oezi

18 ms·

301.
by oezi 1y ago
> The researchers compiled data on 662 energy infrastructure projects covering a diverse spectrum of technology classes and capacities, built between 1936 and 2024 across 83 countries, representing $1.358 trillion in investment. This is
302.
by oezi 1y ago
As someone new to Python: what was ever the appeal of conda that uv doesn't satisfy?
303.
by oezi 1y ago
Pensions aren't a Ponzi scheme. They just need to be calibrated that the years of pay-in and years of pay-out are balanced as to the ratio of the people in the age cohorts. If you have 10 people paying in for 40 years and only 2 people
304.
by oezi 1y ago
> We did not improve people's lifespans Are you sure about that? The statistic I always heard is that we have seen a 1 year rise in life expectancy every 4 years for the past 80 years. I don't necessarily agree with the Ponzi s
305.
by oezi 1y ago
Retirement and old age are expensive due to the cost of medical care and nursing care. If people wouldn't need those then it would be possible to retire much earlier.
306.
by oezi 1y ago
How do we call the models beyond extra large which are so big they can't be served publicly because their inference cost is too high? Do such exist?
307.
by oezi 1y ago
If it runs on 4090, it also runs on 3090 which are available used for 600 EUR.
308.
by oezi 1y ago
I think it is less than 20% of world population are native speakers and will receive audio messages in those languages.
309.
by oezi 1y ago
For transcribing WhatsApp audio messages?
310.
by oezi 1y ago
In some markets this might be right but in others it isn't. For instance, if you have CO2 certificates associated with a product then not buying it won't change emissions. It will make the price of certificates cheaper for everyon
311.
by oezi 1y ago
English, Spanish, Portuguese, Russian only
312.
by oezi 1y ago
> Price is 3-4 times higher but it lasts 5-10 times longer The problem is that if an Ikea furniture lasts 15 years then that's enough. People no longer need 50-100 years furniture, because nobody wants to inherit old stuff anymore.
313.
by oezi 1y ago
I have only tested Qwen2.5-Omni for audio and it was hit and miss for my use case of tagging audio.
314.
by oezi 1y ago
I wish multimodal would imply text, image and audio (+potentially video). If a model supports only image generation or image analysis, vision model seems the more appropriate term. We should aim to distinguish multimodal modals such as Qw
315.
by oezi 1y ago
Yes. I would envision that it is at least 5 years of such updates fixes and another 5 years available for purchase capped at 20% of device price. All manufacturers must pay an annual fee to an insurance scheme which covers the case of insol
316.
by oezi 1y ago
The problem is just one of legislation of liability. Car manufacturers are ordered to recall and fix their cars, but software/hardware companies face just too little pressure. I think customers should be able to get full refund for bro
317.
by oezi 1y ago
Absolutely right. Customers benefit most if there is a working market and monopolies and other kinds of capture are restricted. When I compare the cost cutting happening in my supermarket over pennies and then compare it to the exuberant pr
318.
by oezi 1y ago
You can read up on the advertising industry easily, right? For consumers ads are a net negative. They inflate the cost of the product and favor incumbents and monopolists to retain that position. Ads certainly work on average and Google has
319.
by oezi 1y ago
Pricing you can only achieve because you are the only seller.
320.
by oezi 1y ago
What is baffling to me is why Copilot's chat mode is still so bad. It allows you to use all the good models including Gemini Pro but still feels inferior to pasting to AI Studio. Maybe the logic for gathering the context is wrong, the
321.
by oezi 1y ago
> is effectively free right now using my own API key for Gemini. I think the cost is/was a big factor preventing many from switching from Chat-based coding to Cline/Cursor/Aider. For my hobby projects I don't want to
322.
by oezi 1y ago
Does CommonVoice have any kind of tagging related to the accent/skill of the speaker?
323.
by oezi 1y ago
Did you publish that accent dataset somewhere?
324.
by oezi 1y ago
Paraket is english only. Stick with Whisper. The core innovation is happening in TTS at the moment.
325.
by oezi 1y ago
Sure, but gross labor share of GDP is 60% in the US (salaried employees earn 60% of GDP in total) so it doesn't change the consideration that much. The US is spending roughly 15% of the GDP on health. If 1% of GDP is due to obesity the
326.
by oezi 1y ago
My mathematical intuition says this isn't really much considering US annual GDP is 27 trillion (27,000 billions) USD. So less than 1% of GDP. Your linked article says: Obesity adds between $147 billion and $210 billion to annual U.S. h
327.
by oezi 1y ago
If you pool together doesn't this increase the risk of you being prosecuted for dealing these compounds?
328.
by oezi 1y ago
A key issue seems to me that they didn't do a gradual rollout of their new models and don't have reliable ways to measure model performance. Worse, I would have believed they are running many different versions based on the expect
329.
by oezi 1y ago
Isn't xcodec2 also lossy? I thought it is also just another neural codec (50 tok/s, single codebook). What are people using to upsampling back to 44,1 or 48 khz? Anything fancy?
330.
by oezi 1y ago
Do you happen to know why Orpheus and Llasa use Finetuning for voice cloning? Zonos uses 128-float embeddings for voices and it seems so much nicer. Because you can just mix and match voices without changing the model.

More ›