nutanc

Ok. Guess its different in the US when compared to India. Looks like US system does tie your hands down when it comes to healthcare based on your explanation.

I think it all boils down to goals. If we need a standard health care and standard education, we will need standardization. If we need to handle special personal cases, we need special treatment. Things can be improved, but frankly I am amazed at how process oriented our education and health care has become.

Follow the steps specified by a school, get the minimum marks specified, that means you have learnt the basic skill set needed by you to function in society. If you need to learn extra, there are other channels available.

Same thing with health care. A patient shows some numbers, treat those numbers. If something feels off, the doctor takes the call. That's why we still have human doctors and not AI.

Also, one major difference between education and health care, in education the customer is different from the one who pays. So the goals get differently from the start(make my kid the best mathematician-parent(paying), I want to be a footballer-kid). So who should the school listen to :)

This is cool. Actually if bluesky can do this automatically where it can fetch RSS and show as handle updates it will be really useful. Will help a bunch of people who wouldn't need to maintain the services.

Or maybe there is a potential for a SaaS service?

Agreed. I assume this will open up Bluesky to a lot of potential legal problems. But will it be any different from accesing the content using the app as the content is anyway hosted.

That said, just the other day I was thinking, is the reverse possible. I have a web site/blog. Use RSS and then the RSS updates are posted to a handle on Bluesky. I would assume that's a lot more useful?

That's not a problem of the channel(call) right? That's a problem of personnel.

Had launched something like this in 2016. We had called it as ering.me, so you could have an url like ering.me/handle. Used it in email signatures, web calling etc. It didn't pick up at that time or we didn't market enough :)

Hope the market is mature now and products like this succeed. All the best.

We actually do that. We have just released a web widget which maps to your current phone number IVR or voice bot flow. Have put a quick link on our web site at the bottom left corner, not clean ,but works :)

https://ozonetel.com/ind/

It's a new approach I am experimenting with.

https://gpt3experiments.substack.com/p/a-new-chunking-approa...

Let's keep aside the hype. Let's define more advanced AI. With current architectures, this basically means better copying machines(don't mean this in a bad way and don't want a debate on this. This is just my opinion based on my usage). Basically everything in the Internet has been crammed into the weights and the companies are finding it hard to do two things:

1. Find more data.

2. Make the weights capture the data and reproduce.

In that sense we have reached a limit. So in my opinion we can do a couple of things.

1. App developers can understand the limits and build within the limits.

2. Researchers can take insights from these large models and build better AI systems with new architectures. It's ok to say transformers have reached a limit.

Experimenting with creating semantic chunks of large podcasts. Got the following chunks, https://gist.github.com/nutanc/a9e6321649be5ea9806b4450b0bd6...

Dwarkesh has 18 splits. https://www.dwarkeshpatel.com/i/151435243/timestamps

I got 171. So roughly 9 context discussions in one time stamp.

Since I am a developer I have always created fun games for my kids to play and learn. Almost every single time they just go with the game mechanics instead of the actual learning concept. The only game which worked was a roblox game[1] I created for fun. And that too, its just a basic obby. Nothing fancy. But I told my kid, he can play Roblox for 30 minutes if he can finish this obby for 5 numbers. So, in a way, had to force him. But within a week, he was able to memorize the multiplication tables :)

[1]https://www.roblox.com/games/10489578087/Multiplication-Game

I wouldn't say it has failed completely, but yes, it has not delivered on the promise. Edtech and teaching is close to my heart and I have spent a lot of time analyzing this. My insights are mostly from what I have observed in India.

Edtech has failed in India because there is very little tech and even little education in the apps. Mostly edtech in India equates to exam preparation apps.

Other approaches like the duolingo gamification approaches have also failed to actually educate users though some of them have found success as startups.

This has lead to a disillusionment among parents about the edtech promise. Also, unfortunately the big successes of edtech(Byjus, Whitehat etc) turned out to be bad apples.

But lets take the positives. The main positive is that parents and kids are willing to try and there is a huge market. They just need the right product.

My takeaways are as follows:

1. Too many products try to reinvent the wheel with their own curriculum etc. This creates an extra burden for students who have to study these in addition to their course work. Given the paucity of time, guess what they drop?

2. Too much gamification and engagement farming. The emphasis on gamification instead of learning.

3. Most edtech is about putting videos online and adding quizzes. No emphasis on making the content actually interactive and engaging. In this day and age why shouldn't text books be interactive.

This is what I am focusing on and helping my wife do a startup in this space(I am unfortunately busy with running a SaaS startup, so I just help out of passion :))

So far, we have made the whole of CBSE(India's national curriculum) curriculum for math as an interactive textbook. But it's a long journey. There is no hurry. The revolution has just started :)

I would start simple. Build it and share it with your friends and family. If you are in a college, share it with your college. Solve the problem for your closed group first. If you can convince your group to pay, you have crossed the first step. So first degree of separation is solved.

Remember, most celebrities will be at most at a distance of six degrees of separation. So you can build up to that.

You are right. The wording was wrong. It should have been, "The usage of this app will be harmful for your mental health".

This is awesome. I have been telling that social media is like smoking. When cigarettes came, even doctors were advertising the benefits of cigarettes. Now we know the harmful effects. Same is the case with social media. We just dont know they harmful effects completely yet.

Ban this. I am addicted and can't stop. Or put a warning on social media apps like they do on cigarette packets. Using this app maybe harmful for your mental health.

A little off topic, but if you need a bot for a meeting, do you even need a meeting?

Yes. We have integrated this into our RAG pipeline. Looking for a benchmark which we can try. Have been using this to give feedback to my marketing team on our blog posts.

Also, as I mentioned, now, trying out experiments in the video domain with some pretty cool results. Will post the details this week.

https://x.com/nutanc/status/1849661242027409723

Just a quick perusal, but this looks much better covering datasets also. If we apply this definition strictly, what are the real open source AI models? Allen ai is one?

This is really cool.

Currently, other than my day job, I am obsessed with making sense of the "shape of stories". Mapping embeddings and see how they move through a sequence.

Started of with [1] which showed that there might be some strength to the idea.

Applied it for chunking [2] and web site analysis[3] and got pretty good results.

Just started trying out experiments on video [4] and surprised that the structure seems to hold for image embeddings as well.

I have no clue if this has any value, but it is fun to go down this rabbit hole :)

[1]https://gpt3experiments.substack.com/p/the-shape-of-stories-...

[2]https://gpt3experiments.substack.com/p/a-new-chunking-approa...

[3]https://gpt3experiments.substack.com/p/using-semantic-chunki...

[4]https://x.com/nutanc/status/1849661242027409723

That's possible with the brilliant.org model right. Though the lesson is structured, it's possible to jump to any lesson and brush up.

Agreed. I think there are enough tools that "explain" a concept. Not enough tools to apply those concepts and test. Good point.

Also agree on giving two hoots about students. This is one of the reasons edtech generally is hard to monetize. If you monetize, you come off looking like making money off kids. If you don't monetize,how do you make the best product. I think the Khan academy approach of a non profit makes the most sense. Commercialization of education through edtech has had disastrous effects atleast in India(ex: Byjus etc)

This is really good. Tried out the cloning. It sounded very similar to my voice. But then I did a blind test with 5 people. All of them didn't recognise it as my voice. So is there a bias when we listen to our own voice?

This is true. This is the reason, in many of our experiments we find that using a new algorithm, KESieve, we actually find the planes much faster than the traditional deep learning training approaches. The premise is, a neaural network builds planes which separate the data and adjusts these planes through an iterative learning process. What if we can find a non iterative method which can draw these same planes. We have been trying this and so far we have been able to replace most network layers using this approach. haven't tried for transformers though yet.

Some links if interested:

[1] https://gpt3experiments.substack.com/p/understanding-neural-...

[2] https://gpt3experiments.substack.com/p/building-a-vector-dat...

For my kid, I created a Roblox game so that he can learn multiplication tables by heart. When he started, he would take around 64 secs to complete the obby. When you know the tables by heart, it takes around 14-18 secs for the obby. Have tried with multiple other kids. This is a fun way to get the kids to learn the tables and for us to know if the kids have learnt the tables.

Yes. We are using rolling sentences. Not just pairs. But the whole story. I have written a little about this in my previous post, shape of stories, https://gpt3experiments.substack.com/p/the-shape-of-stories-...

So we can see how as we are appending sentences and getting the embedding we get the movement of the story in the latent space.

Hey, thanks for taking the time to test this out. This has helped us identify a bug in our system(the sentence splitter was rudimentary, similar to LLamaIndex semantic chunker). We will update the API soon. Below is the output we get after fixing the bug.

Segment 1: The programmer, an offshoot of the great ape family, closely related to chimps and gorillas, is distinguished by its minimal bipedal movement and ability to stare at a computer screen for the majority of its lifetime. There's an estimated 30 million specimens alive in the world today. Normal humans use stereotypes to help understand and generalize this unusual variant, which experts estimate are about 99% accurate about 12% of the time. In today's video, we'll take a look at 10 different programmer stereotypes to find out which one you fall into. First up, we have the gear head. This variant owns the bleeding-edge version of everything, like the latest M1 Mac, a big-ass curved monitor, mechanical keyboard, Tesla in the garage, AI-generated synthetic meat in the fridge, and a smart lock on the house to keep it all safe. When programming, he goes wherever the hype train takes him. In '96, it was Java; in '06, it was jQuery; in 2016, it was GraphQL; and in 2026, he'll be first in line at Neuralink to get a chip that can help him write blazingly fast code. It doesn't matter what the tech does—if it's trendy, it belongs in the stack. This stereotype may be true sometimes, but programming can actually push many people in the opposite direction. The guy who works in tech but hates tech stereotype knows exactly how unreliable and dangerous code can be, like that RAC25 incident where a little software bug accidentally killed some people by giving them a massive overdose of radiation. This guy would never buy a car that can be remotely summoned back to Elon when you stop paying the bill, and he would definitely never put a smart lock on his house because the NSA probably has backdoor access—or at the very least, there's an undiscovered exploit in its code.

Segment 2: If you broke into his farmhouse, you'd find a single-monitor Linux machine, a flip phone, some gold bullion, and a shotgun barrel pointed in your face. The most stereotypical programmer, though, has to be the introvert. He's a savant who still sleeps in a car bed, and his vision of the ideal lifestyle is what the rest of society calls quarantine. He's super good at math and can actually program stuff without using Google and Stack Overflow, but couldn't hold a conversation to save his life. Extroverts like Jobs use these nerds like Woz to get super rich. This stereotype used to be 100% true back when programming was hard, like pre-1990s, but as programming has become more mainstream, it's led to a new paradigm: the programmer bro. This guy got a computer science degree while mostly partying with his frat in college. His name is usually Chad, and he has more mating opportunities than the introvert, but it comes at a cost of reduced code quality, which he refuses to test because test-driven development is for losers.

Segment 3: Nah, bro. However, he has better communication skills than the introverts, which is annoying because I wish this guy would stop talking to me. Eventually, he evolves into your manager, where he can torment you with code reviews and team-building exercises. Now, so far in this video, I've been using a lot of masculine pronouns. That's because 95% of my audience is male, which is actually pretty close to the real-world distribution today. What you may not realize, though, is that back in the day, women used to dominate the programming space. Kathleen Booth created the first assembly language, Grace Hopper created the first compiler, and Margaret Hamilton led the team who wrote the code for the Apollo Moonlander—code that was so flawless and perfectly executed that some people think it's proof we didn’t actually go to the moon. That was the apex of code quality. Since that time, everything's gone downhill. The next specimen we'll look at is the influencer, or "code-fluencer." His natural habitat is not a code editor, but rather a social media platform—most commonly Twitter. After figuring out how to print "Hello, World!" in PHP, he immediately rose to the top of the dominance hierarchy in his own mind.

Segment 4: Now, he makes the world a better place by regurgitating code tips and hot takes all day long, and he just landed a better-paying job than you because he mastered the art of virtue signaling. And that's what we call a good culture fit. Another popular stereotype is the hacker. This guy's able to open up a terminal, connect to some remote mainframe, and break all of its security protocols one by one, with awesome fancy animations between each step. This stereotype is what most people think programmers do but is 100% manufactured by Hollywood. Real hacking is extremely tedious and boring and is done primarily by the people who have all the guns. Now, a stereotype that is actually real is the 10x developer. This guy is an extremely rare unicorn that can do the work of 10 other developers combined. Some say they're a myth, but I've seen developers firsthand who write code like Durant plays basketball or Kasparov plays chess. There are people out there with a natural problem-solving ability that just goes far beyond the rest of the population. You'll know a 10x developer when you see one because you'll feel very incompetent and also very jealous. Now, I think the ideal stereotype for most of us to fall into is the lazy programmer. To the outside world, it doesn't look like this guy does much.

Segment 5: He sits at a computer all day hitting the keyboard, and if you glance at his screen, it looks like he's just copying and pasting things from the internet. What he's actually doing, though, is building a million-dollar side hustle so he can retire in his 30s. He also has a remote job with a 400k salary, but he eats ramen for dinner while sharing a crappy apartment with four other dudes. His wardrobe is 50% swag from tech conferences and 50% things his mom bought him.

Segment 6: He leverages code to work smarter, not harder. Now, on the other end of the spectrum, we have the old jaded guy. He has long silver hair and a big white beard. He only codes in C—not C++ and definitely not any of the hipster garbage that you're using. In fact, he probably wrote the compiler for the silly toy language that you're trying to learn. His depth of knowledge transcends the normal ape's idea of reality when he discovered, through psychedelics, that we're all just one entity that found a hack in the universe to experience itself in parallel with primate bodies—and computers are the tool that will ultimately make us one again. And that concludes our presentation on programmer stereotypes. Let me know which one you fall into in the comments below. Thanks for watching, and I will see you in the next one.

Possible titles can be:

"Introduction to Programmer Stereotypes and The Gear Head"

"The Tech-Hating Programmer and The Introvert"

"The Programmer Bro and Women in Programming History"

"The Code-fluencer and The Hollywood Hacker"

"The 10x Developer and The Lazy Programmer"

"The Old Jaded Programmer and Conclusion"

I am comparing with ChatGPT and I see that our chunking can be a bit better to break up into more sections, for example "The Code-fluencer and The Hollywood Hacker" can be broken into "The Code-fluencer" and "The Hollywood Hacker"

Will be sharing the code next week. The basic idea is to find the embeddings of sentences and then finding the distance in the latent space to see if there is too big a jump in context.

No astroturfing here. Atleast I am not doing it. If somebody is doing it on my behalf, why?

2d reduction is only to visualize the flow of an article. When we chunk, we chunk on the full dimensions. Sorry, the article does not have more details. My bad. Will add more details along with the code doing this also.