- Here is Histolog Scanner in action
- value of strong force as function of distance
- Isn't the Java language simple?
- How one can assess quality of #ML #model?
- you may find this benchmark table useful
- reference numbers can save you from a trouble
- There's no such thing as "common sense"
- autonomous car driving that includes natural language reasoning
- Code your own Dahlia
- Too far to Friday, but ...
- NVIDIA Developer Program recently made many courses free
- contemporary methods of manipulation
- remember this terrific past?
- we won the Google Cloud Technology Partner of the Year Award
- Interested in solving hard problems with C++ or Rust?
- Simulation can provide a lot of insight
- VARANKIN: If you need fresh ideas on Espressif Systems MCU
- what does prevent it to construct a whole?
- control #theory depths
- visualize heat convection flows
- Java language "Expression" syntax rule
- ищем Linux System Engineer
- 5+ years on a single button cell
⌚ My Casio hand watch operates 5+ years on a single button cell. It has embedded calculator
to "fix" my decent abilities to crunch numbers. In contrary, the system on a wafer below consumes
enormous amounts of power. As you can see in a video, a major part of design consists of electrical
power delivery units and thermal power dissipation units. That is a typical engineering cost of
computing teraflops of #ML data.
🏭 You may wonder, but entire computing power of the system is concentrated on a single, relatively
small flat module, made of a single cut by sides 300 mm silicon wafer with interconnected processing
cores. Remember, CPUs from Intel Corporation and AMD are being made on similar wafers and technology,
but equipment cuts the wafer along and across to not so big separate CPU chips, that go next to
individual packaging.
- я в поиске новой карьерной возможности
- To get a progress, one needs to move to yjr more energy efficient point
- world of autonomous blow-up fish
- What do you want your money to look like?
- Теперь можно создать "чистое" резюме
- looking for a Software Architect
- What is your favorite Python IDE?
- VARANKIN: Next development step completed!
- There were times when D class audio amplifier was treated cheap
- information existing in a continuous stream or in discrete chunks
- Automatic Number Plate Recognition
- approach of chain of events
- VARANKIN: how to process temporal composite values
- protect sensitive payment card data
- French edition of good old anecdote
- Life in a mixed reality
- we announced Variable Sparse Pre-training and Dense Fine-Tuning
- tech-to-business LLM metrics
- Opinion about AI future
- conversation with Inflection AI
👨⚕️ 👦 Here's my conversation with Inflection AI (the ""Pi, a personal AI — a supportive and
compassionate #AI that is eager to talk about anything at any time""). I pretended a stranger
and asked questions about myself. You may find it a little bit immodest. However, I think every
scientist experimented on him/herself, so did I. At least I haven't been revelated and nobody
was hurt. 😏
🥁 I confirm, and you can see it, it was truly like a talk with a real person. Here we go with
advanced #LLM technology. When I first noticed now famous paper ""Attention is all we need"", I
worked on my own ideas, so the paper leaved just a flash of interesting idea. Who could imagine
then, that after just a few years of work by talented people, a simple increasement of token
context size up to 2048 items and more (along with other improvements), can break some hidden
threshold and open a very impressive way to work with information based on #natural #language.
👿 Many early adopters complaint about errors and hallucinations of the systems they touch. I
would say, it happens because systems work on somewhat static, pre-learned data. Imagine, when
such basement is provided over prospected reasoning engine, we could meet a brand new, alive
idea, that provides a sort of intelligence here. ""
🎈 There are now almost invisible, lightweight signs, that it is a reasoning math that could
be the next major advancement towards real #AGI. Current LLM systems will keep their role to
provide smooth interface (API) to communicate with other smart systems. Including humans, of
course.
🌐 conversation https://lnkd.in/dhT3bwfQ
📜 "Attention is all we need" https://lnkd.in/ddPYKmw2 https://arxiv.org/abs/1706.03762
- Мир потихоньку полнится программами от "юных талантов"
- Embrace the Power of Perspective!
- Dark matter is composed of particles that do not absorb, reflect, or emit light
- Can ML work 99.999% reliably over flat 2-D image?
- Pattern is nothing without right context
- AGI - это другое
- It is so inspiring, a feeling of appreciation...
- ?
- ?
- Software programming is a self-convoluting industry
- Miniature actuator waits for applications
- Wind blows, fire burns, rain falls…
- Will software programmers exist in 2030?
- MSG Sphere, designed by Populous Architecture in Las Vegas
- 1B tokens, declared as a momentary benchmark
- AI code interpreters are getting brilliant
- MSG Sphere in Las Vegas has started using its screen
- Влияние приобретенных шаблонов на поведение
- Evolution of informational exchange
- AI Startup Programme
- Fruit fly brain discoveries
.jpeg)
