Well from the article a dataset is required, but not always the heavier one.

Tho it doesn’t solve the speed issue, where the llm will take a lot more time to do the compression.

gzip can compress 1GB of text in less than a minute on a CPU, an LLM with 3.2 million parameters requires an hour to compress

Hmm, ok. I see. When clicking on the last button in a category tree, it chooses randomly an instance.

Tho for that specific category, I cannot access https://lemmy.sdf.org/. Why? I don’t know, maybe broken or regional block?

If regional block, that is an argument to warts looking at user location. Tho not sure if it would be useful for anything else.

It would also be interesting to categorize more instances. Maybe even put smaller ones to distribute the load. Tho sometimes the smaller ones may not be well prepared for a lot of users. And at the same time the local feed may not be the most active. Tho the all feed may be interesting for new users.

The instance where I am, compuverse.uk is a general tech/computer instance, but had issues with storage for example.

So when does the random instance is chosen? When the website is loaded? When the user clicks on a category?

Maybe currently there aren’t enough instances categorised on the website to get enough randomness.

Is there, or would there be, a weight in the randomness in order to chose closer instances based on the user’s location?

For example sh.itjust.works is an instance based in Canada. When the reddit exodus happened. That instance was slow due to the distance. While others closer to the western europe were faster because they where closer to me.

Tho rn, it’s about the same. So not sure.

Well for it to get a chance in the US where the other side of big tech is, it would need to be sold to another company than Huawei, or for the US to relax the ban.

I have no idea why the US banned Huawei so intensely other than blocking competition from China.

I am sory for not living in a giant home without any 2.4ghz neighbors around me.

Also I myself use 2.4ghz for low bandwidth devices, and other connected devices which don’t have 5ghz. And there are also the neighbors who cast their 2.4ghz network they may use how they will around me.

But I didn’t have many issues on most of the headsets I tried. At least higher end.

There were still some outliers and with bad unreliable connections, even on high latency (jbl tour one m2 for example).

Others like corsair headsets had issues at “high” range (like 10-15m through walls), but no issues at close range, but that is expected from their specs.

Well look at wifi vs bluetooth. Wifi and bluetooth both use 2.4ghz. But wifi has a lot more bandwidth on 2.4ghz. Maybe because it uses more channels (a bigger frequency range)?

But there is room to improve bandwidth, even over distance. Tho how would they have done that, no idea. There is need for more time so devices with the tech can arrive on the market, and be tested.

But also more time so more specialised content can be produced to inform us normal people.

The new creative zen hybrid pro hybrid and pro sxfi seem to be advertised to use bluetooth 5.3 with LC3 and LC3+.

Tho not if they are on the market yet or if there are reviews.

I’m not sure it’s always the only thing people want. Yes stability is needed, even more when a lot of people are using it.

But currently headsets are heavily missing (very) high quality wireless audio and mic at the same time. Maybe this tech could push towards that.

Not really like that.

Right now it’s not possible to connect a headset/headphone via wifi to a device other than some proprietary things.

So a general competitor usable on all devices allowing more data transfer for more audio with less compression. I think it could be interesting.

Not just because it would maybe be better. But because competition on a market is a good thing for the consumer. And push bluetooth maybe further than what it is.

Tho if by some misunderstanding, the chip used isn’t compatible with bluetooth I’m not very sure… As brands would need to include 2 chips, which increases the cost.

Here is another article it seems to have more info https://techcrunch.com/2023/09/25/reddit-will-start-paying-you-real-money-for-your-karma/

redditors need to earn at least 10 gold within a 30-day period — if they don’t reach the threshold, the balance rolls over. For users with between 100 and 4,999 karma, they will receive $0.90 per 1 gold. Once you earn more than 5,000 karma, you can earn $1 per gold.

It’s so bad… and what will happen is spam cheap bot content to get the gold amount necessary to get paid.

There aren’t just concerns, it’s a f prediction.

redditors need to earn at least 10 gold within a 30-day period — if they don’t reach the threshold, the balance rolls over. For users with between 100 and 4,999 karma, they will receive $0.90 per 1 gold. Once you earn more than 5,000 karma, you can earn $1 per gold.

Yeah… I’m not sure about the quality being really better than competition. Tho maybe Linux support may be better depending on what components different laptops use.

System76 is a bit strange. Their laptops are more expensive than other windows brands, while offering Linux. I’m not sure if it’s really worth going with them, as a lot of other laptops ca work with Linux. Tho maybe I missed something about their laptops.

Framework I think I saw some Poole saying it has some issues running Linux with some component incompatibilities. Tho you may want to do more research, maybe even opening some reddit threads and dodging useless comments.

The generative ai have been enhanced with more training on the hands. They hot much better.

Do you have lots of cash to support the copyright and maintenance fees? The Web archive (I think, or another similar website) got sued for having knowledge and book contents on their website. You can’t just publish hacked books.

Most likely you would need a way to generate money for a “business”.

There would be some ways like advertising (can be OK, but not always the best, it depends on how many adds, how scummy ads, and how much tracking/data selling).

Donations? = pretty much no expected revenue. The amount of revenue you can get is very low.

Paid features, but what features could be paid while keeping free access?

Business proposal : destroy my money

For me, it shows 1 subscriber, 1.4k users per month.

It seems to be entirely dependent on the instance.

It can be a pre-planned trade. However they also could have waited for the shares to be sold then make the announcement.

Who knows. I’m not here as a judge.

It can be more than a fine.

They could be forced to sell parts of Google, to stop paying browsers, or whatever the court decides. However if the court decides in favor of the antitrust, then it will also put in place a precedent which will impact the other companies too.

Because no one traches them well what to do and believe.

Social media education and a critica view of the Web is still much behind the current state of social media.

TikTok is the last place I’ll go check for any serious advice. Just opening the app is spamming me with horrible cringe content.

And the rest are very much unreliable. There are some serious creators, but they are coming from other platforms with reputation.

The other TikTok only creators may as well be false content until verified.