

See the abstract of https://arxiv.org/pdf/2106.09685.pdf

a) you shouldn't be demoted because of your political donations, as long as you're not donating to do harm - whether Eich was donating to a harmful organization is debatable

b) This has nothing to do with "wholeness" and everything to do with Google pushing chrome as hard as they can.

- There is android phones, and I am pretty sure some Google services would advertise it to you - Chromebooks are used in most public schools across America

c) Eich later went on to found Brave. Have you ever seen anyone in the real world use Brave? They passed 50M users one year ago: https://brave.com/2021-recap/

Wikipedia states 63% of people used the internet worldwide in 2021: https://en.m.wikipedia.org/wiki/List_of_countries_by_number_...

that is 4.9 billion people. So, brave has a market share of, drum roll, 1%.

In conclusion, the argument that Eich would have led them to success is not a good one, in my opinion.

we could do a rematch tho, where gpt goes first and Bart answers.

about 5 months late, and not public. not looking that great.

of course I can ! just pull in libcurl and then it is just parsing cli arguments

I do not think so.

I see this argument made all the time. While it might be correct at a Micro-Level ("all it does is predict the next word!!!" "it just finds patterns!!!") I dont think that semantics matter here.

It has demonstrated to create new things based on existing ones.

ChatGPT will happily create a fan fiction about the rust mascot Ferris, in the style of the Bible.

I am 99% certain that wasn't in the training set

hn is much more ... extreme

I wonder how fast this badge will get into datasets, and AIs, which will then try to replicate it - as part of the content they generate

they were a non-profit at some point, iirc.

to be fair, that is large part because CH seems to have a topLevelDomain function

There are multiple such measures. Ones I could think of right off the bat is hip/waist ratio, BAI, or simply waist circumference.

Those all fail to reliably predict body fatness in all populations.

All of those measurements should be taken in context with other health signs, like blood pressure, fasting glucose and maybe eating/exercise habits as well.

I think the 3rd point is different for each person. Some people do like "tough love" - being told that they are in a bad place and need to improve will motivate them.

Others, on the other hand, will be pushed further back into the hole by such talk. This is very individual.

also, for a lot of use cases (eg accessing your home-hosted stuff, on the go) simply tailscale, even without funnel, is fine

I'd rather have false lints than miss lints

i think voice assistants can perform actions on phones (eg "open app, message Alice, call Bob, turn off Bluetooth"). This couldn't do that (I think), which is an obvious drawback

Does this mean that MS, with their new stake in OpenAI, makes OAI release a new model, or just that they'll release a model called GPT4?

I feel like GPT* is forever connected with OAI. prove me wrong, tho.

I think most servers will likely just blacklist the FB instances and move on.

Because nix develop is inherently a flakes-bound feature, and Nix Flakes aren't officially stable yet.

because languages that were popular before but aren't now are always a hit.

see COBOL for example.

wait, so did they lock themselves out of their api?

sorry for killing the mood. great article and great progress.

but. didn't openai start as a nonprofit, too?