chaxor

42 ms·

541.
by chaxor 3y ago
A newer but much better system actually reduces the model size while reducing the functionality of the system - similar to training a NN for a very specific task (as was typical several years ago), but now it can happen with far less data
542.
by chaxor 3y ago
This title is hilarious. S is a subset of U. Also headline: |S| > |U|. Incredible. But I suppose this pedantic point doesn't matter since HN hates cryptocurrency so much they lose their minds and start hating cryptography just out o
543.
by chaxor 3y ago
There's nothing really wrong with a preprint making it to the top - there can be genuinely good work that stays in preprint for quite some time. I believe the original ELMo work that spurred the Sesame street gang is still in preprin
544.
by chaxor 3y ago
"OO features appeared to be add-on" No one tell them about the horrors of R. That's a 'yikes' of a language that no one should ever use.
545.
by chaxor 3y ago
Thanks, this is really cool. I found one from gist.github.com/JellyWX/ called jude.goggles that is basically already that.
546.
by chaxor 3y ago
Neural networks were known as being "always second best" for many decades not that long ago... Its very possible ( highly likely even) that elements of gofai end up being implemented into some of the upcoming RL/GNN combo ba
547.
by chaxor 3y ago
The AI pronounced `sudo` as "su-dough" instead of "su-doo". Literally unwatchable :P This is fantastic work, keep up the great job.
548.
by chaxor 3y ago
It's good - but the even more crazy thing is that this can be done by a script kiddy in a few hours - not an expert who spends months or years trying to whittle at some part of the process as it was several years ago.
549.
by chaxor 3y ago
Wow, if 4 degrees out doesn't give 70k nodes it's extended to 7 degrees out? I figured almost any website would yield a near complete graph at just ~6 degrees out. This is a fantastic idea though. I have been looking for something
550.
by chaxor 3y ago
Base wireguard is pretty easy to setup, especially with wg-quick, so idk why anything would be required to make it easier. Also, Rosenpass is quite great and easy to use, which really improves the security further. Hopefully Rosenpass will
551.
by chaxor 3y ago
The models now are probably closer to GPT-2, and gets very close to GPT-3 with decent hardware (larger param models) and effort. They're certainly worse when it comes to 'knowledge' or memorized parts of the training data, bu
552.
by chaxor 3y ago
The most important aspect of langchain is NOT using OpenAI for the LM. The most useful aspect of using langchain is to use it with Galpaca (or vicuna/koala/etc) to spin up an assistant for your home. This way, you can push all
553.
by chaxor 3y ago
Found the js programmer
554.
by chaxor 3y ago
"LLMs are Markov Models"... There's a reason we don't use Markov models anymore though (for language modeling - still used heavily in RL). The density of information and composability you get from MHA makes the proble
555.
by chaxor 3y ago
It's quite fast if you use it at ~4 AM in the US. There's definitely a cycle in time. Putting things in a queue to run while you sleep is a good work around.
556.
by chaxor 3y ago
Of course... that's absolutely true, and I didn't say otherwise. However, the basic idea actually is possible, but only for a very select few of the tests they proposed (which they still unfortunately didn't really get right
557.
by chaxor 3y ago
"the tests they were proposing" - yeah, I would agree that some (probably most towards the end) of the tests they were trying to make part of the product were a big part of why they were so fraudulent. I agree that it was a bad
558.
by chaxor 3y ago
I'm not sure noticing tokens/params or simplicial modeling properties occur requires much critical thought - perhaps it's just a standard first thought for anyone given an LM now. I've worked tangentially to NLP for ab
559.
by chaxor 3y ago
I highly doubt many people will be using models that have this problem in 6 months - and likely people wanting to write scripts will be using local models - NOT OPENAI . The future is much brighter running loca models tailored to the users
560.
by chaxor 3y ago
IIRC, you can put entire binaries in QRCodes, and some have shown this by putting the game 'snake' in a QRCodes, such that if you point your camera towards the QRCodes with the camera on, it will start the game automatically. Fun
561.
by chaxor 3y ago
One important but that is often left out is that ChatGPT is not the first model to come out using RLHF to train LLMs. As is typical in the AI field, Deepmind was key in the development of the process. Deepmind 's Sparrow came out just
562.
by chaxor 3y ago
I don't think it's a business friendly license?
563.
by chaxor 3y ago
This is a bit hard to believe that the system is decent at producing code which captures complex ideas and higher level structure when the tokens/param value is >30 (it's ~200 here? ) The 'good' models (meaning having
564.
by chaxor 3y ago
Depends about on what you mean here. The good scientist mostly give up when trying to 'describe' higher level function (at best maybe generate some analytical expression which may correlate with IQ or something) and focus instead
565.
by chaxor 3y ago
I think most segments against online password companies is that they get hacked so often. The most practical problem is having to switch all your passwords around after such a leak occurs, which seems to be more and more permanent these da
566.
by chaxor 4y ago
If it's a map to the hidden treasure (as in the exact bits of the image of the map) then it's kind of the same thing right? But yes, the non-uniqueness of expressing similar ideas can make things difficult there.
567.
by chaxor 4y ago
It wasn't a viral test, and I didn't lay out details here because I don't want to put all my health information on display exactly, but I have several graduate degrees in both engineering and in biology related fields, so I&#
568.
by chaxor 4y ago
Theranos wasn't a bad idea, and is actually possible. It is extremely unfortunate that it was so badly mismanaged.
569.
by chaxor 4y ago
The last test someone I know had done cost $600 at the physician. We looked it up and it cost a little less to *buy the FDA approved equipment for your house, and run the test every day for an entire month*. It's absolutely heinously
570.
by chaxor 4y ago
I think we can agree that both of those cases aren't worst case scenarios, but rather the only case, or best case scenarios (in the US). The only real way of dealing with health (in the US) is just to never go to a physician and t

More ›