cma

7 ms·

91.
by cma 3mo ago
Interestingly: "For logic proof, it had been thought that AI could handle all logic problems in the near future, hence logic problems of solving conjectures might not be so interesting in the future."
92.
by cma 3mo ago
Someone could have a small private breakthrough tomorrow that gives sample learning efficiency of the brain, online learning, and consolidated memories.
93.
by cma 3mo ago
It could be as simple as do open releases and publication at first to help recruit talent who want that or who want to make a name for themselves. learned from the likes of... OpenAI, Google, Meta, Emad
94.
by cma 3mo ago
It's TBP's fastest growing brand, 50/50 joint venture with Carlson. Mostly sold online, not in stores, lending towards podcast promotion over seeing in a convenience store. I'm not talking about the whole nicotine pouc
95.
by cma 3mo ago
Maybe so, OP mentioned they were very similar and that still seems to hold for stainless. How about the aggregates within? The composite makeup can result in different CoTE than the individual aggregates, I think that's one part of co
96.
by cma 3mo ago
I was just responding to "The harness is what takes these random and hallucinogenic models and make them into something deterministic and useful." You can compare Fable vs Sol vs Kimi in the same harness if you want too and there
97.
by cma 3mo ago
You can run the same harness on fable, opus, sonnet, and see a huge difference between them. It is true the harness is important, and openai has begun encryption its instructions to swarmed sub-agents instead of just encrypting the chain o
98.
by cma 3mo ago
There's also stainless steel rebar
99.
by cma 3mo ago
> Nasdaq 100 has always been marketed as a tech-forward index. It would be a bit ridiculous if they didn’t include the most value tech companies on the market. What if they floated only .01%? What's the cutoff for it being ridiculo
100.
by cma 3mo ago
One example would be go to college/graduate school to invest in skills or get a job now to earn wealth over the same time period.
101.
Android Developer Verifier(old.reddit.com)
3 points by cma 3mo ago|0 comments
102.
by cma 3mo ago
They tested them on formal languages of different power, and saw where they could generalize beyond the training data. Transformers failed to generalize pretty early on at stack machines/brace matching. It was a good bit older of a pa
103.
by cma 3mo ago
> considering that any nicotine product very much does have a cancer risk. Source? The carcinogens are from other parts of tobacco than nicotine. Pouches are newer, but nicotine gum and patches have been studied for years.
104.
by cma 3mo ago
Lots of podcasts advertise them, I think Tucker Carlson, one of the biggest podcasts in the world, has his own brand or promotion deal. I don't know if he is popular with young adults or still has his fox news demographic. It's m
105.
by cma 3mo ago
Not exactly the spam example, but Ireland remained a net exporter of food during the potato famine. The good farmland went to cattle and things like that for export.
106.
by cma 3mo ago
This deepmind paper showed big limitations on nested structures with transformers, LSTMs actually did better: Neural Networks and the Chomsky Hierarchy https://arxiv.org/abs/2207.02098
107.
by cma 3mo ago
But he's also using AI for formally verified math and for ideas in solving math problems. The part about it being ok because it is a supplement just means ok that these aren't formally verified and may have bugs, and may also mea
108.
by cma 3mo ago
Compare Claude at the time to 4o's sycophantic era (when o4 updated in response to Grok), Claude was lightyears better.
109.
by cma 3mo ago
One of the big things that enabled extended context to work was training techniques for extending LLMs with reasoning, part of the thing he was saying wouldn't work.
110.
by cma 3mo ago
> Maybe the new tattoos are just like being racist or something, but that’s hard to do when your heart isn’t in it and they will eventually find some way to absorb that. I think I'd rather read slop than edgelord.
111.
by cma 3mo ago
> What exactly was he dead wrong about that is proven by any of this? He said as you need more and more tokens models will fall apart because each additional token is a chance for a mistake and they will just exponentially fall apart. B
112.
by cma 3mo ago
The scaling with reasoning models is more and more with things like verifiable rewards (coding and math), in line with bitter lesson and also Sutton invented lots of modern RL.
113.
by cma 3mo ago
Why would auto mode turn off thinking?
114.
by cma 3mo ago
Were your question in math areas related to ML? They also restrict model development and research pretty heavily.
115.
by cma 3mo ago
Are you running unquantized GLM-5.2 and getting in loops or quantized?
116.
by cma 3mo ago
By bunch up in a ball I just mean assume the fetal position when powered off, not wheels. They wouldn't have to take up more space in house than a large piece of travel luggage on a shelf.
117.
by cma 3mo ago
Probably related to things like "The Nazis utilized data from routine censuses, tax returns, and municipal police registrations. In Germany, and in occupied countries like the Netherlands, this information was systematically organized.
118.
by cma 3mo ago
Ratified treaties are the supreme law of the land and he's pointing out "all lawful uses" isn't what the admin says it is.
119.
by cma 3mo ago
It doesn't need to have a sleeping surface or stay sprawled out when it isn't in use. It could bunch up into a little ball and fit in the corner of your ceiling. And it doesn't need to be the size of an adult to do most hou
120.
Google's new CAPTCHA asks users to wave hand at camera(cybernews.com)
1 points by cma 3mo ago|2 comments

More ›