kadushka

11 ms·

151.
by kadushka 1y ago
Let’s forcibly take all the billionaires’ wealth, and distribute it equally among the working class people. Oh wait…
152.
by kadushka 1y ago
People do this today - all over the world, only with missiles and automatic rifles.
153.
by kadushka 1y ago
no one is less of a citizen than anyone else This is not true - a citizen by birth can become the president, a naturalized citizen cannot.
154.
by kadushka 1y ago
If you get any accuracy degradation with full 8 bits of precision you're doing it wrong.
155.
by kadushka 1y ago
What's rapidly changing are quantization algorithms, and hardware features to support those algorithms. For example, Blackwell GPUs support dynamic FP4 quantization with group size 16. At that group size it's close to lossless (in
156.
by kadushka 1y ago
Invite candidates to spend 2-3 days working together on some task - as a short trial. See how much they accomplish in that time, and how easy it is to work with them. Does not have to be onsite. Pay them well for their time.
157.
by kadushka 1y ago
If you’re not offering an equal share in the company, then you shouldn’t mention the word cofounder – it’s misleading. You are looking for a regular employee, so evaluate candidates to make sure they’re capable of doing the task you’ll be p
158.
by kadushka 1y ago
Are you looking for a cofounder or an employee? It’s like to be looking for a secretary who will ideally grow into a wife.
159.
by kadushka 1y ago
Catastrophic forgetting would degrade factual knowledge benchmark results - even more than hallucination benchmarks, right? Do we observe this with o3/o4-mini? If not, your hypothesis is invalidated.
160.
by kadushka 1y ago
RL might be making hallucinations worse, that’s true. Why do you think RL is causing catastrophic forgetting? Are there factual knowledge benchmarks showing it for o3 or o4-mini?
161.
by kadushka 1y ago
Hallucinations usually happen when a model never knew the answer, not when it forgot something.
162.
by kadushka 1y ago
Google, last month: “we’re doubling down on our partnership with NVIDIA” https://blog.google/technology/ai/google-nvidia-gtc-ai/
163.
by kadushka 1y ago
Seems like the opposite to me: https://www.datacenterdynamics.com/en/news/google-in-talks-t...
164.
by kadushka 1y ago
it does not "get better" merely by using the existing trainsets with more compute. Pretty sure it does - that’s the whole point of using more test time compute. Also, a lot of research efforts goes into improving data efficiency
165.
by kadushka 1y ago
I also got a quote from Tesla to get solar shingles roof - $110k.
166.
by kadushka 1y ago
Major roof replacements cost 5 to 15k for average homes I just got some quotes to replace a tile roof on my very average 2400 sqft house in FL: 50-60k. Asked neighbors - seems reasonable to them.
167.
by kadushka 1y ago
This is a Codeflash ad.
168.
by kadushka 1y ago
LLMs do about half of my work for me. Today I spent most of my time interacting with o3 and 2.5-pro and I have accomplished what would take me 3 days to accomplish in the past. That’s real impact.
169.
by kadushka 1y ago
I remember how ~5 years ago I said - here on HN - that AI will pass TT within 2 years. I was downvoted into oblivion. People said I was delusional and that it won’t happen in their lifetime.
170.
by kadushka 2y ago
I'm using "enshittification" as defined here: https://www.merriam-webster.com/slang/enshittification I believe there's a real chance of it happening here as a result of this transition. I personally
171.
by kadushka 2y ago
The people on both ends of that conversation (google vs cornell) are clever but the result will probably be enshittification.
172.
by kadushka 2y ago
What does intelligence have to do with having desires or goals? An amoeba can do stuff on its own but it's not intelligent. I can imagine a god-like intelligence that is a billion times smarter and more capable than any human in every
173.
by kadushka 2y ago
My phd thesis, just like 99% of other phd theses, does not have any “truly novel” ideas.
174.
by kadushka 2y ago
Something new != truly novel. ChatGPT creates something new every time I ask it a question.
175.
by kadushka 2y ago
How many people do you know who are capable of doing something truly novel? Definitely not me, I'm just an average phd doing average research.
176.
by kadushka 2y ago
I learned the definition I provided back in mid 90s, and it hasn't really changed since then.
177.
by kadushka 2y ago
Pretty much everyone in high school or college is using them. Also everyone whose job is to produce some kind of content or data analysis. That's already a lot of people.
178.
by kadushka 2y ago
Another general consensus is that humans possess general intelligence.
179.
by kadushka 2y ago
Yes, a general consensus is AGI should be able to perform any task an average human is able to perform. Definitely nothing of Nobel prize level.
180.
by kadushka 2y ago
ChatGPT has 400M weekly users. https://backlinko.com/chatgpt-stats

More ›