Afiq Che Man | This Hormuz thing still on? (@mafqcm) header
Afiq Che Man | This Hormuz thing still on? (@mafqcm) profile picture

🐍 Data science & machine learning. Studying 日本語. Random stuffs / games / anime off-work. Irregular reply sometimes. MPhil @mjiitofficial @utm_my

Malaysia
Joined May 2016
Photo by Afiq Che Man | This Hormuz thing still on? (@mafqcm)Photo by Afiq Che Man | This Hormuz thing still on? (@mafqcm)Photo by Afiq Che Man | This Hormuz thing still on? (@mafqcm)Photo by Afiq Che Man | This Hormuz thing still on? (@mafqcm)Photo by Afiq Che Man | This Hormuz thing still on? (@mafqcm)Photo by Afiq Che Man | This Hormuz thing still on? (@mafqcm)Photo by Afiq Che Man | This Hormuz thing still on? (@mafqcm)Photo by Afiq Che Man | This Hormuz thing still on? (@mafqcm)Photo by Afiq Che Man | This Hormuz thing still on? (@mafqcm)Photo by Afiq Che Man | This Hormuz thing still on? (@mafqcm)
"Don't just skim the surface. Dive deep into ML fundamentals." 2024 🎯
Stuck in a Machine Learning Rut? If you're spinning your wheels in ML, it's time for a change. Here's a no-BS guide to leveling up your ML game. Don't just skim the surface. Dive deep into ML fundamentals. Understand algorithms, not just use them. Basics are your ML backbone. ML is nothing without data. Get good at data preprocessing, cleaning, and visualization. Garbage in, garbage out. Remember that. Sharpen your coding skills. Python, R, whatever your flavor, know it inside out. Sloppy code equals sloppy results. Build stuff. Tutorials are great, but real learning happens when you're solving real problems. Get your hands dirty. ML moves fast. Stay updated. Read papers, follow leaders in the field, and experiment with new techniques. Join ML communities. Share your work, get feedback, and collaborate. Learning in isolation is a slow boat to nowhere. ML is a journey, not a destination. Keep learning, keep growing. The moment you think you know it all, you're obsolete.
5
18
21,377
everything goes back to supervised learning. of course, of course.
7
147
Afiq Che Man | This Hormuz thing still on? retweeted
The more I hang out with top AI researchers, the more I realize how many of them are exploring architectures beyond the standard Transformer. JEPA, diffusion models, Mamba, state-space models, etc. So many radically different approaches being explored in parallel. It seems almost inconceivable that we won’t see a major architectural breakthrough. And by breakthrough, I mean something that delivers better frontier intelligence with significantly less training compute. That doesn’t mean we’ll use less compute overall. Quite the opposite, total compute consumption will keep growing. But we’ll get much better performance for every unit of compute spent.
54
57
6
837
44,542
Afiq Che Man | This Hormuz thing still on? retweeted
Unpopular opinion: The people who aren't intimidated by the "AI will destroy software engineering jobs" hype and continue to learn the fundamentals of computer science and software engineering will be the big winners in the next 2-5 years. Even with all the advances in AI, we will need more software engineers, including junior engineers, as the software market expands.
Students are fleeing computer science. US enrollment in computer and information science programs fell more than 12% for undergraduates and more than 10% for graduate students this fall, per preliminary data from the National Student Clearinghouse. Over two years, undergraduate enrollment is down 18.4% and graduate enrollment 27.1%, reversing years of growth as AI cuts into entry-level coding jobs.
126
119
22
1,980
86,672
Afiq Che Man | This Hormuz thing still on? retweeted
The more I think about Metaphor, more I think how P5/PRemakes fatigue fucked with it. It's genuinely one of the best jrpgs of this decade in everything, but people weren't ready to give it a chance.
“You cannot change the world without power.” Take on formidable opponents in Metaphor: ReFantazio, now available on Nintendo Switch 2! AWAKEN: atlus.link/Metaphor-NSW2
126
477
47
5,582
171,791
Afiq Che Man | This Hormuz thing still on? retweeted
Gotta admire data engineers for (correctly) identifying that ai engineering is just data engineering, and (incorrectly) thinking data engineering will finally get the respect it deserves.
12
17
5
245
9,749
Afiq Che Man | This Hormuz thing still on? retweeted
Replying to @setunasasamura
ppapは6.5億回だからピコ太郎は日本最高の歌手だな。
10
30
4
3,204
271,649
found it a lil funny that nuclear weapons are still relevant in cosmic era literally after nuclear jammer canceller blueprints were found zaft & earth forces immediately pointed nuclear weaponry to one another
81
Afiq Che Man | This Hormuz thing still on? retweeted
追加枠で 米津玄師 King Gnu 藤井風 ヨルシカ RADWIMPSとかないですか?
🔴第77回 #NHK紅白 ⚪️ 出場歌手が決定✨ 紅組20組、白組21組 初出場は、紅組2組、白組5組です。 #紅白歌合戦 公式HP: nhk.or.jp/kouhaku/index.html…
29
184
87
3,749
198,433
Afiq Che Man | This Hormuz thing still on? retweeted
You idiots. We WANT AI to be used for mundane tasks like finding broken code… it wasn’t GEN AI. This is quite literally where we want it to go ahhhhh
Big Walk is being review bombed after House House revealed AI was used on “well under 0.1%” of its code The studio says a chatbot helped fix a Unity bug, with the final solution using around 10 lines of heavily modified code
421
8,364
210
142,007
2,494,471
Afiq Che Man | This Hormuz thing still on? retweeted
Just so you're aware, AMD's CEO is not in support of open models. Jensen Huang was the only man in the room in support amongst all of the other AI oligarchs. His replacement likely won't share his values. I doubt they'll even consider releasing another line of consumer cards.
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models. images.nvidia.com/pdf/Open-W…
46
559
21
5,568
90,736
Afiq Che Man | This Hormuz thing still on? retweeted
Vou antecipar: Ele é quem fazia mais frente ao Dario e Sam pra travar politicas contra as IAs Open-Source que tanto Anthropic e OpenAI apoiam Com ele fora do caminho, estamos a mercê de uma "era das trevas". Vamos ver um corporativismo nunca antes visto Me cobrem em 5 anos
BREAKING: Nvidia $NVDA CEO Jensen Huang resigns.
221
1,685
80
18,460
637,275
good friday after all dapat よくできました time perf rev cuma tula kena improve AI outputs, trim sikit, make it more human readable ✍️
1
3
238
Afiq Che Man | This Hormuz thing still on? retweeted
I never once optimised for my career I just followed whatever I found interesting. If something caught my eye I would go read about it, build something with it, and write about it. Didn't matter if it was useful for my job or not That's why I am as comfortable training models as I am building agentic systems. Evals, memory, agent harnesses, context engineering, inference engineering, RL environments, post training, distributed systems, retrieval etc. All of it came from curiosity and wasn't a roadmap I was following. Curiosity compounds way better than a career plan does in long term
A lot of people have asked how I got a job at a leading Bay Area AI lab. The short answer is pure luck. The long answer is that I showed up every day. I wrote blogs, launched multiple projects, tweeted like there was no tomorrow, and built in public. Increased the surface area of my luck 100x
19
18
7
518
16,899
Afiq Che Man | This Hormuz thing still on? retweeted
Terence Tao says that we still cannot predict how many weights a model needs, or how long training takes, to model a data set well. It is done empirically, with rules of thumb, because some very basic concepts are missing. Communication has Shannon's theory: information comes in bits, with the notion of entropy, and that mature theory is why phones stream megabytes per second and why we know the maximum bandwidth of a spectrum. There is no corresponding concept of intelligence or data complexity. So nobody knows the limits of AI. Models keep getting more powerful as they are scaled, and at some point that will stop, but no one can say where, because the mathematics does not exist yet. Tao says the field will need more mathematicians working on the theory of AI.
11
38
6
209
14,788
Afiq Che Man | This Hormuz thing still on? retweeted
Classic machine learning is BACK, BABY
Introducing e-means, an algorithm beating k-means in 10x fewer epochs. With order of magnitude wall-clock speedups, e-means is released in FAISS and needs no hyperparameter tuning beyond k. It is robust to various scales and datasets. Still skeptical ? Just try e-means [1/10]🧵
25
294
5
5,547
229,475
as seen on threads While all the AI companies were using NVIDIA to train their models, China trained their model to use NVIDIA CEO
1
1
10
295
Afiq Che Man | This Hormuz thing still on? retweeted
Terence Tao confirms the AI-generated Navier–Stokes proof is formally correct (Lean-verified), yet yields almost no mathematical insight. The ~100–165-page text is opaque and poorly written; experts still cannot fully reconstruct the argument.
Terence Tao says that the AI proof on the Navier–Stokes problem is correct, yet mathematicians learned almost nothing from it. The proof is 100 pages, AI-generated and full of issues in its text, but a formal certificate in the language Lean confirms it. Navier–Stokes is a Millennium Prize problem that asks whether the equations for fluids like water can start smooth and then blow up, with velocity or pressure becoming infinite. Human mathematicians came very close, and OpenAI ran its biggest model, with millions of dollars of compute, to get there first. A human proof normally comes with talks on what was tried, what worked and what it connects to, like picking fruit and planting the seeds. Tao calls this one strip mining: the fruit taken, the landscape left barren. One colleague, using lots of AI, could explain only half of it after days of work.
6
6
34
5,165
think it’s nice that in the midst of all these developments in AI, people still pack up to watch comedy made by humans (trust). tho it’s certainly an anecdote on my end since i don’t regularly attend these kinds of shows
77
"data science may be reluctant to use engineer title" me who was an engineer sometime back:
131
Afiq Che Man | This Hormuz thing still on? retweeted
The last thing Sam Altman sees before leaking Astra
144
1,195
89
27,472
628,161