> But even so, I don't think I would have been able to learn much on my own with video lectures, at least not at the start.
This was exactly my situation. Videos can give you a lot of structured, well presented information. And for MIT courses you'd get this knowledge from the very best. The problem is that no matter how well the subject matter is presented, I would hit some conceptual snag that I couldn't resolve just by repeating the sections in the video.
Now, years ago, to clear up the concepts, I would go to math stack exchange, write down exactly what I wanted to understand using mathjax and hope that someone will provide a detailed enough explanation. Most of the time I did learn from the answers, but sometimes the answer would be too succinct. In such cases there would be a need for a back and forth and stackexchange is not really designed around that usage pattern. This hassle would eventually make me give up the whole endeavor.
Now however there are LLMs. They don't need mathjax to understand what I am talking about and they are pretty good at back and forth. In the past 6 months I have gone through 2 full MIT courses with practice sheets and exams.
So I would encourage anyone who went through the route of self learning via videos and found it to be too cumbersome and lacking to give it another go with your favorite LLM.
My only concern with using LLMs to learn new material is being certain that it's not leading me astray.
Too many times I've used LLMs for tasks at work and some of the answers I've gotten back are subtlety wrong. I can skip past those suggestions because the subject is one I'm strong/experienced in and I can easily tell that the LLM is just wrong or speaking nonsense.
But if I didn't have that level of experience, I don't think I would be able to tell where the LLM was wrong/mistaken.
I think LLMs are great for learning new things, but I also think you have to be skeptical of everything it says and need to double check the logic of what it's telling you.
I have the same doubts, it's like the old rule of reading a newspaper story. When it's outside your area of expertise you think they're a genius. When it's something you know a lot about you think it's an idiot.
But it might still help, especially if you think about the LLM as a fellow student rather than as a teacher. You try to catch it out, spot where it's misunderstood. Explain to it what you understand and see if it corrects you?
LLMs are indeed excellent as conversation partners for helping with difficult concepts or for working through problem sheets. They’re really opened up self-learning for me again in math. You can use them to go much deeper with concepts much deeper than the course you’re taking - e.g. I was relearning some basic undergrad probability and stats but ended up exploring a bit of measure theory using Gemini as well. I would go so far as to say that an LLM can be more effective for explaining things than a randomly selected graduate student (though some grad students with a particular talent for teaching will be better).
What the LLM still does not provide is accountability (a LLM isn’t going to stop you from skipping a problem set) and the human social component. But you could potentially get that from a community of other self-learners covering the same material if you’re able to pull one together.
I spent a good minute looking at the exponential in graph, ignoring all the actual data points, thinking to myself that the experiment does show an exponential relation. Where's the lie?
The line is the predicted, not actual. How would you derive that line from plot of noise?
>> I drew an exponential through my noise.
The issue is that there was supposed to be a curve according to his reading, but the actual had no measurable trend. It's possible that the data was measured on the wrong scale. If you zoom out, those noise plots become a line segment. Then again, the predictable line is on the same scale (and we're assuming that it's correct according to his reading or the best he could fit) so zooming out would probably be a different form of lying with statistics via overfitting.
I'll tell you my experience as someone who's been using Math Academy for past 6 months.
Math Academy does what every good application or service does. Make things convenient. That's it. No juggling heavy books or multiple tabs of PDFs. Each problem comes with detailed solution so getting them wrong doesn't mean looking around on the internet for a hint about your mistake (this is pre ChatGPT era of course, where not getting something correct meant putting down MathJax on stackexchange).
> better than just prompting ChatGPT/Claude/etc
The convenience means you are doing the most important part of learning maths with most ease: problem solving and practice. That is something an LLM will not be able to help you with. For me, solving problems is pretty much the only way to mostly wrap my head around the topic.
I say mostly because LLMs are amazing at complementing Math Academy. Any time I hit a conceptual snag, I run off to ChatGPT to get more clarity. And it works great.
So in my opinion, Math Academy alone is pretty good. Even great for school level maths I'd say. Coupled with ChatGPT the package becomes a pretty solid teaching medium.
> the fact is that less people take the IIT-JEET (a couple of hundred thousand), than some of our exams here - like for law school and medical school
Do you have some source backing this up? A cursory search reveals that more than 850,000 applicants attempted the IIT-JEE exam in 2023. Total available college vacancies slightly more than 16,000. This number would be around 5000 for the OG IITs.
For US medical colleges, for 2022-2023 admission cycle, 22,712 of the 55,188 students who applied to medical school matriculated. Perhaps these numbers belongs to a specific US region?
I would say that attempting JEE is now more or less a cultural thing. 90% of the applicants (pardon the number out of thin air) sit the exam because that's just something you do after finishing school.
This is correct to some extant, and not just for getting education. As an example, in many competitive exams conducted by universities etc. to select students, the selection cutoff marks clearly show this pattern of intentional discriminatory admittance.
For someone belonging to "general" category, anyone designated to an upper caste, the cutoffs percentage for selection could be as high as 97-99%. For "reserved" category candidates, those from the lower castes, the cutoff can be as low as 10-20% for the same test. This means someone scoring 95% would not get get an admission offer while someone else scoring 15% would.
This reservation system is also a part of government jobs. I have seen "reserved" teaching positions being filled by candidates who score negative marks in the selection exams. But since no other type of candidate can fill the reserved position, the highest negative score gets the job.
Does this really alleviate the issues borne out of historical oppression of the lower cast, I don't know. Perhaps. Is this overall a good thing for a nation and its people? Again, I have no idea.
Also, the example I have provided are from 15 years ago, when I appeared for these exams. I'm not going to bother finding verifiable information from back then but the figures I mentioned are pretty much what I saw.
You and I are referring to completely different scores. It is true that a candidate needs to have scored a minimum percentage of around 50-60% in their high-school/intermediate board exams (so 10th/12th standard) to be allowed to sit in the selection examination.
The cutoffs I'm talking about are dependent on overall performance of all candidates. In the link I have shared this cutoff is 18% (89/480) for reserved category students. There are more details in the article about the cutoff for reserved category students being 60% of that of general category ones. There is further elaboration on how many students have to drop off because of their poor performance. These vacant seats are rarely filled.
Right, it is. Like I said, the figures I've pulled are from what I remember from 15 years back. Unfortunately I'm not able to find any sources with light searhcing.
I'd like to point out that there is a high possibility that there is an upward trend in reservation category cutoffs. And if it is indeed the case then I'd be the happiest.
Data science requires a very strong mathematical background. Thee are libraries and software that do take care of some of the most complicated processes, but I don't believe someone can become a good data science engineer by always relying on such libraries/software.
Hoe rigorous is the treatment of mathematical topics in the AI course you offer?
Do you teach the concepts of probability/statistics, linear algebra and calculus required for the course, together with some testing or examination relevant to the subject material being taught? Or is your approach similar to Andrew Ng's Coursera course where he does give some introduction about the maths involved without going into details because they are not required, resulting in acquisition of, at times, half baked knowledge about core concepts.
Your observation is on point - a strong understanding of maths is crucial for data science engineers. The topics you've mentioned - probability/statistics, linear algebra, calculus- are covered in our course, and our learners are expected to build a solid understanding of them gradually.
To ensure effective learning of these topics, we space out these topics over nearly all of the course. It's a change from our initial version where we've had it concentrated early in the study - however, we saw that such an approach was quite demotivating to many learners.
Next, we use spaced repetition. This happens during our daily standups and project reviews, where senior team leads (expert data scientists) regularly ask questions about topics that might have been covered in previous modules. The questions also tend to focus on understanding (e.g., why something is relevant, how it can be used in business situations) rather than simply recalling formulae or definitions.
Compared to Andrew Ng's Coursera course, we require our learners to understand these topics deeper. However, upon graduation, the level of most learners will be less than PhDs graduates, who spend half a decade on learning these topics. Nevertheless, our students will have strong practical skills to do data science, which isn't properly taught in academic data science education (we hear this a lot from hiring partners). What we have as a key goal, however, is to give our graduates enough understanding to:
a) be able to work on junior-level tasks effectively from day 1. The libraries you mentioned are helpful here, even though not enough on their own;
b) develop the capacity to continue learning maths-related topics independently so that the learner, even after graduation, can continue getting better at maths and feel comfortable with it instead of fearing it
What would you suggest as a starting point for a regular software engineer who'd like to look into this? See how everything comes together at a high level I mean? Because to someone on the outside, the field of game development appears to be very daunting.
There are a lot of libraries in different languages that are sort of in a middle place between drag n drop engines and programming straight to a graphics API like Direct3d or Vulkan. You should look around for a popular one in a language that you're already comfortable with and read up on tutorials, documentation, and such first.
For reference, the ones I used to love and seem to still be kicking were Phaser.js for web games and Libgdx (on Kotlin) for mobile/desktop. Both were incredibly fun to learn, and I went at it very slowly on my own pace with a clear, small-scoped, end game goal each time. It was very hard but worthwhile to not let myself fall to scope creep, and after a couple of completed prototypes I felt ready to tackle 3d. I got some projects on C# that used Direct3d to draw things with triangles and went nuts trying to implement the concepts I read about in books like realtimerendering.com
I know it all looks daunting at first, but always remember that it's supposed to be a very long, very fun journey, meant to satisfy your intellectual curiosity at every step of the way. Games and graphics are one of the most rewarding problem spaces since every time that you learn a new concept and get it working, you get to see it and play with it and it's the best feeling ever!
And while I'm here I want to share the most wonderful tutorial on how to build hexagonal grids for games and things, in case you want to do something hexagonal at some point like I did, from what I think is one of the best educational resources I've ever seen: https://www.redblobgames.com/grids/hexagons/
Starting with something easy, like 2d animations in browser using libraries like pixi.js. From there you can expand on many topics, just focus on one at once, be it shaders, networking (for multiplayer games) or whatever might interest you.
You normally get a blue card with a period of validity, which is normally 4 years. While the blue card is valid you can freely switch between organisations within EU. The only restriction is that you cannot change the field of your work. So a software engineer can only apply for other software engineering positions.
To get a blue card you primarily need an unlimited contract. Companies do help with the process of procurement, like setting up an appointment. Once you get your blue card you also get another "green slip" that mentions your current employer along with your job role. Anytime you move to a different job, this slip needs to be updated, which is then done by your new employer.
After playing it for about 5-8 hours in the last 2 days, I'd say for a casual gamer like me, there is no perceptible latency. Even in an FPS like Destiny. But again, yesterday was the first time I played a AAA game after two years so maybe I'm just a bit rusty.
You don't know what you're missing until you have tried lower latency.
My regular gaming setup is a 144 Hz monitor with a PC. Everytime I try gaming on a console with a 60 Hz TV I feel like I'm completely drunk because of the noticeable input lag. And that's not even cloud gaming. That's just the Console + TV input lag...
This was exactly my situation. Videos can give you a lot of structured, well presented information. And for MIT courses you'd get this knowledge from the very best. The problem is that no matter how well the subject matter is presented, I would hit some conceptual snag that I couldn't resolve just by repeating the sections in the video.
Now, years ago, to clear up the concepts, I would go to math stack exchange, write down exactly what I wanted to understand using mathjax and hope that someone will provide a detailed enough explanation. Most of the time I did learn from the answers, but sometimes the answer would be too succinct. In such cases there would be a need for a back and forth and stackexchange is not really designed around that usage pattern. This hassle would eventually make me give up the whole endeavor.
Now however there are LLMs. They don't need mathjax to understand what I am talking about and they are pretty good at back and forth. In the past 6 months I have gone through 2 full MIT courses with practice sheets and exams.
So I would encourage anyone who went through the route of self learning via videos and found it to be too cumbersome and lacking to give it another go with your favorite LLM.