Barry Ugochukwu (@small_chuks) header
Barry Ugochukwu (@small_chuks) profile picture

Data scientist, AI & ML⚡// Agric Econs// Technical writer📝// love books❤// Chess♟️ was my first love. This profile belongs to an amazing individual 🥰

Joined November 2019
Photo by Barry Ugochukwu (@small_chuks)Photo by Barry Ugochukwu (@small_chuks)Photo by Barry Ugochukwu (@small_chuks)Photo by Barry Ugochukwu (@small_chuks)Photo by Barry Ugochukwu (@small_chuks)Photo by Barry Ugochukwu (@small_chuks)Photo by Barry Ugochukwu (@small_chuks)Photo by Barry Ugochukwu (@small_chuks)Photo by Barry Ugochukwu (@small_chuks)Photo by Barry Ugochukwu (@small_chuks)
Open market is the next best thing that will happen to you as a trader after profits. Everything is literally free! Drop everything and check them out today! @openmarket_xyz
575
Nature agrees, millions of seeds. Few trees. But I won’t call it a waste…. It’s data.
"99% of all effort is wasted."
18
Call it opportunity cost. The more options you have, the more important it becomes to choose wisely.
The irony is that competence solves one problem and creates another. At first, the challenge is becoming valuable. Later, it’s deciding what deserves your attention.
14
The irony is that competence solves one problem and creates another. At first, the challenge is becoming valuable. Later, it’s deciding what deserves your attention.
Skills increase opportunity, which can increase distraction if you're not intentional with your time. That's why focus has to level up with competence.
1
40
Barry Ugochukwu retweeted
One advantage of having many distinct sovereign nations instead of a single world government is that they can try different policies and we can see which work best.
1,175
2,212
289
17,614
1,831,399
Barry Ugochukwu retweeted
curious about the training data of OpenAI's new gpt-oss models? i was too. so i generated 10M examples from gpt-oss-20b, ran some analysis, and the results were... pretty bizarre time for a deep dive 🧵
119
481
80
5,438
1,008,610
The crazy thing about this is that it's not a reasoning model.
LMAO!!! No one was ready for this! Did Google just dropped open-source SOTA??? Google Gemma 3 is thrashing o1-preview and o3-mini-high and it's only 27B parameters. The second-best open model, only behind DeepSeek-R1 FREAKING NUTS 😂
1
70
Man, AI update is moving at warp speed—I can’t keep up. Years of improvement has happened in a week!
32
Randomly inserting "let me know if you have any other questions!" at the end of code docs is our equivalent of "thoughts and prayers." I KNOW YOU HAVE OTHER QUESTIONS, the answers just aren't here! 🥲
1
1
40
Man, it feels like just a few months ago everyone was saying 'RIP Google' thanks to OpenAI. Now, it's 'RIP OpenAI’. Could we be saying 'RIP Deepseek' before long? Very likely.
1
1
83
Barry Ugochukwu retweeted
Which NVIDIA GPU is best for your AI, ML, or high-performance computing needs? @small_chuks compares four of NVIDIA’s powerhouse GPUs: the A100, L40s, H100, and the new H200 GH Superchip. Find your best match today in this blog: civo.io/3DGFWzm
1
1
4
315
Barry Ugochukwu retweeted
Replying to @emollick
lol this was fun
2
1
19
479
One underrated skill for developers: naming things. Variables, functions, classes, even PR titles. Clear names = better collaboration, fewer mistakes, and future-you thanking past-you for not calling everything data123
29
In the end, every great project boils down to this: •Understand the problem. •Break it into small parts. •Tackle one piece at a time. The rest is just tools, frameworks, and syntax.
31
Barry Ugochukwu retweeted
Which NVIDIA GPU is best for your AI, ML, or high-performance computing needs? @small_chuks compares four of NVIDIA’s powerhouse GPUs: the A100, L40s, H100, and the new H200 GH Superchip. Find your best match today in this blog: civo.io/3CczmzM
2
5
347
Think Nokia
Imagine being Sundar Pichai now: - you had the largest continually updated data set of any company to train AI on (the Google Index) - you invented the underlying technology of LLMs like ChatGPT in 2017 called Transfomers - you had complete search dominance: all you had to add was AI and you'd own the market And yet: - you managed to complete fumble your massive head start and was late to everything - you made your APIs so hard to use nobody seriously integrated it into their apps and people instead went Anthropic and OpenAI - you now see your search dominance quickly slipping away to Perplexity and yesterday's launched ChatGPT Search This will be a business case studied in universities for decades
52
Barry Ugochukwu retweeted
Worth repeating: Do not confuse retrieval with reasoning. Do not confuse rote learning with understanding. Do not confuse accumulated knowledge with intelligence.
Replying to @VictorTaelin
mini rant: the illusion LLMs are intelligent comes from their massive scale. it is hard to visualize, but these things memorized the WHOLE internet. everything you've ever asked it, either has been solved before, or is a simple combination of existing solutions. but that's still an illusion. when LLMs face a problem that require a NEW solution - one of original shape, one it has never seen before - they fail. it is that simple. that's what my example shows. i took as simple problem - inverting a binary tree - and added a few constraints to make sure the solution is unique and not in the dataset, forcing it to actually SOLVE it itself. and, surprise - it can't! and I must stress this isn't about THIS problem. but about all. LLMs can't solve ANY problem, at all. it can only spit a memorized solution. if nobody posts this solution online, not even GPT-6, opus-5, or o3, will be able to solve *this very prompt*. I'm betting on that. the inability to create new solutions imply LLMs won't invent new science. yes, they will completely change the world as we know it. they'll have a higher impact than computers and the internet. but, unless a new kind AI emerges, we're still on our own when it comes to curing cancer or making superconductors.
234
856
101
5,745
621,331
Barry Ugochukwu retweeted
LLMs are trained on much more than the whole Internet -- they also consume handcrafted answers produced by armies of highly qualified data annotators (often domain experts). Today approximately 20,000 people are employed full-time to produce training data for LLMs.
Replying to @VictorTaelin
mini rant: the illusion LLMs are intelligent comes from their massive scale. it is hard to visualize, but these things memorized the WHOLE internet. everything you've ever asked it, either has been solved before, or is a simple combination of existing solutions. but that's still an illusion. when LLMs face a problem that require a NEW solution - one of original shape, one it has never seen before - they fail. it is that simple. that's what my example shows. i took as simple problem - inverting a binary tree - and added a few constraints to make sure the solution is unique and not in the dataset, forcing it to actually SOLVE it itself. and, surprise - it can't! and I must stress this isn't about THIS problem. but about all. LLMs can't solve ANY problem, at all. it can only spit a memorized solution. if nobody posts this solution online, not even GPT-6, opus-5, or o3, will be able to solve *this very prompt*. I'm betting on that. the inability to create new solutions imply LLMs won't invent new science. yes, they will completely change the world as we know it. they'll have a higher impact than computers and the internet. but, unless a new kind AI emerges, we're still on our own when it comes to curing cancer or making superconductors.
67
291
50
2,619
376,390
Barry Ugochukwu retweeted
ML twitter this morning 😂
24
519
59
5,082
267,602