HN user
Pre-training data is pre-tokenized ahead of time before being used to not waste any GPU compute.
A massive speedup like this is a nice efficiency savings on some of these data pipelines for sure.
The word you're looking for is nuclear. Everywhere.
One of the claims it asks LLMs to grade is "Artificial intelligence will cause widespread job loss among software engineers."
Yea man this benchmark is really really bad.
This has been a non issue when using proper routing libraries that push history entries on the stack properly and render routes from the top of the component tree down.
You hate BAD react SPAs that break the fundamentals of how the web works. Good ones take care to not do that.
React fundamentally doesn't cause this issue either. You can use a different framework than react or even vanilla JS and still produce the same bugs.
I love Claude but I hate waiting a minute or two for any inference to start. I hope they can get their xAI capacity online ASAP and that it helps!
Alternate title: "I did not understand the current limitations of AI and assumed it could do large software design and it generated spaghetti slop"
Yea, that's why engineers are still very important for now (until models can do this type of longer term designs and stick to them).
This whole thing has just been a huge PR stunt the whole time. Even the original leak of the blog post was just more fuel to the hype.
This is such copium for AI haters. I stopped working almost any single line of code at the beginning of this year and I've shipped 3 production projects that would have taken months or years to build by hand in a matter of days.
Except none of them are open source so they don't show up in this article's metrics.
But it's fine. Keep your head in the sand. It doesn't change the once in a lifetime shift we are currently experiencing.
With my kids we did Le Robot from hugging face over the Christmas break, it was a fun project to put together the kit and get the follower arm to follow the leader. You can also train ML models with it etc https://huggingface.co/docs/lerobot/en/so101
"Well you're all f***, good luck. I'll take my millions and go live on my micro farm"
Its top 100 or so subreddits are moderated by the same ~10 or so individuals who impose their ideological views on the subs and delete posts or ban anyone who dares challenge them.
A great example of how community moderation inevitably slides a platform to one side or the other of the political spectrum.
I honestly don't think mods on reddit should be allowed to moderate more than 1 or 2 of these top sub-reddits, this would at least force some semblance of diversity of thought on the platform.
Bud Light's stock performance last year would like to have a word with you.
This is such a US centric take.
"Sucking superintelligence through a straw"
Reddit was forced to clean it up when they started eyeballing an IPO.
I've been thinking about this and using Rust for my next backend. I think we still lack a true "all in one" web "batteries included" framework like Django or RoR for Rust.
Maybe someone should use AI to write the code for that...
Sugar free coke is not as bad as sugar-ful Coke but it's still bad. Many of the cheap sweeteners have been linked to cancer. They still fuck with the brain and hormones and make you want salty foods and/or more sweet tasting things.
So yea, how about drinking water as your primary source of hydration?
If you are poor, the last thing you need is Diabetes, Cancer, Hypertension, Cardiovascular disease, etc.
The problem also is there is a huge amount of fraud with SNAP with people claiming benefits for multiple people and then reselling their SNAP cards to just make cash. The people buying the endless cases of Mountain Dew often have just bought a 50% discounted SNAP card off some other person who isn't starving at all.
From the author: > at some point we started benchmarking on wikipedia-scale datasets. > that’s when things started feeling… slow.
So they're talking about this becoming an issue when chunking TBs of data (I assume), not your 1kb random string...
The sad thing is it doesn't have to be this way.
I worked on an internal tools team for a few years and we empowered engineers to fix user issues and do user support on internal support groups directly.
We also had PMs who helped drive long term vision and strategy who were also actively engaging directly with users.
We had a "User Research" team whose job it was to compile surveys and get broader trends, do user studies that went deep into specific areas (engineers were always invited to attend live and ask users more questions or watch raw recordings, or they could just consume the end reports).
Everyone was a team working together towards the same goal of making these tools the best for our internal audience.
It wasn't perfect and it always broke down when people wanted to become gatekeepers or this or that, or were vying for control or power over our teams or product. Thankfully our leadership over the long term tended to weed those folks out and get rid of them one way or another, so we've had a decent core group of mid-level and senior eng who have stuck around as a result for a good 3 years (a long time to keep a core group engaged and retained working on the same thing), which is great for having good institutional knowledge about how everything works...
So are photos that are edited via Photoshop not art? Are they not art if they were taken on a digital camera? What about electronic music?
You could argue all these things are not art because they used technology, just like AI music or images... no? Where does the spectrum of "true art" begin and end?
GraphQL was created to solve many different problems, not just overfetching.
These problemes at the time generally were: 1) Overfetching (yes) from the client from monolithic REST APIs, where you get the full response payload or nothing, even when you only want one field
2) The ability to define what to fetch from the CLIENT side, which is arguably much better since the client knows what it needs, the server does not until a client is actually implemented (so hard to fix with REST unless you hand-craft and manually update every single REST endpoint for every tiny feature in your app). As mobile devs were often enough not the same as backend devs at the time GraphQL was created, it made sense to empower frontend devs to define what to fetch themselves in the frontend code.
3) At the time GraphQL was invented, there was a hard pivot to NoSQL backends. A NoSQL backend typically represents things as Objects with edges between objects, not as tabular data. If your frontend language (JSON) is an object-with-nested-objects or objects-with-edges-between-objects, but your backend is tables-with-rows, there is a mismatch and a potentially expensive (at Facebook's scale) translation on the server side between the two. Modeling directly as Objects w/ relationships on the server side enables you to optimize for fetching from a NoSQL backend better.
4) GraphQL's edges/connections system (which I guess technically really belongs to Relay which optimizes really well for it) was built for infinitely-scrolling feed-style social media apps, because that's what it was optimized for (Facebook's original rewrite of their mobile apps from HTML5 to native iOS/Android coincided with the adoption of GraphQL for data fetching). Designing this type of API well is actually a hard problem and GraphQL nails it for infinitely scrolling feeds really well.
If you need traditional pagination (where you know the total row count and you want to paginate one page at a time) it's actually really annoying to use (and you should roll your own field definitions that take in page size and page number directly), but that's because it wasn't built for that.
5) The fragment system lets every UI component builder specify their own data needs, which can be merged together as one top-level query. This was important when you have hundreds of devs each making their own Facebook feed component types but you still want to ensure the app only fetches what it needs (in this regard Relay with its code generation is the best, Apollo is far behind)
There's many other optimizations we did on top of GraphQL such as sending the server query IDs instead of the full query body, etc, that really only mattered for low-end mobile network situations etc.
GraphQL is still an amazing example of good product infra API design. Its core API has hardly changed since day 1 and it is able to power pretty much any type of app.
The problems aren't with GraphQL, it's with your server infra serving GraphQL, which outside of Facebook/Meta I have yet to see anyone nail really well.
Please please please, if you have young kids learning to read or who will need to soon, educate yourself by listening to the "Sold a Story" podcast from NPR (it's on Spotify and other places).
There is so much bullshit out there about how kids should be taught to read, and too many schools unfortunately still use wrong methods disproven by science.
What works is phonics, old, tried and true. If your school isn't teaching it, you need to do it yourself at home or your kids risk never being good readers.
I used OVH years ago when "the cloud" wasn't a thing. Always super reliable. Always super pro. You love to see good competition in this type of market.
Oh great another device that has access to way too much. I skipped the "smart speaker" that listens to all your conversations phase, I will skip this one too, thank you very much.
You assume there is a symmetrical relationship between the US and other nations here. There is not, hasn't been since WW2.
Let's remember it was recently proven that there is little to no evidence that serotonin levels are linked to depression: https://www.nature.com/articles/s41380-022-01661-0
It's pretty much likely to be the same deal here.
Since there is no chemical or biological test one can do to confirm a schizophrenia diagnosis, it's a subjective diagnosis by practitioners, treating such subjective diagnosis with powerful brain-chemistry altering drugs. Worse, we do not yet know whether these new drugs will be easy to get off of. Some of the other anti-psychotic type medication is VERY hard to get off of, such as abilify, where most drug treatment centers will not deal with Abilify withdrawals or take on those patients because those are so severe (extreme violence, self harm, etc during withdrawal).
Until Psychiatry can reform itself to become a real science, using the scientific method (repeatable, provable results and not just a "theory of how brain chemistry maybe works sorta but we're not sure"), it will continue to just be a big cash cow for Big Pharma while hooking many patients who do not need these drugs onto them for life, while failing to effect ANY cure on the things they are treating (Have you heard of someone getting "better" after taking anti-depressants and being able to come off them? No. They have to take them for life. That's not a cure.)
That's because the sample size is probably small and for niche prompts or topics.
It's very hard to evaluate whether a model is better than another, especially doing it in a scientifically sound way is time consuming and hard.
This is why I find these types of comments like "model X is so much better than model Y" to be about as useful as "chocolate ice cream is so much better than vanilla"
Morals is the code of conduct which a certain group adheres to in principle, without those rules having the force of law.
Ethics is about your own conduct or the conduct of a group and whether that conduct leads to survival or improvement for yourself and others.
Justice is when a larger group sets some rules (usually laws) because it believes those rules are beneficial to the group. When a member of that larger group fails to follow these rules set by the group, the group comes down on them to force them to apply the rules set out or to punish them for not applying these rules.
Morals is rules that can be enforced or not, ethics is about doing the "right thing", justice is about enforcing the laws set out by a larger group (like a city or state or country etc)
Define "things you don't need" ? That's a very subjective qualification. Even things which may only bring joy to the end consumer but very little objective value otherwise still have the effect of raising the spirits of said consumer a little (or a lot).
Like are Magic the gathering cards "something you need"? No. Do they bring endless joy to people who enjoy playing MTG, do they help build supportive communities, etc? Probably. Is that net good for the world? Probably.
This reminds me of the rant of a friend of mine who works on movies and TV shows. Someone was telling him the entertainment industry was useless and doesn't bring value to the world. He replied "cool so if we took away all your streaming platforms, all your music and all your books you'd be totally cool with that?" . Obviously the answer was no.