Mbongeni Ndlovu

The aim of this article is to provide a clearer understanding of what is a "computational sports scientist" The future leaders in sports science will be those with a strong background in computer science, capable of bridging the gap between technology and sports There is a growing need for sports scientists who possess advanced skills in data structures, algorithms, data science, and machine learning. These experts will drive the next wave of innovation in sports Having a deep understanding of both sports science and computer science (and ideally, coaching experience) provides a unique perspective on the tools and technologies that will shape the future of coaching and athletic performance It gives these professionals the chance to build these tools that bring sport much closer to the future to benefit athletes and athletic development teams of coaches

This is next level!

Computer science is gradually returning to the domain of physicists, mathematicians, and electrical engineers as large language models automate much of what we currently call software engineering. The field’s center of gravity is shifting away from manual code writing and toward deeper theoretical thinking, mathematical insight, and systems-level reasoning.

This is the biggest change coming in the software industry: No more Leetcode-style interviews. Tolan is giving candidates a real problem and the AI tools they need to solve it. Then they discuss the solution and how the candidate would productionize it. This tests real skills, capability with actual tooling, and provides the conversation necessary to gauge a candidate's proficiency. I was pro Leetcode-style interviews because they were the best we had. Now the industry has changed and interviews need to as well. Read their article for more info: tolans.com/relay/how-we-h…

One important lesson I hope I can get through to my kids: You'll get better at whatever you spend your time doing. It took me way too long to internalize this.

My team at Google is hiring ML and software engineers in Pittsburgh and Mountain View! We're looking for someone who is: - Curious - High agency - Excited about solving difficult problems If this is you, send me a DM. (If you've already sent one and I didn't respond, ping me again!)

AgentCraft v1 is live ⚔️ Control your agents like it's an RTS game! It's early. It's rough. It's fun. npx @idosal/agentcraft

@Ricburton Good read!

@MillionInt Very true

Sometimes a single person telling you "Aim high, you can do it!" can change your life forever

@giffmana Seen too many of these now

So during phd a good fraction of my lab was working on semantic segmentation, which is a classic computer vision task where you assign a class to every single pixel. The input is an image like the top one below, and the output is an 8-channel image of the same size, with each pixel a softmax vector of size 8 (number of classes in this example dataset). This is typically visualized as an image with one color per class, like the bottom image below. Now, one day, one of my co-phd-students doing his first project in our lab got super excited! His model managed to score more than 95% accuracy! CVPR, we are coming! He started writing an email to the professor full of excitement. Another more experienced and generally pessimistic co-phd-student was skeptical. This was too good. He requested to take a look. He looked at everything carefully, and found the bug: the code was loading the ground-truth image (the bottom image in the pic below) as input to the model! Just a couple letters wrong in the filename... Both for training as well as for testing! The modern LLM version of this (true) story is pretty much what happened here for SWE-bench. The git history was not pruned, so the model simply looked at future commits which contain the solution. This is not the first team this happens to, btw. The lesson here is to always be very skeptical of results that are just a tad too unexpectedly good.

@michelleefang I'm building - would love a sub to accelerate

if you're vibe coding or building over the holidays, i want to gift one of you a 6 month subscription of claude pro to support <3 just drop a comment below. merry christmas!

@parmita Do you have the full video - would love to hear the whole thing

WHY HAS NO ONE DONE THIS BEFORE.

@DenisLabelle Merry Christmas and Happy New Year Denis 🎅🎄❤️

The aim of this article is to provide a clearer understanding of what is a "computational sports scientist" The future leaders in sports science will be those with a strong background in computer science, capable of bridging the gap between technology and sports There is a growing need for sports scientists who possess advanced skills in data structures, algorithms, data science, and machine learning. These experts will drive the next wave of innovation in sports Having a deep understanding of both sports science and computer science (and ideally, coaching experience) provides a unique perspective on the tools and technologies that will shape the future of coaching and athletic performance It gives these professionals the chance to build these tools that bring sport much closer to the future to benefit athletes and athletic development teams of coaches

I'm seeing a lot of "why would you want AI to play games for you" comments. The answer is - of course you don't. That's not the point. Sure there's a couple use cases, like better NPCs and such, but what this is really about is agents. There's all these billion dollar companies spending money trying to automate stuff with LLMs. The issue is, they kinda still suck at being agents. More than you'd expect them to. They're getting pretty good at code, but planning & operating your warehouse logistics operations? Forget it, too unreliable. Of course it's training that's the issue, there's no particular reason to expect models trained mainly as chatbots and text completion tools to be good at that, regardless of how many parameters you increase them to. A chat is a VERY simple environment. Just a user prompting it, all information is generally in the prompt. No need to explore, experiment, deal with the unexpected Obviously this has improved somewhat, tool calls as part of the chat, code agents and such, anything where it's easy to get a lot of text data and kinda fits in the "there's a chatbot and it's helping a user" space So, what to do? How the hell do you get enough data that's challenging, diverse, requires not just single shot reasoning but planning, goal setting, prioritising information, and not just text but full multimodal environments? Games OBVIOUSLY. If you had an agent that could truly dominate a good range of games, it'll hopefully start to generalise to more boring but economically important environments. You don't want a chatbot to run your logistics, you need the best damn simulation of a high skill Factorio player you can get your hands on There's a silly amount of potential tokens available in games, all verifiable, and many that potentially transfer to other tasks. That's why game agents matter more. There's a huge amount of unlocks if you build models to be agents first, chatbots second. World models extend this principle to the environments themselves, which is going to be huge IMO, but in the meantime there's huge amounts of high quality data on Steam waiting to be unlocked

John Carmack on the importance hard work: “For decades, I worked 60 hours a week” “I was never one of the programmers who would do all-nighters or work for 20 hours straight,” programming legend John Carmack begins when asked about his work routine. “My brain generally starts turning to mush after 12 hours or so. But hard work is really important, and for decades I would work 60 hours a week. I would work 10 hours a day, 6 days a week.” He continues: “I had a little thing in the back of my head where I was almost jealous of some of the programmers who would do these marathon sessions. Like Dave Taylor, one of the guys we had at id Software, would be one of those people who would fall asleep under his desk sometimes and all the classic hacker tropes about these things. Part of me was always a little bothered that wasn’t me. I wouldn’t program 20 hours straight because I’m falling apart and not being very effective after 12 hours . . . There are people who can work on 4 hours of sleep and continue to do good work, but there’s a lot of people who just fall apart. I always try to get 8 hours of sleep . . . you can work 100 hours a week and still get 8 hours of sleep if you prioritize things correctly. But I do believe in working hard.” John disagrees with the backlash against hard work and voices support for game developer’s comment that “40 hours a week is kind of a part-time job.” “If you’re doing what you think is important work that you’re passionate about, working more gets more done. It’s really not possible to argue with that if you’ve been around the people who work with that level of intensity.” He believes people who argue that you’re less productive if you work more than 40 hours a week are misinterpreting things: “Your marginal productivity for an hour after eight hours is less than one of your peak hours, but you’re not literally getting less done. There’s a point where you start breaking things and literally going backwards, but it’s not 8-12 hours.” John illustrates this point with a fictional example: “Imagine there’s an asteroid that’s going to crash into Earth and destroy all human life. Do you want Elon Musk and the people working at SpaceX building the interceptor that’s going to deflect the asteroid clocking out at 5pm because they’re going to do worse work if they work another couple hours? It seems absurd . . . It’s the truth: working longer gets more done.” Video source: @lexfridman (Aug 2022)

@gabriel1 thats what i've done too

im taking a break from x i've tried to put less energy into it, but it's pretty clear the general trend is towards it taking up more and more of my attention and mental energy, and is taking up valuable time to think about other things like research

Will try my luck - would be great to work on sports specific evals with a frontier lab

@onlySigz always happy to chat about this if you want

@Mbounge_ A conversation to be had around this.

There’s so much work to be done in sports analytics with the tech we have now I also think the sports industry as a whole should stop obsessing over performance based metrics and capture more data on non performance based metrics

@OfficialLoganK @GoogleAIStudio Would love to test early

Big upgrade to vibe coding in @GoogleAIStudio lands in Jan, but if you want to test early… 👇🏻