One thing I thought was interesting about this paper [1] on understanding LLMs was how the models associate words/concepts in different languages with each other in what they call Multilingual Circuits.

So the example they give:

English: The opposite of "small" is " → big

French: Le contraire de "petit" est " → grand

Chinese: "小"的反义词是" → 大

Cool graphic for the above [2]

So while English is the lingua franca of the interenet and represents the largest corpus of data, the primary models being built are able to use an English dataset to build associations across languages. This might create significantly stronger AI and reasoning even for languages and regions that lack the data, tech and resources to build local models

[1] https://www.anthropic.com/research/tracing-thoughts-language...

[2] https://www.anthropic.com/_next/image?url=https%3A%2F%2Fwww-...

i think this mightve been as early as 2008 even

2007 in fact, as this was the only way to write "apps" for the original iPhone.

Honestly, I think you should go on Shark Tank and find a way to get millions of people to see your product. The products are visually so distinct from anything anyone has seen before. I think they'd do better with a stronger marketing /presentation.

Looks amazing! Good luck!

I'll give you some senior advice too, which is probably a bit of a different direction than what you expect. In regards to this:

> However, I have some strong FOMO about not being able to learn much technical stuff from my peers or seniors. I am trying my best to learn and pick things up on my own, learning design principles, getting code reviews from ChatGPT, etc. But even then, I'm afraid I am not producing the software to the highest standards of the industry since we don't have any rigorous cross-checking, and might be missing out on a lot of learning.

On having the right mentors:

This sucks, and I think the rest of the comments have good advice (look to open source, read more, etc.). I'll also say, think of the upsides. You have a lot of autonomy. You have a lot of latitude to implement your ideas. You have the ability to ship something and see it fail, learn from that, and improve it.

At some of the big companies, there are so many talented people hungry for good jobs or empire building that it’s really hard to have your ideas implemented.

On high-tech quality:

I spent many years working at Airbnb and Amazon. I have friends at Facebook and Google. The honest truth is that all of these places have lots of bad code. The entire industry is held together by duct tape and glue. If Google (clear tech leader of the 2000s-2010s) and Airbnb (Tech-darling built in a post-AWS environment) both have "bad" code, where is there "good" code?

1) Companies that succeed solve business problems and don't focus on fixing tech debt that doesn’t fundamentally matter to customers.

2) Real businesses are complex and don't fit nicely into good design patterns. Every big refactor and rearchitecture I saw at Airbnb started with good design but got complicated once they hit reality.

3) Good engineers deliver good results despite bad code. Take the time to deeply understand your tech stack. Figure out which parts of the code are brittle and figure out ways to improve them. Make the case to your business partners that fixing things will improve developer time, reduce mistakes, and save them money.

And you’re in your first job, not your last job. Take the time to learn the ropes, and when you feel like you've outgrown your role, find a new one. There’s always something to learn.

This is the best advice I’ve read so far.

Real Q is, do you care about oil and gas? Probably answer is no. This sucks too.

The skills you gain from autonomy might translate into, you have a good idea for a new business, you ask your rich boss for a $25k-$100k check investing in your new thing when you leave. I assume it will not be in oil and gas because nobody really cares about mineral extraction sincerely.

This deal sounds nice but in reality there are millions of people your age who get a $25k-$100k check every year from their parents to do something they care about, as opposed to some boring ass business.

Your instincts to find mentors are good. The hard truth is you will have to quit to find them, you have correctly ascertained that you are hitting a ceiling, and developing a relationship with these guys is kind of worthless.

This is the real value of “Google” and “Airbnb:” they’re so big and they’re household names, even if you worked on some bullshit meaningless thing there, you can get a lot of people to pay attention to the new thing you want to do that you are actually excited about. The skills you develop will not be useful for getting hired at these companies, unfortunately that is a huge crapshoot right now, and you’d have to do a career reset by taking grad school or something, because I cannot emphasize enough how little people care about mineral extraction.

Could have nothing to do with Apple's restrictions. Might just be the version of the app they built with. It's very hard to logistically to drop old version of iOS after you support them so oftentimes new apps will launch only supporting the latest version

ELI5, why are GPUs not a commodity piece of hardware that any sophisticated chip manufacturer can't also produce?

This is one of the first time in the tech industry where the value was fully reaped by the hardware itself and not by the differentiated software that ran on top of it

they are very small, very advanced, and production yields are poor. The equipment is also stupidly expensive and takes reams of smart people just to run.

Silicon is currently the most advanced tech humans have ever made and those GPUs are on the cutting edge of that

Amazon has a some AI chips they created themselves. Same with Tesla. They are both behind Nvidia and probably won’t catch up BUT raw performance isn’t always the most important part. Often it’s the cost per compute unit and companies are quickly getting close to Nvidia.

Can more less sophisticated chips perform as well as fewer more expensive chips?

Intel for a few decades didn’t do to shabby.

Yes, they did have great run, but I think over the long run the OS they were tied to was more important

While you are right, the trick is to find the new unturned stone. Google, Facebook and Uber developed playbooks for their respective environments. One cannot replay their playbook and expect it to work

I agree with this

I would argue this is implicitly paid in salary differences

Looking Glass is a scratch-like software designed to teach kids to code in a 3D environment

I don’t think it is quite like a video game. It more allows you to make a scene. Perhaps this can help scratch the itch and can let your child be in the drivers seat

https://lookingglass.wustl.edu/

(I also I used to be an HCI research for the team when I was an undergrad)