Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
Closi
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
34 ms
·
661.
▲
by
Closi
3y ago
It's easy to invest a million in some index fund, but it's not easy to beat the market.
662.
▲
by
Closi
3y ago
Also note some of the limitations with Lexical - e.g. you can't even drag and drop highlighted text within a block.
663.
▲
by
Closi
3y ago
There isn’t - you are competing against big bank trading bots in a world where the best bots win.
664.
▲
by
Closi
3y ago
It also works well for questions that have easy/medium answers and aren’t asked a lot. Complex questions that involve things that aren’t well documented are obviously going to be harder for an LLM to answer, particularly when the LLM d
665.
▲
by
Closi
3y ago
They will have a different set of scaling problems to what Facebook had in 2005 - their decentralised architecture will throw out new challenges that will need time to be worked through. Has anyone else created a decentralised social networ
666.
▲
by
Closi
3y ago
If they opened the floodgates with a not-ready product and also fell over, that's a much worse look. Besides, 'limited' account creation to form communities is a pretty well-established scaling strategy. They have scaled from
667.
▲
by
Closi
3y ago
Parent comment said PC-exclusive, Witcher 3 and RDR2 are examples where the PC port came out way after the console release and they are both fairly old games at this point.
668.
▲
by
Closi
3y ago
IMO if you look at how much better GPT-4 is at coding compared to GPT-3.5, and advances in letting GPT test and debug it’s own code, it’s not going to be “cheaper and worse” in the future. GPT and LLMs will allow good and seasoned programme
669.
▲
by
Closi
3y ago
It’s like eBay - clothes are listed and shipped by users. If you did have to sort clothes from bags though you would probably just do it manually to start with, and if you had to scale up you would probably empty the bags onto a belt convey
670.
▲
by
Closi
3y ago
> Everybody remembers the fantastic opening sequence but seems to forget that it ends with dogs flying airplanes. That’s not the end of the story - that’s the start of the third act. It actually ends grounded in reality - a return back t
671.
▲
by
Closi
3y ago
We would also have to account for profit (let’s say 3%) and all overheads (let’s say 15%). Average car insurance charge is c$2000 so we can say $380 of that is probably overhead and profit. That leaves $1,640 left. $340B then divided by the
672.
▲
by
Closi
3y ago
Need to account for the fact that not all Americans drive. To get to the $1,035 per person they have taken the total cost ($340B) and divided it by the population, but if you divide it by the number of drivers (i.e. people actively insured)
673.
▲
by
Closi
3y ago
That’s not exactly the modern oculus experience - the Quest is “strap and go”.
674.
▲
by
Closi
3y ago
Well if another competitor starts using your protected term and you don’t protect it, it is more likely to become generic. (Ie if Apple does not protect the word iPhone and then Google and Samsung use the term iPhone for 5 years unchallenge
675.
▲
by
Closi
3y ago
It’s not entirely a myth, see a list of trademarks that have been “lost” due to insufficient protection or other reasons here: https://en.m.wikipedia.org/wiki/List_of_generic_and_generici... There are definitely exampl
676.
▲
by
Closi
3y ago
If you didn't learn anymore, could you still be conscious? IMO the answer is yes.
677.
▲
by
Closi
3y ago
Article is pretty thin on strong arguments. It seems to hypothesise that a computer could never be 'conscious' just because it is made of switches, but we are conscious and just made of bundles of neurons. The whole argument just
678.
▲
by
Closi
3y ago
These are maybe true for GPT-4, but probably don't hold for GPT-10 or whatever the landscape looks like in another handful of iterations.
679.
▲
by
Closi
3y ago
Future transport is unlikely to be single mode - it’s almost definitely multimodal and will include both cars (why not self driving?) and rail.
680.
▲
by
Closi
3y ago
Adoption sometimes takes that long, we just always forget the predecessors to the devices that made it big (eg the worldwide success of mobile phones was not overnight, and was limited to business people and the rich at first).
681.
▲
by
Closi
3y ago
I mean it's a stunning achievement in the same way counting out loud to a million is, but it doesn't necessarily mean I think it was a good use of time in the context of making a movie.
682.
▲
by
Closi
3y ago
LLMs can reason, and it’s surprising. I think some people get caught up on the “next word prediction” point, because this is just the mechanism. For the next word prediction to work, the LLM has all sorts of internal representations of the
683.
▲
by
Closi
3y ago
The Russia example there is a genuine example of de-dollarisation though, showing why exports matter.
684.
▲
by
Closi
3y ago
80 people to manage African countries from a finance, sales, marketing, support, legal, localisation and licensing perspective for a company the scale of Spotify is way too light. Localisation isn’t just about language - consider payment re
685.
▲
by
Closi
3y ago
> And looking at the Framework laptop shop right now, I see a 16GB DDR4 DIMM for $60, and you can throw in a standard M.2 storage module To be fair to Apple, theirs is LPDDR5 rather than DDR4, and it’s built into the SOC so it’s a differ
686.
▲
by
Closi
3y ago
Well they are programs capable of some amount of reasoning… Basic/moderate reasoning tasks are absolutely in the capability of GPT-4.
687.
▲
by
Closi
3y ago
Well you are right - Spotify operates in 184 markets, so even if you only had a small 10 person team for each country handling sales/marketing/support/localisation, that's already c2000 employees before you start on anyt
688.
▲
by
Closi
3y ago
Not always - take oil as the obvious example where transactions are usually done in dollars regardless of the shipping from / shipping to location. If you were in England and wanted to buy oil in Uganda for use in a subsidiary in Brazi
689.
▲
by
Closi
3y ago
Disagree - AI is getting much closer to solving this fast. Latest generation robotic arms with vision are taking on a greater variety of tasks (e.g. bin/case picking from a mixed tote), more reliably, with a better ability to adaptivel
690.
▲
by
Closi
3y ago
But humans eyes often look away from the road, close during a sneeze etc, and have a very narrow viewing angle compared to a car surrounded 365 degrees by cameras... so there are plusses and minuses. Human vision isn't that perfect for
More ›