hmottestad

7 ms·

61.
by hmottestad 2y ago
Kagi uses R1 through Fireworks.ai, Together.ai and Groq. https://help.kagi.com/kagi/ai/llms-privacy.html
62.
by hmottestad 2y ago
I tried R1 through Kagi and it’s similarly censored. Even the distill of llama running on Groq is censored.
63.
by hmottestad 2y ago
I tried to ask it about Java exploits that would allow me to gain RCE, but it refused just as most western models do. That was the only thing I could think to ask really. Do you have a better example maybe?
64.
by hmottestad 2y ago
I tried the R1 distill of llama 8B, which did refuse direct questions about the massacre. Haven’t tried this new model locally, but I agree with you that it looks like there is a secondary censorship going on. If I ask it to list the 10 wor
65.
by hmottestad 2y ago
Just don’t ask it about the tiananmen square massacre or you’ll get a security warning. Even if you rephrase it. It’ll happily talk about Bloody Sunday. Probably a great model, but it worries me that it has such restrictions. Sure OpenAI al
66.
by hmottestad 2y ago
I remember pushing the R1 distill of llama 8B to see what limits had been put in place. It wasn’t too happy to discuss the 1989 Tiananmen Square protests and massacre, but if I first primed it by asking about 9/11 it seemed to veer mor
67.
by hmottestad 2y ago
You can now safely use virtual threads to run code that uses "synchronized".
68.
by hmottestad 2y ago
Anyone know any good infinite canvas app for the iPad? I’ve used Goodnotes for years, but it doesn’t have an infinite canvas unfortunately. Otherwise it’s an amazing app and nothing really comes close. I’ve been trying Apple Freeform as an
69.
by hmottestad 2y ago
Life expectancy in the US is behind 47 countries according to https://www.worldometers.info/demographics/life-expectancy/ Infant mortality rate is behind 53 countries: https://www.cia.gov/the-world
70.
by hmottestad 2y ago
The MoE architecture allows you to keep the entire active model on a single GPU. If two consecutive tokens use the same export then the second token is going to be much faster.
71.
by hmottestad 2y ago
If the same expert is chosen for two consecutive tokens then it’ll act like a 37B model running on the GPU for the second token since it doesn’t need to load that expert from the main RAM again.
72.
by hmottestad 2y ago
That’s probably the main performance benefit of using the GPU. If you’re changing the active expert for every single token then it wouldn’t be any faster than just running it on the CPU. Once you can reuse the active expert for two tokens y
73.
by hmottestad 2y ago
If you’re running just one GPU your context is limited to 1024 tokens, as far as I could tell. I couldn’t see what the context size is for more cards though.
74.
by hmottestad 2y ago
iMessage gets synced automatically, but SMS needs to be enabled explicitly on the phone for it to forward them to your Mac.
75.
by hmottestad 2y ago
Are you the CISO?
76.
by hmottestad 2y ago
Do you run a paid bug bounty program? I saw an interesting presentation from Finn.no about how they got most of their vulnerabilities through that, basically none from their security.txt file, and a handful from people contacting the CISO o
77.
by hmottestad 2y ago
Yeah. I've found out that you can start out with o3-mini-high and then switch over to o1-pro, or the other way around. Helps to iterate a bit faster.
78.
by hmottestad 2y ago
For me I was sold on the consistency. o1 does really great at several programming problems, but o1-pro does great on these problems 4 of 4 tries. I get a good answer more often with o1-pro than with just o1, or even o3-mini-high.
79.
by hmottestad 2y ago
Funnily enough a lot of the open source world has landed on an API that is basically a copy of OpenAI. So if you develop against OpenAI it’s almost a slot in solution to switch to an open source solution.
80.
by hmottestad 2y ago
Would be interesting to know if dragonflies get red eyes when you take a picture of them with flash. Human eyes reflect some of the red light from the white flash, making it look like the pupil is red instead of black.
81.
by hmottestad 2y ago
It seems to be available in Norway now.
82.
by hmottestad 2y ago
OpenAI Deep Research isn’t available in Norway at least, or the rest of Europe basically :(
83.
by hmottestad 2y ago
I don't think there are all that many MRI machines that just sit there, unused, most of the day. There might be some hospitals that want to reserve their MRI machine for emergencies, so that it's less likely that you'll have
84.
by hmottestad 2y ago
I think the question was if Deep Research is available from the API or not. As far as I know it’s only available through ChatGPT and not the API. Same for o1 Pro Mode. Anyone using ChatGPT is paying a flat rate and not paying per token. Tho
85.
by hmottestad 2y ago
It wasn’t really until GPT-3.5 came out as ChatGPT that the world really started paying attention to OpenAI. I guess the author must have been one of the very few that had actually experimented with GPT-3 before the whole ChatGPT era.
86.
by hmottestad 2y ago
I still don't know how to quit Vim without googling for instructions :P As an anecdote from my time at uni, I can share that all our exams were either writing code with pen on paper for 3-4 hours, or a take-home exam that would make up
87.
by hmottestad 2y ago
You can stop it at any time, then modify what it's written so far...then press continue and let it continue thinking and answering.
88.
by hmottestad 2y ago
Yeah, that's my point. If they do end up surpassing OpenAI then it would seem likely that they aren't just relying on copying from o1, or whatever model is the frontier model at that time.
89.
by hmottestad 2y ago
At the pace that DeepSeek is developing we should expect them to surpass OpenAI in not that long. The big question really is, are we doing it wrong, could we have created o1 for a fraction of the price. Will o4 cost less to train than o1 di
90.
by hmottestad 2y ago
I don't use it much for programming, since I mostly use IntelliJ for that. Where Sublime shines for me is opening large files, and opening pretty much any file. So many times MacOS wants me to open a json file in Xcode, or a txt file i

More ›