Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
MacsHeadroom
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
9 ms
·
31.
▲
by
MacsHeadroom
2y ago
That's the purpose of the "public training set." You don't take an exam before reviewing the instructions on how to fill out the answer sheet.
32.
▲
by
MacsHeadroom
2y ago
It's easily closer to 90% than 10%.
33.
▲
by
MacsHeadroom
2y ago
The Chief Compliance Officer of Circle, makers of the USDC stablecoin, sits on the board of the Illicit Virtual Asset Network alongside U.S. government representatives from the FBI, DEA, etc. https://www.businesswire.com/ne
34.
▲
by
MacsHeadroom
2y ago
They aren't Open Source, so I would not expect them to be supported. The point of this project appears to be to keep full control of your intelligence pipeline.
35.
▲
by
MacsHeadroom
2y ago
BSD or MIT license would be nice.
36.
▲
by
MacsHeadroom
2y ago
DeepSeek is a scrapy tech startup fully funded by a successful hedge fund owner without any cooperation with the government. Meanwhile OpenAI's board includes the most recent head of the NSA and Altman just announced he'll be in D
37.
▲
by
MacsHeadroom
2y ago
The people he "tried to kill" were fabrications of two FBI agents in an entrapment scheme who both ended up in prison shortly after Ross.
38.
▲
by
MacsHeadroom
2y ago
It's a tiny island nation with a single sea port, single bridge, and single airport. Meanwhile western nations are so porous they can't keep millions of undocumented people out.
39.
▲
by
MacsHeadroom
2y ago
This is real. Female engineers are overrepresented in big tech something like 3-4x the graduation rate. There just aren't any left over for startups that can't afford FAANG rates.
40.
▲
by
MacsHeadroom
2y ago
lmsys is a poor judge of coding quality since it is based on ratings from a single generation rather than agentic coding over multiple steps.
41.
▲
by
MacsHeadroom
2y ago
DeepSeek is open source and has a context length of 128k tokens.
42.
▲
by
MacsHeadroom
2y ago
This sounds like an argument in favor of AI personhood, not an argument against AI experts.
43.
▲
by
MacsHeadroom
2y ago
Paying local market rates is not exploitative.
44.
▲
by
MacsHeadroom
2y ago
A cap of less than 1% of an average developer's pay is "really expensive"?
45.
▲
by
MacsHeadroom
2y ago
Estimates from earlier this year are over 60 Million Linux powered desktop PCs globally. That's not a huge portion of total market share but is still major by some measure.
46.
▲
by
MacsHeadroom
2y ago
> A generalizable strategy for synthesizing LLMs differentiated by their training parameters is a tokenization is isolating data sets and then establishing a lattice in uniformity within the field of technics. This comment appears to be
47.
▲
by
MacsHeadroom
2y ago
The system as a whole has reasoned twice over, verbally and then logically.
48.
▲
by
MacsHeadroom
2y ago
Prediction markets like Polymarket solve this by not having bookies. Just bet against other players. Why bet against a house who takes an enormous cut even when you win when you could bet in an efficient market and keep 100% of your winning
49.
▲
by
MacsHeadroom
2y ago
As discussed elsewhere in this thread, butterflies retain some memories formed as caterpillars after every cell in their body changes drastically. They never die but they change immensely and still retain memories, so I'm not sure we k
50.
▲
by
MacsHeadroom
2y ago
This paper does not support your position any more than it supports the position that the problem is tokenization. This paper posits that if the authors intuition was true then they would find certain empirical results. ie. "If A then
51.
▲
by
MacsHeadroom
2y ago
~50¢
52.
▲
by
MacsHeadroom
2y ago
GenAI costs go down 95% per year. So next year it will be $3.40/hr and more reliable.
53.
▲
by
MacsHeadroom
2y ago
Wow Tether transactions are half as likely to be involved in illicit activities vs fiat transactions? Who knew! Thanks for sharing.
54.
▲
by
MacsHeadroom
2y ago
I see a whitespace expert, a punctuation expert, and a first word expert. It's interesting to see how the experts specialize.
55.
▲
by
MacsHeadroom
2y ago
Yes, this is how o1 was trained. Math and programming, because they are verifiable. This is also why o1 is not better at English. Math skills transfer to general reasoning but not so much to creative writing.
56.
▲
by
MacsHeadroom
2y ago
Our constitution established a republic, not a democracy. The drafters famously despised democracy. https://founders.archives.gov/documents/Adams/99-02-02-6371
57.
▲
by
MacsHeadroom
2y ago
How are you going to prevent fraudulent double spending of these digital coins? You'd have to invent some kind of Byzantine Fault Tolerant (BFT) consensus method.
58.
▲
by
MacsHeadroom
2y ago
OpenAI defines AGI as "autonomous systems that outperform humans at most economically valuable work." ChatGPT is not autonomous or capable of doubling global GDP.
59.
▲
by
MacsHeadroom
2y ago
o1 is about $2-$4 per message over the API. I'm probably costing OpenAI less than 24hrs after my subscription renewal each month. Voice mode is around $0.25 per minute via API. I don't use that much, but 3 minutes ago per day woul
60.
▲
by
MacsHeadroom
2y ago
Autonomy and rule following are at odds. Humans have the same problem. The solutions we use for ourselves work amazingly for LLMs (because they're trained on human data). Examples: Give an LLM an effective identity (prompt engineering)
More ›