stri8ted

11 ms·

1.
by stri8ted 13d ago
I suspect the only reason those alternative providers have better up-time and more generous quotas, is because they don't have nearly the same amount of demand. Notice that Deepseek recently had to increase their pricing, once it gaine
2.
by stri8ted 13d ago
> instead the queue of people in front of it will increase. This can ultimately result in the system breaking. I don't think Anthropic engineers are that much worse than their peers, such that they are 10x more prone to causing outa
3.
by stri8ted 13d ago
Reduced weekly limits, higher prices than competitors, paying well above spot price for space-x compute, all point to supply issues.
4.
by stri8ted 13d ago
Demand > supply. It's impressive that customers have not migrated en masse to other providers, given the frequency of these outages. Perhaps switching costs are greater than some would believe. Or, qualitative differences between mo
5.
by stri8ted 2mo ago
Translation, moderation, classification, guardrails, etc..
6.
by stri8ted 2mo ago
This is brutal, especially for small transactions. Note each failed attempt is another screening fee. Unfortunately, there is no good competition.
7.
by stri8ted 2mo ago
Demand > supply
8.
by stri8ted 2mo ago
Same
9.
by stri8ted 2mo ago
By what measure is there an overbuild? Every metric I look at, shows inference unable to satisfy current demands.
10.
by stri8ted 2mo ago
20k is small potatoes for the marketing impact.
11.
by stri8ted 2mo ago
They are doing both. Distilling Mythos down to affordable models, so they can continue to fund the business. And training Mythos level models at the high-end, to expand the frontier.
12.
by stri8ted 2mo ago
You seem to assume training on pelican would not result in improved performance on other similar tasks. Why?
13.
by stri8ted 2mo ago
Still has a 1.78T market-cap. It will be OK.
14.
by stri8ted 2mo ago
Its actually useful, when you are launching into a long monologue and want periodic acknowledgement that its "listening".
15.
by stri8ted 3mo ago
Way too expensive. Google vision OCR (which they failed to compare against), is $1.50 per 1k pages. Vs $4 from Mistral.
16.
by stri8ted 3mo ago
What metrics are you using to classify it as a downfall?
17.
by stri8ted 3mo ago
Is there something in the post that you find implausible or don't believe to be true?
18.
by stri8ted 5mo ago
I doubt this is representative of real world usage. There is a difference between a few turns on a web chatbot, vs many-turn cli usage on a real project.
19.
by stri8ted 5mo ago
Entire segments of the podcast sphere are making their money talking about these so-called unspeakable subjects. Why don't you share what you really think.
20.
by stri8ted 5mo ago
Those are the same thing
21.
by stri8ted 6mo ago
48 GB is not consumer hardware. But fundamentally, there are economies of scale due to batching, power distribution, better utilization etc.., that means data center tokens will be cheaper. Also, as the cost of training (frontier) models in
22.
by stri8ted 6mo ago
Do you have any evidence to support this view?
23.
by stri8ted 7mo ago
Price Input: $2.50 / 1M tokens Cached input: $0.25 / 1M tokens Output: $15.00 / 1M tokens https://openai.com/api/pricing/
24.
by stri8ted 7mo ago
Flock has put out a report claiming 10% crime in the US is solved using their technology. There are of course counter argument, that claim this is not valid. https://www.flocksafety.com/customers/how-many-crimes-do-aut.
25.
by stri8ted 7mo ago
I take into account publicly available information (news articles), factor in personal anecdotes, and reason about human nature and incentives. I know the extent of reported abuses, and I do my best to extrapolate. It's not perfect, bu
26.
by stri8ted 7mo ago
It's clearly true there have been abuses as a result of this technology. And its also clearly true criminals have been caught as a result of the cams, that otherwise would not have been. If you believe the costs of the the abuses, and
27.
by stri8ted 7mo ago
Can you show some comparisons for WER and other ASR models? Especially for non english.
28.
by stri8ted 7mo ago
How is this content related to HN? Are there any submission criteria?
29.
by stri8ted 7mo ago
Exactly. As far as I'm concerned, the benchmark is useless. It's way too easy and rewarding to train on it.
30.
by stri8ted 7mo ago
That is where the money is.

More ›