12B is pretty small, so I’m doubting it’ll be anywhere close to internvl2 however mistral does great work and likely this model is still useful for on device tasks
1,141 karma · joined March 12, 2012
For example, on mobile phones it used to be the case that the application processor wasn't actually the main processor of the phone, it was actually the baseband, and relatedly, this meant that the application processor could be shutdown and woken up when the phone received a text message or phone call. Similarly, modern arm processors have a big/little configuration where you can switch to the efficiency cores when the system is asleep.
While I'm aware that at some point intel started offering efficiency cores, the vast majority of devices that run windows, AFAIK, don't have any way of being awake in sleep mode other that just leaving the CPU running with the screen off.
Maybe s0ix will make more sense as big.LITTLE and other secondary CPU schemes become common, but today seems like it's just upsetting existing users for no apparent gain
> Humble Games is an indie video games publisher owned by Humble Bundle that is dedicated to bringing you great games the Humble way.
So, humble bundle owns humble games, and this doesn't (explicitly) mean anything will happen to the humble bundle DRM free game store front
With "only" 16gb of ram, a relatively meagre 8 core 6800 series APU, and small screen it wouldn't make sense for most software developer workloads, and because of the attached controller(s) it's not super portable so not great for content consumption.
Other than gamers, who likely don't even care that the steam deck runs linux (and in fact are hindered by it in some ways) is there a group you can imagine that would appreciate preinstalled linux so much that the steam deck makes sense over the surface pro or even a framework?
Based on what? Anecdotal I know, but my biggest frustration is I can't convince friends and family they should buy a small car instead of an SUV because they always say "I might need to move a couch/dresser/friends/etc one day". I'm convinced they would come out way ahead buying a small car and renting a truck the once or twice a year they need it but again, I've yet to convince anyone that's the case
[0] https://www.theverge.com/23820078/matters-biggest-problem-ap...
Presumably the fear here is that iPhones/ipads/macs will be border routers by default and since they likely have internet access they would "force" the system to accept messages from outside the mesh.
One example that stood out to me though:
> GraphQL discourages breaking changes and provides no tools to deal with them.
GraphQL the standard doesn't provide tools no, but I've been very successful using Apollo Studio to solve this as (in my experience) the workflow maps to how you generally develop applications:
1) agree on some schema 2) make a (Apollo studio) "branch" with the schema change 3) mobile & backend devs (typically) code gen the objects they need for the schema 4) backend dev deploys a preview branch of their implementation, mobile dev points the client to it 5) test it, merge it, publish the schema on the main "branch" - deal with the breaking changes if Apollo warns you of them.
So maybe you can accomplish this with other tools, and maybe it's not fair to compare "the standard" to a particular tool, but I usually say I stick to gql (where appropriate) because there is a tool for it that works so well for the problem(s) I'm typically solving.
I was however tripped up by this sentence close to the beginning:
> we encountered a significant challenge with RAG: relying solely on vector search (even using both dense and sparse vectors) doesn’t always deliver satisfactory results for certain queries.
Not to be overly pedantic, but that's a problem with vector similarity, not RAG as a concept.
Although the author is clearly aware of that - I have had numerous conversations in the past few months alone of people essentially saying "RAG doesn't work because I use pg_vector (or whatever) and it never finds what I'm looking for" not realizing 1) it's not the only way to do RAG, and 2) there is often a fair difference between the embeddings and the vectorized query, and with awareness of why that is you can figure out how to fix it.
https://medium.com/@cdg2718/why-your-rag-doesnt-work-9755726... basically says everything I often say to people with RAG/vector search problems but again, seems like the assembled team has it handled :)
The fuel economy is actually quite reasonable unloaded - I'm getting 11L/100km vs 13 on my 370z.
The point I'm trying to make though is that it's probably not worth it to look at saving cents/kw when you're likely to come out ahead either way.
And yes, maybe more to the point of the original article - yes, if you have no car and you're only concerned with the cheapest to operate vehicle I'm confident something like a Prius driven gingerly and only filled up while gas is historically low (over a week window or whatever) you could beat an EV. On the other hand, that doesn't describe most people with cars already so for most people even living in an apartment they'd likely save money.
For example to go “500km” in the model y charging at home with the cheapest electricity it’s about $6. At a super charger I pay about $30. 5x more expensive but for my f150 I pay $120-150 for the same distance. Between the two options charging at the super charger every time is still 4x cheaper than gas even though I could save even more charging at home
My first impression was that you'd "just" upload my pictures to OpenAI with a prompt and call it a day.
Maybe highlight that it uses ML running locally? (I see that it's in the FAQ, but in the title)
I'm a huge fan of kagi and have been paying for it for as long as paying for it has been possible - that said, I think the author is spot on about the long term viability of the project considering their limited funding, limited employees, very wide (yet unproven) interests AND a leader who's maybe not so receptive to feedback.
For example I was part of the Orion beta and I left feedback in the discord that it took ~30 seconds on the then top of the line iPhone (13 Pro Max?) to load the interface which made it hard to use and I thought it was unreasonably slow and he said something like "that's not slow it's totally reasonable" and since then I decided it wasn't worth leaving any more feedback and have since left the community.
I also would prefer not to see consolidation but also I imagine it's hard to make money on this kind of product so maybe that drove the decision to merge?