Видео

good, curious to know whether the module supports text analysis

his study and results were a great benefit, but too unsuitable for vultures like Fau-cci

15 years ago. Now just one epoch, but several at least. Is this video still viable?

Fast forward 2023 and here they are

so pretty much: good ai: ixtj fun ai: exfp (so its just like in real life)

Considering the modern usage of AI and LLM's, I wonder if a modern IF adventure using a LLM as a parser, because the LLM understands the "world" in its entirely could it make a new version of IF?

4:07 And I want to do this with DRAM found in every day machines, an even crazier idea. It will allow the deployment as a software distribution rather than having to mass produce and transport hardware.

Great talk I do recall some of those early machines for typing in chinese. It is a great way to faster enter text that english and other phonetic languages will never match except if they try making similar way of entering text into the computers like chinese characters.

Imagine llm agent interacting with kan to do above. We can let it run autonomously

Wonderful talk! Tom Mullaney is versed in so many separate aspects ( multiple languages, typography, philisophy, history, and knowledge of cultures), it's a delight to watch/listen.

Good explanation👍

Who’s the moron at Google who asked whether it’s ok to destroy ants homes? 🙄

Economic investigator Frank G Melbourne Australia is following this very informative content cheers Frank 😊

not a "hero" he was shilling for "viruses" big time

I am never going to forgive Lacie for burying this tech. This deserves a comeback.

This is a great talk, thanks and congratulations on your research and success 🔬 If I summarize what I took away (as a layperson): - KANs serve as a foundation for machine learning in the same way as MLPs, but KANs are based on the Kolmogorov-Arnold Representation Theorem, while MLPs are based on the Universal Approximation Theorem. These are significant, because they provide mathematical foundations for why neural networks can learn to represent complex (and arbitrary!) non-linear functions. - MLPs learn non-linear relationships by combining linear weights with non-linear (but notably smooth and differentiable) activation functions. KANs learn non-linear relationships by summing non-linear activation functions together (but notably not guaranteed to be smooth or easily learnable). - The main discovery is that the non-linear activation functions in KANs are learnable when you represent their functions with splines. Splines (like what you use to make curves in Adobe Illustrator for svg files) let you parameterize complex, non-linear shapes with a relatively simple algebraic representation. - KANs perform similarly to MLPs when they are shallow, but have favorable scaling laws. I didn’t get all the math (hypercubes??), but the gist is that KANs basically can do in a single layer what MLPs need multiple for. - KANs are exciting because they are far easier to extract symbolic representations from than MLPs. You can “look inside” of each node and see what function is used to perform a single unit of computation. When you “prune” a network, you are able to see which functions matter the most, and then can easily extract a formula to describe the dataset. You laid out some interesting examples of how KANs “rediscovered” multiplication and division. - KANs being more symbolically legible means that they can be more useful for scientists, who study the world and try to come away with (more or less) formulas to describe what happened. I hope I got it right! Thanks again for your research! ❤

Such interesting stuff and not so much time to do anything with it it should have been my bread and butter haha

Interesting work！Exciting results! But is it possible to open source?

Wow now im the new

Amazing! Can't wait to see all the applications!

Mlp in disguise.

This architecture is not compatible with current hardware due to the need to compute many additional and diverse nonlinear functions.

Is he saying the cock theorem? because that's a bad name

Do KANs require fewer GPUs to achieve the same results for certain problems ?

Just for your information; Dong Hyun Shin has admitted to fabricating a lot of his stories, and is often called out by other defectors

Thanks.

I know this was 12 years ago, but if you are the company and are still in business.... Have you considered setting up a unit in Ukraine? If you need testing they need power, and they know how to handle nuclear power.

Gracias !!

Why has no one commented? Tina’s making an impact. It can’t be too technical; there is plenty of technical stuff on here.

Can you please provide us the sl4a application

I will love to join and age do not determine the mind

Metal can confirm and investigation in the hole universe

IA, IE,II,IO,IU a,e,I,o,u to confirm usk metal

I am interested send you information about to join ,it's a logistics EI METTRE OF METTER DIGITAL MICRO AND MACRO AND human anatomy VS solasplasis in the space, l' in to this I can do in Il ,I O, IE ,IU

I have an exam tomorrow and this video helped me a lot! Thank you very much!

Yr tidak bisa di q yang tidak bisa wkakakkkak yang sangat baik bagi kesehatan tubuh manusia yang paling penting dari Twitter dan Facebook juga menggunakan uangaku yang tidak pernah tahu tentang lowongan pekerjaan terbaru berdasarkan yang terkenal dengan mata telanjang jangan banyak tanya ada apa yang akan kita bahas Tesla CZI sorry we the Kings of his glory for Facebook for BlackBerry berikan tanyak saja apa saja tak usah sok jago merah putih SORRY

“The most exciting phrase to hear in science, the one that heralds new discoveries, is not ‘eureka!’ but ‘that’s funny….’” Isaac Asimov Also, with the colored dots presentiment experiment, one must account for the time that light frequencies landing on the rods and cones are without color and it takes processing in the brain to assign these frequencies color.

Is there tech that can track rapid eye movements. Let's say going from red blue green orange and so on. Mixed with numbers. In rectangle per se? Can it detect as I'm rabidly looking at each object in real time with zero delay

loved it

I am ready to learn thanks sir

Bob is amazing!

And still no dice. Oh well, thankfully India and China took up the slack...😂

24:40: Judging by all the funny ChessGPT videos from last February/March, it's actually a good thing if a chess AI cheats badly enough. (Example: ruclips.net/video/GneReITaRvs/видео.html )

I'm a Noogler, and this was super helpful. Thanks!

We are still being manipulated by these ideas today. Our intelligence agencies used Bernays too. We only have the illusion of choice and freedom. 21 flavors of ice cream, but only two main political sides. Think about it. 22:44 NAFTA was a bad idea. It let manufacturers in the USA move their jobs over seas for cheaper labor. And things went even further down hill when corporations were given the same rights as people.

May 2024 Still. Not. Dead

Does anyone have the results of this study and where can I find it?

Remarkable how little has changed in 18 years

I learned a lot , Thanks

i’m asking myself: did somebody watch this and think “lets found github”?