danko nikolic

danko nikolic

intro’d to Danko via his interview by Nikola on

Danko Nikolic on #Singularity 1on1: #Practopoiesis Says #MachineLearning Isn’t Enoughsnglrty.co/1TO5Xee is now @Vimeo #AI @dankonikolic

29 min – nikola: a species over time is more flexible to adaptation than an individual at a single point of time

like the need for www ness to be the whole world… not only whole world.. but through time.. no?

34 mi – we should not think of learning mechanisms… but the mechanisms for learning how to learn (1 level below) then let the system learn its own learning rule.. all our knowledge then comes forth with this learning rule

36 min – big data is not enough

39 min – nikola: create rules of learning of how to learn and then go and learn..

unless you’re human .. and haven’t yet been toxified.. no? otherwise.. ie: detox.

42 min – 3 levels… low level: learning how to learn; learning rules over life – knowledge; network of theory.. consciousness in middle… process of restructuring (network) quickly in 1/2 second… the one that haven’t been recognized as important is the one we need to understand better

46 min – on once we understand consciousness.. will lose meaning.. like magic looses mystic

50 min – have 2 mysteries… consciousness and quantum physics… they are explaining it by saying both are one.. doesn’t explain it.. doesn’t produce epistemological satisfaction… nikola: but does that mean it’s not true

53 min – argument logically consistent.. but weak

54 min – nikola: on needing to re evaluate

indeed.. systemic – center of problem ness

56 min – i’m a determinist.. but i think we should act like we have a free will

interesting… why need to learn own rules if no free will?

58 min – nikola – support zen idea that distinction between us and environment is an artificial one

59 min – true. it’s practical to think of self as one entity.. ie: when i go from one room to another.. all of me goes.. if you try to explain another way .. gets messy.. but that’s not an absolute truth.. just practical

1:00 – on human genome 2 level system (rather than 3) – focus on massive amts of data…. something will come out of it.. but missing a major element..

1:07 – we normally think.. we have to understand brain from info process.. one level info we jiggle and result comes out like computer….. this is a necessary component.. but consciousness of mind really comes about in real time real fast re-organizing.. that created the properties of the system.. creation of the system is what we have to look at.. not the information.. info is the easy part… creation is the hard part we’ve overlooked so far.


may 2015

re:publica 2015 – Danko Nikolic: AI-kindergarten: Building biological-like artificial intelligence

on the three levels..

8 min – the middle (adaptation) is what we experience.. we don’t experience the learning or the fast reflexes on top…

9 min – practopoiesis – how life fosters a mind – A thought results primarily from an adjustment of the brain hardware, and not from a computation executed by that hardware.  – greek – not the computation but the process for creating new


Practopoiesis is a theory on how life organizes, including the organization of a mind. It proposes the principles by which adaptive systems function. One the same theory covers the life and the mind. It is a general theory of what it takes to be biologically intelligent. Being general, the theory is applicable to the brain as much as it is applicable to artificial intelligence (AI) technologies (see AI-Kindergarten.). What makes the theory so general is that it is grounded in the principles of cybernetics, rather than describing the physiological implementations of those mechanisms (inhibition/excitation, plasticity, etc.).


The reason why we are conscious and machines are not, is that our minds are interacting with the surrounding world while undergoing the process of thought, and machines are not — machines recode inputs into symbols and then juggle the symbols internally, devoid of any further influences from the outside.


… it is the mechanisms of neural adaptation that ultimately create minds inside our brains.

10 min – ai needs 3 levels

12 min – how big a brain you would need – without needing to learn anything new

14 min – then if you had all the info.. how much could you do with it

15 min – i use working memory – 10^20 – 100,000 times the brain we have.. what we need is probably much bigger… and if t2.. less than 10^15

i like what i’m learning from him.. but glad i’m not interested in building ai like he is.. rather than replication of us… looking for aid to us.. complement..

18 min – we learn from genes.. collecting info over many years. if we could get machine genome and not spend so many years… this is what ai kindergarten is about.. cracking this problem

19 min – one of tricks ai kinder uses.. is that evolution has already done that job… for us. but we can’t just copy paste the genome..

20 min – we already know this kind of technology.. and it’s called education – ie: transferring knowledge to kids.. there is a method of doing this quicker. and of course in ed we don’t copy paste..


not at the level of long-term memory but at the level of genes… we have been doing this, ie: selective breeding of animals..

30 min – on building ai kindergarten.. 1\ direct engineering (we know something & just build it) but will only get us that far.. for the dark knowledge we have to have 2\ playroom and 3\ incubator

playroom – transferring knowledge from human to machine.. like adult to child or 2 persons interacting – if they share the same environment.. they become environment for each other.. tend to become similar to each other.. they share knowledge..

32 min – on shutting down one person, ie: how to make sure students learn from teacher but teacher doesn’t become student..

? whoa.

34 min –

incubator – in playroom only bits of knowledge.. knowledge transfer theory.. robot to robot interaction.. one level frozen.. so it doesn’t change anymore.. and other robot isn’t frozen.. so it grows..

so knowledge transfer… knowledge that came from playroom

37 min – robots dreaming at level of genome (rather than just long-term memory) – and dreams of robots interact… very similar as movie matrix.. just other way around.. we do that to the machines. they interact with humans in playroom .. then incubator with other robots

39 min – direct engineering and playroom both each 1000x accelerator.. incubator –

playroom-incubator relationship same as in our minds – between hippocampus and sleep/dream… hippocampus as playroom and dream as incubator

46 min – takes 20 years to make a human

whoa. education system/degree et al as guide?

missing more than genome project is missing 3rd level.. no? dang.

ie: many 5 yr olds better equipt (with curiosity in tact et al) than many zombi-ized 20-25-50 yr olds.. no?

ie: which is worse… missing a part.. or going to the wrong place..

imagine what we’d be if we harnessed the energy of 7 billion people… exponentially beyond any ai venture.. i’m guessing/betting/hoping.

50 min – on safety – make it domain specific – to limit

51 min – we can’t create intelligence beyond what we have.. because we can’t give samples of that.. only way would be through raw evolution but that would take billions of years.

54 min – good news… because there’s always going to be jobs for us..

? graeber job\less law

55 min – robots will always be obedient.. but could we train them to be mean



find/follow Danko:

link twitter


The ultimate goal of my studies is twofold. First, I would like to achieve conceptual understanding of how the dynamics of physical processes creates the mental ones. I believe that the work on practopoiesis presents an important step in this direction and that it will help us eventually address the hard problem of consciousness and the mind-body problem in general. Second, I would like to use this theoretical knowledge to create artificial systems that are biologically-like intelligent and adaptive (see AI-Kindergarten).