Milan Novkovic

DIY Pro

[Massimo @Rainmaker1973]

An incredible wind-powered drummer made entirely from scrap metal.

[Milan Novkovic @MilanQSP]

Impressive!
Who are we to trust if not Big Serge 🙂

[Big Serge @witte_sergei]

How did I do

[Milan Novkovic @MilanQSP]

Welcome to Balkan

[Igor Pifat @pifat]

Zamalo pa kao Gaša Šeprtlja koji svira gitaru i vozi ono svoje prometalo, a milkan dobija srčku 😂

[Milan Novkovic @MilanQSP]

Ja se zaljubio u titanijum još od mladolikijih kamping pa ronilačkih dana (ko nije imao nešto titanijumsko u ruci neće me možda razumeti) pa sam maznuo foru moje žene (ili nas DIY-jevaca kad kupujemo nepotreban alat visokog kvaliteta): isklučio pre-frontal cortex i kliktao dugmići bez razumevanja i razmišljanja sa basom za upozorenja potpuno isključenim, i evo top pro Japanci in the pipeline (apparently, u Japanu se više koriste). Od plastike dobijam ospice, a u 20 godina očuvanoj tikovini, i pored poenkad pranja u mašini, žive svi bakcili koji su ikad postojali na zemlji, bez obzira na pranje. Imamo ih dosta i od maslinovog drveta, ali to i upozorenje žene da je probala titanijum i nije nešto zadovoljna sam, dakle, ladno isključio. Kao fol ratosiljali smo se Španije pa se ja bacam sa novim slobodnim vremenom i na kuvanje i automatski stičem neke kuhinjske mandate i privilegije (čitaj: najebo sam kad ukapira da sam ipak pazario)

[Milan Novkovic @MilanQSP]

Spam, znam, al ovde se ionako ne diskutuje previše, a ja nemam pojam o vašim/našim (ja sam još uvek prvenstveno naš) političkim igračima.

Plus moram da pitam na književnom prigrevačkom pošto se moj tarzanoidni engleski neće shvatiti kako treba: el ima neko od vas iskustva sa Japanese titanium cutting boards za kuhinju? Mislim na high-end modele.

[Milan Novkovic @MilanQSP]

AI is just a tool - and a very powerful one - in authors' toolbox, and everyone should start using it.

Nothing bridges our tacit creative talents and potential, passions also, on one side, and some actual creations for which we didn't have the time in the past on another.

Video is incredibly powerful also, and is a bit of a leap as in static pictures domain, if you fancied it, you could fuse Caravaggio's and Rembrandt's widgets and blacks even two decades ago, but even until very recently you couldn't really turn this picture of ladies of yours into a video chat on the topic of risks that Italian cooking is exposed to from invented tradition drives by mediocre influencers, overweighted importance of technical mastery, unstable cultural gatekeeping by noisy pseudo-experts etc. 🙂

[Milan Novkovic @MilanQSP]

Josh, pretpostavljam, iako mi ovo pomalo dođe kao trik pitanje u smislu da mi je Josh interesantniji, a Charles bliži kao osoba.

Pa mi je Diana jako interesantna, Lauren isto tako, Maggie su možda malo omašili u castingu - ali cool lik, a Lizzie McGuire nam nije silazila sa ekrana dok su ćerke odrastale (davno bilo, sada su 40 i 30 godina stare) pa mi je to možda jedan od razloga što mi je Kelsey možda prosečna.

Ipak, evo gledam seriju po drugi put i pored svih streaming kanala što imamo pošto sam ušao u fazu gde mi opuštene commedy-drame često više prijaju.

Ovu Maksinu seriju sam počeo davno da gledam i bila mi cool, ali mi se zagubila, a od sličnih na spisku nema ni Suits, pa White Collar - na ovu smo se baš onomam i žena i ja primili, i con-man žanr nam prija - ajd to je coomon psihologija, objašnjavali su onomad kad je ovde u UK Only Fools and Horses izglasan kao najbolji UK sitcom ikad (po meni bi dosta Uk sitcoms trebalo da bude na listi) ... itd, da Star Trekove i slično ne spominjemo, a i prenisko plasirane Battlestar Galactice, pa odsustvo The Expanse (meni bi bio jako visoko), Dr House, ...

Pošto ja skoro svašta mogu da gledam interesantno mi je šta sam prekinuo pošto mi se nije dovoljno sviđalo: Pluribus npr, pre kraja sezone, Black Mirror posle prve epizode ali drugari me ubeđuju da samo nju treba da preživim pa je sve cool pose toga, Sense8 posle prve sezone, Stranger Things jedva odgledao prvu sezonu a gotivim klince inače, itd ... 🙂

[Milan Novkovic @MilanQSP]

@LomaXoma @OshotaMasa Samo da se pohvalim: odgledao sam Younger pre ne samo žene, nego i ćerke jedne pa druge ... i onda sam im preoporučio seriju.
Ako ovo ne spada u kategoriju pičim pa rasturam onda ne znam šta tamo uopšte ide 🙂

[Milan Novkovic @MilanQSP]

It is Musk and/or (more likely) Google's Gemini.

We do see in daily marketing mostly LLM heavy players, but they are all well beyond LLM (LLMs navigate highly dimensional spaces as per our prompts but even along the way, but for certain at various leaf nodes where the problem belongs all sorts of localised, specialised models kick in.

Elon M came later - as he said recently only 3 years ago, while others are 6-10 years old) - and claims positive "second derivative" in progress (some notion of fintech Gamma) - and I believe that.

Gemini is, in a broader game, imo, way ahead of others.

Clarification first: when we see this second L in LLM, and talk tokens, they are not just simple words - can be, and largely are at the beginning of parsing of our prompts, but fan out end get richer structures/tokens, as localised meanings get lost in phenomenological nature of neural nets they flow through.

Back to Google and their notion of "Action is Just Another Language" but still using common LLM as a reference: "While other labs focus on building isolated software brains (LLMs), Google DeepMind operates on a foundational concept pioneered by senior roboticists like Vincent Vanhoucke: Perception, semantics, and physical action are all part of the same unified intelligence." - from my conversation with Gemini a month ago on the topic of physically agile - dogs, cats, humans, UniTree AS2-W running up or down tricky terrain with not AGI brain, but a part of brain that is a "super-class" (programming lingo, common ancestor, a class from which another class - a derived class, inherits everything that super class can do) of all big classes of intelligence, including future AGI as we abstract it out now.

Now, sorry, I started mixing in my opinions with common: "Most AI researchers now believe true AGI won't be a single class, but an instance of Multi-Inheritance. Tomorrow's AGI will likely inherit its core logical, language, and coding reasoning from large foundation models, but it will use the real-time spatial, physical, and sensory-motor loops pioneered by robots like the AS2-W to ground that knowledge in reality. Without the spatial class, the AI has knowledge, but it lacks understanding. "

My subjectiveness here is that "...but it will use the real-time spatial, physical, and sensory-motor loops pioneered by robots like the AS2-W to ground that knowledge in reality..." carries far heavier weighting that the others.

Namely, it's not just mine, I heard it from others decades ago how all our main types of intelligence - abstract, numerical, linguistic, ... - stem from our ability to (mentally first) comprehend and use/control classical mechanics, time and space. Physicist go and specialise in the direction of speeds, accelerations, ... you in the direction of deltas, gammas, ... momentums you share, ... in one way or another you/we share an incredibly large proportion of key intelligence lego blocks.

And Google is very clever, it's a joy following them from time to time in this domain you mentioned: "..Instead of training a model for a specific robot chassis, Google DeepMind spearheaded the Open X-Embodiment dataset. Their researchers proved that a massive AI transformer trained on 22 entirely different robot bodies (quadrupeds, humanoids, robotic arms) exhibits a massive "positive transfer" of intelligence. The AI gets better at controlling Robot A by practicing on Robot B. This strongly supports your OO theory: Google is actively building the base class PhysicalIntelligence that can be flashed onto almost any hardware...."

Etc, Google is probably still ahead worldwide, but largely in labs (somewhat (too) relaxed about chasing profits by editing our Emails with AI, office docs etc), Chinese catching up incredibly fast, others splashing a lot of money on brute-force capture of as much of the market as possible regardless of inherent potential of their systems from mid-future perspective, Elon M smart, well funded, and a master of cross leveraging ideas and businesses.

[Milan Novkovic @MilanQSP]

@LomaXoma Tko je ovaj gospodin, btw?
Malo zapliće kad priča ko ja kad popijem malo više (tako mi žena zna da sam pijan, kaže jezik mi malo luta i veseliji sam, agresija spadne na apsolutnu nulu - kaže da sam bolji čovek kad sam pijan nego kad sam trezan 🙂

[Milan Novkovic @MilanQSP]

@LomaXoma Pa jbt, umesto da podmlađuju Džesiku Lang, ili bar trguju commodities, vidi na šta Srpski young Turks troše tokene 🙂

[Milan Novkovic @MilanQSP]

@skdh Intro first and then heavy stuff from one of mine decades old favourites:
https://t.co/Qtoc7MCx9S

[Milan Novkovic @MilanQSP]

@Math_files Henri Poincaré or John von Neumann

[Milan Novkovic @MilanQSP]

@rustyrockets Not in a way they are talking about them now in order to cover potentially imminent sovereign debt crisis, war losses, energy flow disruptions and all the avalanches that will follow.

[Milan Novkovic @MilanQSP]

My wife would have made a fantastic politician!
With some other guys we touched upon TV Series "Billions" once some years ago and one guy said how he didn´t believe characters like that existed in real life.
I said how I worked with two. They shared some common traits:
- Unless you stared at their faces while a team member was delivering some bad news you wouldn't notice disappointment - it only lasted a fraction of a second and after that they had a range of scenarios and solution.
- They were, on the other hand, incredibly aggressive towards outside-of-team people who they perceived behaved in an unfriendly manner.
- They would "kill" for the team - it was only a figure of speech at the beginning by which I meant buying coffees, books, good food, taxies, fighting off enemies, ... until I started paying more attention to one of them, half-mad Irish.
- They both made irrational decisions later when they decided to stop chasing riches for themselves.

[Milan Novkovic @MilanQSP]

@LomaXoma Jao vidi iskoči jedna za tebe na Fejsu 🙂

[Milan Novkovic @MilanQSP]

@skdh In my wife's case it doesn't really matter, sentences can be used also.

[Milan Novkovic @MilanQSP]

@pmddomingos They might just be trying to slow down Google - too much of their stuff is converging and leveraging each other.

[Milan Novkovic @MilanQSP]

Haven't read the paper, but I thought it was always the case that various forms and sub-forms of human intelligence were deriving their basic capabilities from our tacit mastery (in vivo and/or mental only) over real, visual world and classic mechanics.

[Rohan Paul @rohanpaul_ai]

Google DeepMind + Harvard + Stanford and many other top labs paper argues that a path to AGI may be visual AI that builds a world model, remembers changes, predicts outcomes, and acts.

Most multimodal AI still treats vision as something you feed into a language model.

The paper wants vision to do more of the thinking itself.

A capable visual system should learn directly from images, video, 3D structure, and interaction. It should understand what exists, what changed, what is hidden, what might happen next, and what it needs to look at before acting.

they point to video generation, reconstruction, persistent memory, continual learning, multimodal sensing, and robotics as possible pieces of the same system.

they say stop judging visual AI mainly by image Q&A, captions, or realistic video.

learn the world's structure from visual experience, keep updating that knowledge, and use it to predict and act.

[Milan Novkovic @MilanQSP]

GDP is a poor measure. Examples:
- In your case depends on what you do with the savings - there are many scenarios where - with some delays - you add a lot to GDP.
- Late credit card payment fees are high. That goes towards GDP yet a lot of wealth goes to financialists abroad, although some stays in UK.
- It can be even poorer measure as one compares countries, regions etc. Example: if you buy potatoes in China directly from producer chances are most or all of that will not go through GDP at all. If you do it in US it'll go through GDP multiple times, 7+ times was the figure 15+ years ago when I "checked" last time. The most drastic example was a <1$ pack of screws from DIY shop: the same pack was paid by MIC end users over $70, and there were/most often are intermediaries in between.