triton-lang

Just caught up to this 😅 : so OpenAI's Triton is a pythonesque DSL for exposing hardware primitives using a "tile" paradigm.

Handles hardware for you

• - numpy : CPU only
• - tensorflow, pytorch : anyPU

You handle hardware :

• - CUDA : nVideaPU only
• - triton / C/++ / SYCL / openCL : anyPU

Triton is a dialect of MLIR@LLVM ...

localised LLMs?

Kita bincang pasal tek saja ya. 

Pentafsiran dan penterjemahan dalam LLM berlaku pada lapisan tersembunyi, iaitu model mempunyai gagasan ( rangka data graf ) yang tidak terikat kepada mana-mana bahasa. Ini sistem A

Dari situ, penterjemahan dari mana-mana bahasa kepada sistem A, lepas tu dari sistem A kepada mana-mana bahasa adalah satu transformasi khas. Lebih sesuai bagi SLM / MLM. Inilah sistem B.

Saya tak tahu lah rekabina dalaman LLM yang tuan sebut tu ... 😂 ... tapi saya rasa, kita kena asingkan. 

Negara kita kena imbang tujuan dan had ekonomi. Gergasi luaran sudah hampir menyempurnakan sistem jenis B. Kita boleh khaskan bagi loghat la, tapi ini kira penyesuaian bukan penyelidikan dasar.

Kita juga boleh menyelidik secara mendalam sistem jenis A, sebab budaya kita ada graf pengetahuan yang lebih khas. 

Yang paling jijik : jangan kita dengan duit yang terhad cuba nak jadi juara dalam setiap jihad sekaligus. Lebih-lebih lagi, jangan orang kita cuba buat kerja malas dan nyahmaju dengan cuba gabungkan sistem jenis A dan B. 😘🤣🤣🤣😭

---

Further reading :

• 1. visual recognition as a latent space, under verbal languages as feature spaces : LENS / LLMs Enhanced to See, 2023 : https://arxiv.org/abs/2306.16410 ( if you know anything about my views, this is a big deal to me 😛 - so far this is the most anthropomorphic implementation I have seen documented )
• 2. separate training of latent platonic spaces, under verbal languages as feature spaces : Zipper-LoRA / low-rank adaptation, 2026 : https://arxiv.org/abs/2603.17558 ( I'm zooming into this area a bit, since yesterday's public conversation about how some Malaysian localised LLMs don't seem to perform efficiently )

bab kelarian kemahiran

Fenomena ini menarik. Ada akarnya. 

1. Kita mahu keadilan melalui ekonomi sama rata.

2. Kita mahu keadilan melalui ekonomi tidak sama rata.

3. Kita mahu pertengkaran berhenti.

4. Kita mahu ekonomi berkembang.

5. Kita mahu kemahiran menduduk dan tidak melari keluar.

Kebanyakkan pertengkaran adalah antara (1 vs. 2), dan antara (4 tanpa 5, vs 4 wajib 5).

Agak-agaknya, (1,3,4,5) adalah satu pihak, dan (2,3,4, tapi neutral terhadap 5) adalah satu pihak lagi. 

Dalam analysis ini, hanya (3,4) boleh jadi tumpuan senegara. Maka pemimpin kena bermula setiap strateginya dengan mendasarkan (3,4), dengan tolak ansur terhadap (1 dan 2), jadi tercapai lah (3,4) dan akhirnya [(5) yang menyokong (3,4) dan merendahkan lagi hati (1 dan 2)].

Fasa 1 : mendasarkan ekonomi

Fasa 2 : tolak ansur antara, kesamarataan dan keistimewaan ekonomi

Fasa 3 : ekonomi kembang

Fasa 4a : juara kesamarataan dan keistimewaan kedua-duanya direndahkan coraknya

Fasa 4b : kemahiran mula dibela dan bukan diusik keluar

Kemahiran politik itu memang kena tertumpu pada 2/ tolak ansur, dan 4/ rendah hati. Jadi pakar pengendalian tisu-tisu hati ini lah yang kita harap dinaikkan dalam PRU dan kerajaan masa terdekat.

getting started on computer vision

Brief dip into computer vision jargon. Anthropomorphism. Humans learn to recognise gross gestalten, before understanding mereology ... and humans learn to recognise 2D persistence before 3D persistence. But as conscious processes develop, they discover more efficient storage formats for data, by relying on 3D persistence graphs.

So while we could train a vision system to be more human like by beginning with gross gestalten, we may want to skip forward a little bit to endow the system with the capability to immediately begin forming a database of 3D graphs, representing object topologies in the wild.

LFG :

0. Statistical globbing via neural networks should be used as a final optimisation technique, only after all known structural optimisations ("physics") have been ("logically") programmed into the system - not as a lazy shortcut from zero.

1.

1.1. Let us throw away the concept of colour, and begin only with one dimension of magnitude per unit of space - luma / luminance / brightness.

1.2. Let is ignore 3D structure and begin only with the gross phenomenology of the field of vision. 

2.

2.1. Now, mereology : we must have a way to describe subsets of any field which are interesting. The smallest item is a point of brightness.

2.2. SCALE invariance : the concept of an item growing to take up a larger fraction of the field, or shrinking to take up a smaller fraction, must exist. See SIFT, SURF, ORB, et al.

2.3. Location : the field of vision must be referenced by an address space.

2.4. Multiplicity : there may be many items of interest.

2.5. Extension in 2D : items may be linked, giving them 2D structure in the field. See "graph convolution networkS / GCNs"

2.6. ROTATIONAL invariance : the 2D structure is different if you spin it around with respect to the address space. Where you apply this to spinning everything around the system, see "sphere node graphs", though in 2D that's probably a "cylinder node graph".

... at this point we have five to six concepts that must be uniquely addressed in the cognitive space of the system. A sort of introductory transcendental aesthetic.

3.

3.1. Extension in 3D : Now we can revisit the notion of 3D space. 

3.2. Location ("distance") in 3D is associated with scale in 2D (2.2.).

3.3. Occlusion in 2D is associated with rotation in 3D. See "aspect graphs".

... and, I think that's all we need to get starting building a computer vision system, to undergird the visual ontological comprehension of any anthropomorphic computer.

4. Oh yes, we can then revisit colour and motion in the end.

Malaysian public school history textbooks should have complete academic citations

OK - I've figured out the theoretical solution to Malaysia's public education historical sources controversies.

kita kena ajar pembaca buat lit review la ... sebab ramai orang tak tahu sistem petikan / citation yang berdasarkan peer review tu

... come to think of it, zaman kini : besnya kalau kita start setiap buku teks KPN itu ada citation betul-betul pada akhiran buku

Lalang!

Lalang!

Bagi segelintir warga Malaysia, kitorang tiada kepercayaan terhadap perkauman. Perkauman tidak membawa sebarang makna kepada kita, kecuali makna yang dibawa kepada warga senegara kita.

Patut lah kita menahan undi, dalam politik perkauman ( dengan itu, menyokong majoriti secara pasif ) ?

Ataupun patut lah kita mendakwahkan kepercayaan kita, bagi jihad menghapuskan gagasan perkauman secara am ( dengan itu, derhaka terhadap perbadanan yang terkandung dalam perundangan negara ) ?

Di mana tulisnya "kaum" di atas, ada juga segelintir yang boleh menggantikannya dengan "agama". Tapi itu pulak memerlukan penelitian yang lebih rumit. Setiap kepercayaan, walaupun mengenai tuhan, alam, atau orang, atau ketidakpentingan mana-mananya, boleh dikatakan sesuatu yang diikat kepada "agama". Jadi perkara agnostis "agama" adalah terlebih am daripada perkara agnostis per"kaum"an.

loading feedback

Today shall be dedicated to the study of feedback mechanisms, via hormones and nerves, during load/ exercise.

First on my checklist, "various stress hormones need to spike, to their respective activation thresholds, in order to activate certain negative feedback loops". So, loading below certain stress thresholds does not meet this criteria for development.

Second, by the time severe negative feedback is triggered, it is almost certain that many lesser thresholds for positive feedback have already been triggered. These two series of triggers may have an overlapping range, so chronologically it is expected that there will be intermittent triggers of both positive and negative feedback, until the ratio between these starts to skew too strongly towards the negative. This bears observation intra-session of loading.

Third, both positive and negative feedback mechanisms will interact with nutrient availability across the entire timeframe from days before to days after each loading session. As the windows for some of these interactions are so long, loading more than once a week results in a series of overlapping frames. So inter-session observation is also needed.

em-dashing

It's used similarly to a colon, or ellipses. 

I : paused - then paused again ... and again. 

I - paused ... then paused again : and again. 

I ... paused : then paused again - and again. 

I : paused ... then paused again - and again. 

I - paused : then paused again ... and again. 

I ... paused - then paused again : and again. 

Oh, a comma also works in all positions above. Lol

Disability in trust

In trust, there is a gradient of aptitude. 

1. Low context movers, get down to business quickly, keeping risks limited. 

2. High context movers, need to spend a lot of time watching each other, keeping risks limited. 

Aptitude in this case is linked to how anxiety is managed. One of these is a disability, but it is not polite to point it out. 

JEPA is still not inspiring

Finally got around to reading a definition of JEPA. Based on this limited skim, my first impression is that it's still operating on data structures which are too abstract, from a top-down approach, instead of just building a world-view from the bottom, upwards.

I'll keep tabs on it, but of course I don't have time to dive in right now. I have already discussed what I consider to the the basics : anthropomorphic AI must have (1) sensory data structures (2) spatio-temporal modelling (3) hypothesis formation and testing.

Meanwhile the probabilistic relationship between high-level abstract ideas like "dogs in walkers" is simply "less reified" or "more abstract", and basically will display the behavioural inefficiencies associated with reasoning about platonic forms in general, without instantiation.

https://youtu.be/oM4neOyZOi0?si=YFY5qnse8vb3MyWh

Update : oh, hypothesis testing is in the 2022 paper, we're good

Language priorities

 ( a reverse list of what I would prefer to forget first ) : 

• 1. English : 
• Base
• 1. English : 
• Logistics & Business Management
• 1. English : 
• STEM
• 2. Spreadsheets : 
• mainly GOOG, MSFT
• 2. C : 
• because I want access to memory layouts -> in modern terms, it leads to Rust
• 1. English : 
• History & Pop Culture
• 2. JavaScript : 
• as lingua franca, lowest common denominator -> most scripting languages are not far from this
• 3. Malay : 
• because I live in Malaysia and want to participate in citizenship dialogue
• 4. Chinese : 
• because I happen to know a bit of it already

frequency domain analysis

• 1. The simplest technological lay-example is the Graphic Equaliser (GEQ) which lets you change how your music sounds by boosting high or low frequencies
• 2.1. If you search for the origins of such study, you may be directed to Fourier's 1822 CE theorem
• 2.2. However, any student of music will be familiar with the notion of harmony in music - Pythagoras' 600 BCE notes on this occur rather late in the history of human cultures. The Rigveda is dated to 1500-1000 BCE, the Hurrian Hymns to 1400 BCE, and bone flutes and ivory instruments ( that have been found and documented ) go back at least as far as 43,000-40,000 BCE
• 3. All of this is tied back to the phainomechanics of humans - how we perceive the world, as our brains process information, and give it internal representations within our processing systems

the point, orf debate

Some people are debating :

- to kill time

- to be annoying 

- to promote a point of view

- to make a specific change in the world

- to find out the truth

- to find out what their counterparties are made of

There are many reasons to debate. Beware 🤷🏼

phainomechanics, qualia, & quantification

Phainomechanics. I may start using this word to talk about physical mechanics in phenomenology.

Synesthesia. Mysterious, perhaps. Inexplicable? Absolutely not - it may seem odd that a 1D signal like sound translates into a 2D signal like vision. But hold on ... vision is a 2D phainomechanical process on an intrinsically 1D signal. Essentially we animals hallucinate colour as a matter of convention : the three (or four colour) receptors we have, are discretising samplers of a continuous EMF spectrum.

Imagine if sound was discretised in such a way. Not hard at all, with a bit of practical thought. Imagine if smells, pressure, vibration, heat, and pain were too. Actually we do not really need to imagine ... we simply have many of these discretisations, already.

All human qualia are hallucinations from a physical point of view. The intrinsic nature of qualia are therefore most practically characterised as data types / data structures, in an information system.

Calorific timing hypothesis

• - saturate supply ( food ) within waking/W+ 0-to-8 hours
• - saturate demand ( motor load ) within W+ 4-to-10 hours
• - leverage on resulting hormonal yield for desk work within W+ 8-to-16 hours
• - sleep from W+ 16-to-26 hours

bayangan bukan-bukan

Hari2 medsos mengalir kepada kePERLEMBAGAAN berita. Kisah seterusnya tak mungkin- buat bayangan saja. Klo pada 1957, ini permudahannya:

• * Kelas rakyat ada 2 saja: Raja & Marhaen. Tiada konsep bukan/bumi.
• * Undang ada 1 saja- Syariah Berpersekutuan. Rakyat bukan Islam, juga di bawahnya.
• * Kebebasan agama mutlak. Rakyat boleh murtad. Tiada drama keluar/masuk agama.

Saya rasa, tumpuan ekonomi 1957-2026 jauh lebih tinggi, kalau bukanlah kenyataan kita, yang saya cirikan, Kedukaan Sebangsa M'sia 

• ( given a summary of single and dual legal systems in GCC countries )
• Fahamkan anda, Kedukaan Sebangsa Malaysia : bukan disebabkan perkembaran Islam dan bukan-Islam, tapi wujud akibat penambahan kerumitan perkara-perkara perundangan perkauman. 👈yang dipermainkan lagi oleh tokoh politik kita 😕

perisian sumber terbuka & kerajaan Malaysia

Bab perisian sumber terbuka kerajaan ni : masalah teras ialah Malaysia tidak mencita2kan ketamadunan STEM yang berkepimpinan di dunia. Dengan itu, halaju orbit tak tercapai. Unsur2 masalah dasar :

• 1. "membayar pakar > penting daripada menjadi pakar"
• 2. "berdagang > penting daripada membina"
• 3. "menyewa IP > penting daripada memilik IP"
• 4. "pembinaan berdekad/berabad > pembinaan pakai buang"
• 5. "kita jaga kita > kita jaga dunia"
• 6. "rohani membolehkan jasmani > zahir menyempurnakan batin"

🤣🚧💥

-- I'm not the sort of person to be passionate or sentimental about my country. Commentary like this is, to me, the equivalent of picking up rubbish where I happen to live.

a study of my prejudice

I need to work on my underlying prejudices, but maybe I will be carrying them for life, because I am not done with the work which aligns with these prejudices.

1. I am of the notion that people who struggle to stay alive are stupid. (I am stupid too, but at least I am aware of it.)

1.1. Partly I attribute this to people not having a grasp of the context of their environment, so they fall back to animal programming. ( Animal programming is a given, but one should develop the capability to think it on and off, based on available information. Of course, all of that is part of being an animal. Some animals are just smarter than others, to paraphrase a children's book.)

1.2. People who are generally like this tend to be in business, so I am unfortunately prejudiced to believe that being in business is the mark of a stupid person.

1.3. Corporate life is generally working for businesspeople or being paid by businesspeople to do non-business work, e.g., politics, war, art, sport, and charity, so all of these people fall into my domain of prejudice. 

2. I have been aware of my prejudices, but maybe my choice of how to deal with them has not been ideal. 

2.1. My plan as a teenager, was to live with, work with, and study in the most thorough way, all the stupid people above, in order to be able to maximally appreciate their ways of life. 

2.2. In my 20s and 30s, I executed this plan to the best of my ability, despite limited interest. 

2.3. The execution was intrinsically rewarding, but it did not improve my liking for these people. If anything at all, I have been constantly reminded in every interaction for 20 years about how dumb most people are. 

Now on sabbatical, I have some freedom to review my work. That is what I am writing about this minute.