Habeeb Shopeju

8.1K posts

Habeeb Shopeju banner
Habeeb Shopeju

Habeeb Shopeju

@HAKSOAT

I tweet often about Search, Machine Learning, Judo, Cooking, Biographies and YouTube videos. I write at https://t.co/uVvOd4J3ky ML Engrng @ADSKResearch

Lagos, London Katılım Temmuz 2011
2K Takip Edilen2.1K Takipçiler
Sabitlenmiş Tweet
Habeeb Shopeju
Habeeb Shopeju@HAKSOAT·
My father passed on some weeks ago. It has been a challenge to process this loss, but I feel a sense of gratitude. In this article, I write about some of the lessons I learnt from him. I got a flush of emotions doing it, but it was worth it. haksoat.com/lessons-from-m…
English
5
2
16
550
Habeeb Shopeju
Habeeb Shopeju@HAKSOAT·
I should find a way to watch this 🙃 Zhang Xiaojun is a very good interviewer. So... what tool do I use to translate a seven hour YouTube video?
张小珺 Xiaojun Zhang@zhang_benita

@sainingxie 一起挑战7小时播客!他刚和Yann LeCun踏上“世界模型”的创业旅程(AMI Labs)。这是他第一次Podcast、第一次访谈。 2026年2月雪后的一天,我们在纽约布鲁克林,从下午2点,开启了一场始料未及的马拉松式访谈,直到凌晨时分散去。 这篇访谈的中文标题叫做《逃出硅谷》,但他又不厌其烦地枚举了影响他学术生涯的每一个人,并反反复复口头描摹这些人的人物特征(侯晓迪、何恺明、杨立昆、李飞飞…)正是这些,让这篇“逃出硅谷”的对话充斥着人性的温度。 By the way, 下面是访谈的YouTube版本,我们提供了中英字幕。 And yes, 我们是在用播客给这个世界建模😎 A 7-hour podcast with Saining Xie. He has just begun a new journey on world models with Yann LeCun at AMI Labs. This was his first podcast appearance and his first long-form interview. A day after the snowfall in February 2026, in Brooklyn, New York, we started recording at 2 p.m. What followed became an unexpected marathon conversation that lasted until the early hours of the morning. The Chinese title of the interview is “Escaping Silicon Valley.” Yet throughout the conversation, he patiently listed the people who shaped his academic life, repeatedly sketching their personalities in vivid detail: Hou Xiaodi, Kaiming He, Yann LeCun, Fei-Fei Li, and others. These portraits are what give this “escape from Silicon Valley” conversation its human warmth. By the way, the YouTube version of the interview is below, with Chinese and English subtitles. And yes, we are using podcasts to model the world 😎 A 7-hour marathon interview with Saining Xie: World Models, AMI Labs, Ya... youtu.be/rIwgZWzUKm8?si… 来自 @YouTube

English
0
0
0
26
Habeeb Shopeju retweetledi
Kyle Chan
Kyle Chan@kyleichan·
I’ve been working my way through this epic 7-hour interview with Xie Saining at AMI Labs. I also asked Gemini to give me top 10 takeaways. Biggest ones are that he turned down Ilya twice and believes world models, not LLMs, are the key to AGI. 1. Non-Linear Path to AI & Academic Freedom Xie emphasizes that his journey wasn't a standard, hyper-competitive path of a "genius." During his time in Shanghai Jiao Tong University's ACM class, his "highlight" was playing video games in his dorm, teaching him the value of unstructured exploration over rigid academic competition [11:46]. He believes the best research is never linear; if a project ends exactly how you initially planned it, it's likely a "boring idea" [02:09:58]. 2. Rejecting OpenAI & Ilya Sutskever (Twice!) In 2018, Xie turned down a job offer from OpenAI in favor of Facebook AI Research (FAIR), which led to an angry phone call from Ilya Sutskever [01:21:04]. More recently, he declined an invitation to join Ilya's new startup, SSI, because of a fundamental philosophical disagreement: Ilya believes vision is a "solved" problem and focuses on language, while Xie believes vision and physical world modeling are the true frontiers of AI [01:25:57]. 3. Silicon Valley is "LLM-Pilled" Xie argues that the tech industry is currently hypnotized by Large Language Models (LLMs) [05:46:51]. While he acknowledges LLMs are revolutionary communication tools, he insists they are not true "world models" because they operate purely in a digital, text-based space and lack the ability to process high-dimensional, noisy, continuous signals of the physical world [04:29:36]. 4. The Definition of a True World Model According to Xie, a true world model must go beyond text and video generation. It must be a "predictive brain" that understands the physical world, possesses associative memory, can reason and plan, and can predict the consequences of actions in the real world [04:31:32]. 5. Founding AMI Labs with Yann LeCun Disillusioned by the current Silicon Valley narrative that treats AI research as a "finite game" of benchmark-chasing and product cycles, Xie co-founded AMI Labs with Turing Award winner Yann LeCun [05:00:58]. The startup acts as an "underdog" aiming to build true predictive world models based on LeCun's JEPA (Joint Embedding Predictive Architecture) vision, separate from the dominant LLM narrative [06:04:42].
张小珺 Xiaojun Zhang@zhang_benita

@sainingxie 一起挑战7小时播客!他刚和Yann LeCun踏上“世界模型”的创业旅程(AMI Labs)。这是他第一次Podcast、第一次访谈。 2026年2月雪后的一天,我们在纽约布鲁克林,从下午2点,开启了一场始料未及的马拉松式访谈,直到凌晨时分散去。 这篇访谈的中文标题叫做《逃出硅谷》,但他又不厌其烦地枚举了影响他学术生涯的每一个人,并反反复复口头描摹这些人的人物特征(侯晓迪、何恺明、杨立昆、李飞飞…)正是这些,让这篇“逃出硅谷”的对话充斥着人性的温度。 By the way, 下面是访谈的YouTube版本,我们提供了中英字幕。 And yes, 我们是在用播客给这个世界建模😎 A 7-hour podcast with Saining Xie. He has just begun a new journey on world models with Yann LeCun at AMI Labs. This was his first podcast appearance and his first long-form interview. A day after the snowfall in February 2026, in Brooklyn, New York, we started recording at 2 p.m. What followed became an unexpected marathon conversation that lasted until the early hours of the morning. The Chinese title of the interview is “Escaping Silicon Valley.” Yet throughout the conversation, he patiently listed the people who shaped his academic life, repeatedly sketching their personalities in vivid detail: Hou Xiaodi, Kaiming He, Yann LeCun, Fei-Fei Li, and others. These portraits are what give this “escape from Silicon Valley” conversation its human warmth. By the way, the YouTube version of the interview is below, with Chinese and English subtitles. And yes, we are using podcasts to model the world 😎 A 7-hour marathon interview with Saining Xie: World Models, AMI Labs, Ya... youtu.be/rIwgZWzUKm8?si… 来自 @YouTube

English
12
64
533
146.7K
Habeeb Shopeju
Habeeb Shopeju@HAKSOAT·
Have you seen project hail mary, question. I really liked it, amaze amaze amaze. 🤖
English
1
0
0
99
Habeeb Shopeju retweetledi
Ali Hatamizadeh
Ali Hatamizadeh@ahatamiz1·
If you’re an AI PhD student just starting out, don't be discouraged by the hype of "autoresearch" automating scientific discovery. It won't. AutoML made the same big promises in 2017, and we all know how that turned out. Ignore the noise. Master the fundamentals and learn to do research from first principles. Trends fade, but a solid foundation is how you will actually thrive.
English
69
151
1.5K
108.3K
Habeeb Shopeju
Habeeb Shopeju@HAKSOAT·
I've been doing the exercises in Weston's "A Workbook for Arguments" lately. I'm surprised how much gaps are in my analysis when I do the exercises. But I can see how thorough I'm getting as I am going through them. I wonder if anyone on my TL has done them before.
English
0
0
2
52
Habeeb Shopeju retweetledi
Moe
Moe@Mochievous·
update: to avoid any back and forth, if you hold a valid UK sticker visa, you can follow these steps to get the electronic copy of your vignette (sticker visa). That way you have both. Ps: note that according to UK immigration your sticker visa on its own is still valid
Timeyin@kokomaestro

gov.uk/evisa/set-up-u… - Create a UKVI account - register using your passport number - you'll need to have the application number from the actual application that you used to obtain the visa on hand. GWF******269 - download the app I mentioned - you'll do some biometric capturing on there using your phone camera. This is all I remember doing.

English
1
40
212
22.2K
Habeeb Shopeju
Habeeb Shopeju@HAKSOAT·
@OnukoguFavour That's sad. Probably a good idea to keep backups if you choose to have your own thing.
English
0
0
1
4
Ozioma Onukogu| 🔜Where The World Takes Me
I can't believe I am inspiring people to write through my Substack articles. Funfact I had a blog for 4 years, then I completely lost it and started again and lost that one when the app closed It took me a long time to pick up writing again... The response has been amazing and I'm so glad I started again. You can read my Substack wildflower.eth.substack
Ozioma Onukogu| 🔜Where The World Takes Me tweet media
English
1
0
2
155
Habeeb Shopeju retweetledi
matospiso
matospiso@matospiso·
I’ll close this opinion-piece by saying that even though I pointed out an aspect of the eval that I’m not satisfied with, I still believe the results are overall convincing and the model seems strong. Research is complex. For a final takeaway, let’s pretend for a minute this whole discussion is about knives, not embeddings. Knives are used for numerous tasks but probably most commonly for cooking - and so there’s a huge market for versatile kitchen knives: chef’s knives. It only makes sense that many consumer comparisons test knives on kitchen prep work - cutting bread, dicing onions, fileting fish, you name it. However, there’s another major group of customers buying knives whose needs require something way more precise than a chef’s knife. Surgeons use scalpels - probably because performing a surgery with a chef’s knife is a bad idea. To choose which scalpel to buy, I assume the procurement team in a hospital would be interested in the scalpel’s performance on, e.g., a cataract surgery task. Back to embedding models. Single-vector models are like chef’s knives - if you need one model to cover all your needs, they are probably your best bet. BM25 is like a scalpel - it’s much more precise but less versatile. For example, it’s not clear how to extend it to modalities other than text. Multi-vector representations are like a hypothetical combination of the two - versatile like single-vector but with much higher precision thanks to the rich MaxSim scoring operator. ⭐️ Here’s, in my view, the main point of the release: @mixedbreadai’s new model demonstrates that it is possible to achieve both versatility and surgical precision in a single model; they forged the combined knife type with very strong performance on a whole bunch of tasks. The results are undeniably impressive and they will inspire others to try out multi-vector representations. But outperforming just chef’s knives on a cataract surgery benchmark will not impress eye surgeons. That’s all. (8/8)
matospiso tweet media
English
0
2
11
846
Habeeb Shopeju
Habeeb Shopeju@HAKSOAT·
@oyinakindele Finishing means you did great. After a while, you'd go without stopping at all.
English
1
0
0
19
Oyinlola Akindele
Oyinlola Akindele@oyinakindele·
I gave up and decided to stop. Brought out my phone to stop the apps and saw I’ve already done 2km. Ahn ahn, I was amazed. So I decided to push through.
English
1
0
1
96
Oyinlola Akindele
Oyinlola Akindele@oyinakindele·
Decided to step out of the gym today and do a 5km run. Glad I made it through cos halfway, I almost gave up. 😂 It’s my third run ever and I’m proud. So you guys, how did I do?
English
4
1
20
967
Habeeb Shopeju retweetledi
Habeeb Shopeju retweetledi
Adepeju Jaiyeoba
Adepeju Jaiyeoba@pjlalah·
Let me tell you about the picture in Frame 1. In August 2017, as a litigation lawyer in Lagos, I had work to do that required me to be at the office during the weekend. There was only one problem that weekend. I had a 16 months old son, no nanny and no one to leave him with at home so I asked my boss if I could bring him with me and she was gracious enough to agree. While I was working, my son fell asleep and I had to lay him down on the floor beside my table. In that moment, I took a picture of him sleeping and made a promise to myself that if I ever had my own firm, at some point in that journey, no woman working with me who had a caregiver emergency and needs to be at work will have to lay her baby on the floor. That same day, I thought of my struggles as a breastfeeding mother in the workplace and promised again that no woman working with me who had a baby will ever have to struggle for a private space to use her breast pump at work, relieve her aching breast to focus and save some quality milk for her baby. Frame 2 Today, we finished the renovation of our office building complete with a nursing room. Litigation is tough for women. The firm should not make it tougher. Strichland LP has a nursing room and today, by the grace of God, I kept that promise. ❤️ #WomenInLitigation
Adepeju Jaiyeoba tweet mediaAdepeju Jaiyeoba tweet media
English
536
2.6K
10.5K
383.2K
Pawn ♟️
Pawn ♟️@_iamodb·
Jorgensen ke Absolutely bad change to the team
English
2
0
2
262
Habeeb Shopeju retweetledi
Wálé Akínfadérìn
Wálé Akínfadérìn@WaleAkinfaderin·
A colleague who recently went through a series of AI/ML interviews has published this study booklet. It connects foundational concepts with more modern techniques like PPO RL or constrained LLM inference. It's really comprehensive. peymanr.github.io/aiml_interview…
English
7
129
1.1K
69.5K
Habeeb Shopeju
Habeeb Shopeju@HAKSOAT·
Ryan Peterman's interviews are so good.
English
0
0
0
42
Baby Girl
Baby Girl@jesuis_ST·
Baby girl, PhD chemical engineering, imperial college London! هَٰذَا مِن فَضْلِ رَبِّي
English
70
91
1.7K
33K
Adùnréwà
Adùnréwà@Hallyfet·
@HAKSOAT May Allah forgive his sins, overlook his shortcomings and ease your affairs for you.
English
1
0
1
45
Habeeb Shopeju
Habeeb Shopeju@HAKSOAT·
My father passed on some weeks ago. It has been a challenge to process this loss, but I feel a sense of gratitude. In this article, I write about some of the lessons I learnt from him. I got a flush of emotions doing it, but it was worth it. haksoat.com/lessons-from-m…
English
5
2
16
550
Habeeb Shopeju
Habeeb Shopeju@HAKSOAT·
@ridwan_on_x If the PhD is related to the job in mind, it should count as work experience.
English
0
0
4
972
Ridwan
Ridwan@ridwan_on_x·
This is one of my fears starting a PhD now. The job market is currently brutal and I'd rather face it head-on today and—in five years—have 5 years of solid work experience than start job hunting with zero years of experience after spending 5 years on a PhD.
Deepti Sreeram@anthronotes21

Every day as I inch closer to submission, I am petrified thinking of the job market. It is hopeless and beyond redemption. Can't for the life of me encourage students to do a PhD now. Only do it if you have wealth.

English
5
6
41
13.1K
Udo
Udo@dokasto·
@HAKSOAT so sorry man, hope you are keeping well
English
1
0
0
35