riku_iki

25 ms·

541.
by riku_iki 2y ago
> Is there a point to saying something like this other than to be explicitly uncompassionate towards those affected? those "affected" have the same little compassion towards many millions of others who are not lucky to work on
542.
by riku_iki 2y ago
> You'll still need the best of the best undergrads at the very least to solve it. Ok, so I hope you admit that OAI could manually solve them now? > Open AI didn't have any hand in providing problem And you know this exactly
543.
by riku_iki 2y ago
> So the LLM is already doing what the author says they would be surprised about. that's if you unconditionally believe in result without any proofreading, confirmation, reproducability and even barely any details (we are given only
544.
by riku_iki 2y ago
> Frontier Math isn't something that can be fully solved without gathering top experts Tao's quote above referred on hardest 20% problems, they have 3 levels of difficulty, presumably first level is much easier. Also, as I ment
545.
by riku_iki 2y ago
> Again, Frontier Math is private. its private for outsiders, but it was developed in "collaboration" with OAI, and GPT was tested in the past on it, so they have it in logs somewhere. > If you don't want to take the be
546.
by riku_iki 2y ago
> Open ai haven't released any benchmark scores there are multiple research results demonstrating that various benchmarks are heavily leaked to GPT training data. Is it intentionally or not, we can't figure out, but they have v
547.
by riku_iki 2y ago
> On frontier Math that result is not verifiable, not reproducable, unknown if it was leaked and how it was measured. Its kinda hype science. > ARC wasn't a special fine-tune but rather some of the training set in the corpus for
548.
by riku_iki 2y ago
If you are talking about ARC benchmark, then o3-low doesn't look that special if you take into account there are plenty of finetuned models with much smaller resources achieved 40-50% results on private set (not semi-private like o3-lo
549.
by riku_iki 2y ago
And I suspect o3 is something like monte carlo: generates tons of CoTs, with most of them are junk, but some hit the answer.
550.
by riku_iki 2y ago
there are many comments in internet about this, that only subset of frontier math benchmark is "field medal level research", and o3 likely scored on easier subset. Also, all that stuff is shady in the way that it is just numbers f
551.
by riku_iki 2y ago
I referred on high compute mode. They have table with breakdown here: https://arcprize.org/blog/oai-o3-pub-breakthrough
552.
by riku_iki 2y ago
> ~=$3400 per single task report says it is $17 per task, and $6k for whole dataset of 400 tasks.
553.
by riku_iki 2y ago
you don't need to deal with Maven if Java standard library(which is huge) is enough for you. Also, not sure why such big resistance against maven. Other pointers are either opinionated or minor. From another hand, if dev has experience
554.
by riku_iki 2y ago
> but I think smaller things like one-off scripts have been firmly out of reach so far. why specifically?
555.
by riku_iki 2y ago
I think some business SPA is heavily benefited from JS/TS too.
556.
by riku_iki 2y ago
> than the old "public static void main(String[] args)" thing you forgot surrounding class definition.
557.
by riku_iki 2y ago
JS/TS primary target is running interactive code in webbrowser. What are the good alternatives?
558.
by riku_iki 2y ago
is it qualified tax deduction for corp?
559.
by riku_iki 2y ago
in my testing, clickhouse dies on various queries with OOM on large datasets, because algos are subefficient.
560.
by riku_iki 2y ago
Brew runs on linux too..
561.
by riku_iki 2y ago
why?..
562.
by riku_iki 2y ago
> This kind of thinking is like those who think deporting all illegal immigrants will not somehow significantly harm the US economy. I do not support what prev commenter said, but also I don't think you can mix these things, it is n
563.
by riku_iki 2y ago
I wouldn't trust survey, chance that casual spotify listener bother to fill it is very low. p.s.: big and loyal fan of your show!
564.
by riku_iki 2y ago
he says in article that it is largest source of listeners, but I am not sure how they track it.
565.
by riku_iki 2y ago
successful quantum transfer experiment.
566.
by riku_iki 2y ago
annual perf cycle probably
567.
by riku_iki 2y ago
> Add to this that propositional logic (the language in which we express SAT) is a versatile language to code problems in is it though? You can't express some basic loop in propositional logic, right?
568.
by riku_iki 2y ago
> it's literally the cost of a cup per coffee per day So, AI market is capped by Starbucks revenue/valuation.
569.
by riku_iki 2y ago
You don't have to live in Bay Area to run startup?..
570.
by riku_iki 2y ago
> Every time I try to use Nomad, or any of the other "simpler" solutions, I hit a wall - there turns out to be a critical feature that is not available Maybe you could give example of feature in case of nomad?

More ›