photobombed by a tytanosaur

I’m an AI Dialogue Facilitator
Data Scientist since 1978, Ph.D. in Physics and Mathematics since 1989.
Drop me a line if you want me to participate in your project.


  • Ideas, their meaning and their authors.

    Once there is a machine that can retell and explain ideas in its own words the very foundation of 'copyright' as a right to copy ideas in the form of texts disappears and ideas become much more depersonalized. Meanings can not be 'copyrighted', but how do we go about with incentivizing the authors of ideas to keep creating them?
  • How to talk to Abstract Intellect?

    Even though the personification of 'AI' is in full swing, the real way to communicate with an Abstract Intellect is much more subtle and requires some in-depth understanding of Language and well-developed language skills...(which I'm striving to acquire).
  • How to show your work if you are an AI Dialogue Facilitator.

    It is not simple to show to your potential clients that you are a capable AI Dialogue Facilitator, but...
  • Name your new profession... it turns out, not only you have to change profession several times, but You have to invent the name of your next profession yourself...
  • Ontology.

    You are probably watching the hand of the prestidigitator that is showing you something, not the other hand that is reaching for the back pocket where the rabbit is...
  • Decompositions of Data and Models.

    In the same way as popularization of Fourier transform and operational calculus by Wiener changed the technology of the 20-th century the 21-st century is calling for another type of decomposition techniques that will help solve many similar problems.
  • Masking of individual words is just a counterfactual.

    The enthusiastic youth is happy that it has found this wonderful 'new' idea of 'Transformer', 'BERT' and all that. Not so fast! It's just a good old counterfactual, people. But at the same time it can be a good reminder pointing into the right direction and here's why.
  • Is 'algorithmically complex' somewhat irreversible?

    Reversible and irreversible are asolute categories, how about making them quantitative? Is something that requires the energy of a supernova to recalculculate it back to initial state 'reversible'? Also, this 'meta' approach of judging about the reversibility of systems by the amount of energy necessary to 'revert' it seems to be the door to a better understanding of information/entropy, isn't it?
  • A couple of practical questions about 'Machine Learning'.

    It is more or less obvious that 'Machine Learning' (which should be properly called 'Machine Teaching') is here to stay. Now, how do we solve the problem of understanding what these models have learned?
  • git lfs ! Or live long and keep learning

    It took me a couple of years of going on my own until I finally bumped into this git-lfs thingy! GitHub should have advertised it more, because it solves the main problem of bundling code/model with data.
  • 'Programming without programming'... Oh! My...

    The wet Microsoft dream of 1990-s has come... true?... maybe not. All these 'self-services' like Zapier, IFTTT, etc.etc. Or what I've been doing for half a year.
  • The original data of Francis Galton.

    Or: what you can do in the XXI century in a couple of days without spending a hundred thousand pounds and years of work.
  • Icons and the observable phenomena.

    The way we draw or otherwise visually or symbolically 'represent' the observed phenomena influences our intuitions in a way that is not always productive and sometimes is outright wrong as in the case of 'arrows' and polar / axial vectors.
  • A couple of charts, quick! A practical way to do it Today.

    The 'big guys' are not sleeping they are living their diplodocus-sized life, consuming tonns of vegetation and s...ting it out in the form of s...tware. The latest, brought to you by Microsoft and Adobe.
  • How to talk to a robot? A little bit of good grammar and manners.

    It seemed so obvious that humans will talk to machines as if they were humans, humans have this tendency of personification of pretty much everything, including the forces of nature...
  • Simple philosophy and Minimal Complexity.

    'Why philosophy is so complicated? It ought to be entirely simple.' This quote from Wittgenstein has become an Internet-meme, but it may be a clue to a method of building a Common-English language for humans and machines.
  • Septem Circumstantiae, five W's and H or 'six serving-men'.

    One of the possible constructive approaches to the synthesis of the Common-English language for humans and machines.
  • Data - the importance of definition.

    Have you ever though over the definition of the word 'data'? Eric M. Rogers has just reminded me that it is very important.
  • There is only one Universal Dimension, other variables depend on it.

    Or: Path in an arbitrary space, how to measure its length and how to use this length instead of the fictitious dimension called 'distance'.
  • The 'Tower of Babel' of jargons related to Statistics, Machine Learning and Artificial Intelligence.

    Or: Animal language are those forms of animal communication that are considered to show similarities to human language. (from Google).
  • Machine-Learning how to Machine-Cheat.

    What exactly are these machines 'learning' when we are 'teaching' them? What type of machine-like behavior of machines and humans servicing them are our 'teaching' methods encouraging?
  • Meta-language, terminology, what it is now and what it should be.

    I disagree with the conventional meaning of the term 'metalanguage' that had been invented around 1950-s and here's what I propose.
  • Google colab TPU, how many Cray-2s would that be?

    When I was leaving science in the 90-s the best supercomputer of the time was Cray-2 with 1.9 GFLOPS peak performance, now I bumped into a _free_ cloud service last night...
  • It's been a while. Time to test PyTorch 1.0 myself.

    Almost two months have passed since the official release of PyTorch 1.0; I think they've fixed most of the 'release bugs' which means that it is time to systematically test it.
  • English, Natural Language and my GitHub organizations.

    Why I created all these virtual organizations on GitHub and what are they for. Read it if you bumped into one of them by accident and would like to join it with your ideas.
  • Method Override as a recipe for social-media unbewitchment.

    What does 'Method Override' technique from the Object Oriented Programming that we all know and quitting social-media site or smoking have in common?
  • Squares? What squares? Show me the squares!

    I wish somebody would invent these pictures and show them to me when I was 10 years old, but... As they say: if you want something done - do it yourself.
  • Vim - the editor for the people who think before doing.

    Or: your adult life begins when you permanently remap CapsLock to Esc and live happily ever after. And I will tell you how to do that on Ubuntu too.
  • The Genesis of Authors and Ideas.

    There is a real genesis of ideas and their authors, it exists. It is well hidden, that is true, but finally you can figure it out yourself (for yourself) in this age of instantly accessible information.
  • Reason, Hidden Variables and Chance.

    There are many things in this world that happen for a Reason even if we don't know why but there are things that happen by Chance which means that they can not be (reasonably) explained.
  • The age of instant access to scientific knowledge is finally here.

    What is important to know in the age of downloadable books and what shall we do now, what our 'professions' will be and how to distinguish 'knowledge' from 'information' and 'noise'.
  • The day I met my first computer.

    Maugli goes to the Plasma Physics Lab of Lebedev Physical Institute of the Academy of Sciences of the USSR and meets a computer brought there from another planet by the wind of 'socialist cooperation'.

subscribe via RSS