YouTube
QME
Alpuha Pushistovna
1.9K posts
Alpuha Pushistovna
@AlpuhaP
Once upon a time, a fluffy animal was discovered. It had four short paws, one tail, a soft tummy, a pair of miniature ears, and the desire to conquer the worlds
Hydaelyn Katılım Nisan 2018
39 Takip Edilen32 Takipçiler
@quindenam @ShitpostRock2 they do but in japan (and tbf other east asian countries too) the teachers and the ones writing the textbooks don't actually speak english at a proficient enough level to be able to actually teach anything useful in english
sentences like how are you? fine thank you are the limit
English
@TheOpeningMove1 @deepseek_ai This is misleading.
"He called exactly this scenario" is doing Olympic-level work here. Huang warned about ecosystem bifurcation, not DeepSeek API pricing being tied to a chip production schedule.
Real signal, fake precision.
English
the story nobody's covering: v4-pro's $1.74/M is tagged "bottleneck-constrained; expected to drop once ascend 950 supernodes mass-deploy H2." deepseek just tied its api economics, in writing, to a chinese chip production schedule. huang called exactly this scenario "a horrible outcome" on dwarkesh 9 days ago.
English
🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.
🔹 DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
🔹 DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.
Try it now at chat.deepseek.com via Expert Mode / Instant Mode. API is updated & available today!
📄 Tech Report: huggingface.co/deepseek-ai/De…
🤗 Open Weights: huggingface.co/collections/de…
1/n
English
@nullbotto @PMinervini - visible red
- infrared
- far infrared
- then into microwaves
so the "12mm" color selector for an item would be something like "microwave beige" 🤣
English
Let me say this clearly: airplanes cannot fly.
Flying is an evolutionary mechanism that first insects and then dinosaurs used to avoid danger etc.
Valerio Capraro@ValerioCapraro
Let me say this clearly: LLMs cannot feel emotions. Emotions are evolutionary mechanisms. They push us to avoid danger or approach what is beneficial. We experience emotions because we are alive, and we want to stay alive. LLMs are not alive. Yes, emotional language may be encoded somewhere in the LLM. Yes, it may even be associated with some LLM output. But that is just a superficial property. There is nothing deeper behind it. For a very simple reason: LLMs do not have an intrinsic and inescapable drive to stay alive. This is what we call “motivation fault line” in our paper describing seven fault lines between human and artificial intelligence. * Paper in the first reply
English
@nullbotto @PMinervini Didnt you know on aliexpress they had color option variants as wave length instead of colors 🤣
English
@PMinervini Give them ability to change environment, discuss, argue, and some stats they can lose. Or even die… and observe what happens
And my intuition says if a system cares about not dying, it will have properties of a living being. Might even develop emotions. Consciousness even.
English
@PMinervini If you continue this thought process, a human is a closed loop system where it have to monitor thousands of parameters, keep alive and emotions is simply a mechanism to break out of destructive loops or stay in creative ones.
Give artificial neurons enough embodiment and dangers
English
@alexvcs_2 @cmdlinefanatic @Rothmus Anyway, I digress🤣
Don't make your logs source of truth, this is disgusting behavior!
I'm sick of lack of effort to make things nice and tidy for operations / debugging / analytics. I wasted too much time looking at Datadog/Coralogix. I'm exhausted 😩
English
@alexvcs_2 @cmdlinefanatic @Rothmus 3. When releasing a new feature that might break, you should be able to toggle logging on/off simply by changing it from admin panel: not re-deploying, not changing code. A toggle via Web/CLI. OK to log info/warning for first 1-2 weeks, full request trace.
4. Granularity: service
English
@cmdlinefanatic @Rothmus This isn’t logs.
There are analytics platforms like Tableau that make it visible on dashboard: clients, devices, average active time on the app, and so on.
This can be sourced from logs, but it’s as efficient as watching random TikTok reels to become PhD in computer science 🤷♂️
English
@yoheinakajima When you type properly, slowly, and thinking what comes out from your thoughts, it makes you think.
Compress knowledge. Clear your mind.
Delayed thought can eventually become good.
Rushed one is always bad.
English
@sensibleposter @allgarbled Coding something on first day isn’t a challenging thing if people prepared good first day intro problems to solve. Could be as simple as adjusting alert thresholds, documentation, or even updating a library.
It doesn’t have to be extreme programming day 1.
Just getting to prod.
English
@sensibleposter @allgarbled Planning out something useful isn’t needed. Everyone have backlog. All items in there have proper description what needs to be done. And honestly an agent could do it but it still needs human validation and judgement.
English
@OriOriOriOnion @cyber_razz I have around 14 million files currently
English
@OriOriOriOnion @cyber_razz Real. I just store entire system in a single folder. Max depth is 0.
Sometimes files are rewriting each other but that’s fine, I just download it again from backup 🤷♂️
I never had to run cd 😤
English