yes, it was a really short meeting. They ask Deepseek if he was trained on H100 and it said yes. xDDD

I thought DeepSeek explained to the world exactly how they beat everyone at a fraction of the price.

deep seek is cheaper because garbage in gives garbage out, its not about the amount of research you do its about the quality of the data you have. there actually comes a point when you have put so much garbage data into a LLM model or a baby that it slows down its learning been watching what deepseeks being doing for a while its been available in chat bot arena for some time, since last yr these models are obsolete deepseek is giving its deepseek out open source just like a bunch of models where based on open AI so future models are going to be constructed off deepseek china needs us chips as much as it needs us AI you guys need to stop drinking your own koolaid. china was already 1 of the worlds largest chipmakers just as its now the leader in AI china is already making its own 12 nanometer chips that are faster as US 7nm's thanks to better programing

Yes I want AI agent to scrap the internet for ways it can make money for me that day. That is demand.

I really don't care about "integrating AI into my workflow" or whatever. In fact, I find AI quite boring and would rather spend my time talking to other people and making human connections.

You should check out a KZbin video called "This 20+ AI Agent Team Automates ALL Your Work" by Ben AI. It’s a pretty long video, but man, I watched it in one go at 1x speed. It offers a great perspective and some fascinating thoughts about AI agents. Just imagine designing such agent networks for specific tasks!

no all the problems have been solved. we are useless. good talk

I saw this Reddit post too, the crypto comment: after a few minutes of figuring out what it was doing, i moved on. I still don’t understand. If you figure it out let us know! Thanks for the video.

1.21 gigawatts?!?!?!?!?!?!??!!!!!!!!!!!!!!!!!!!!!!!

Ahhhh China!!!!

neural network consisting of different agents each working on a set of subtasks within the workflow or task

really been enjoying your videos can tell you're probably gonna do well at this whole youtube thing, keep it up

For me it's Go. It's very productive like Python allowing me to solve issues quickly and get a working solution done and I like that it's compiled (static binary so no dependency headaches) with stupidly easily concurrency and memory safety features. It doesn't really have any drawbacks when compared to Python imho and it's got a lot of batteries included in the standard library and it's third party library management is great. Just fits really nicely into that middle ground for me.

Everyone who can't code Hello World is under the impression that we're living in Person of Intererest.

As a software architect, and an IT nerd since childhood, I'm actually extremely disappointed by the current AI hype and general push for AI. It has significantly soured my attitude towards corporations, and eroded my trust in them. I think their entire goal is simply to addict and enslave people and companies to using their paid or soon-to-be paid services as much as possible, being completely careless and irresponsible towards the potential damage they are doing to people's life, livelihood and career with their hyperbolic shilling (of actually broken or half-baked services). I find it completely antisocial and anti-human for these heads of companies making categorical statements about what people and professions won't be needed in a few years. Meanwhile I can't remember meeting anyone who genuinely wanted this supposed 'revolution' to happen, and genuinely felt positive towards it instead of fearing it or being annoyed. It's entirely forced, it's entirely serving to further increase the profits of the richest corporations of the planet, and it seems to be that it's against us, and not for us. As toxic, as deluded, and as morally bankrupt as the US generally has become. That said, the amount of slack and lack of work ethics a lot of people in the industry are exhibiting is also extreme, I think. Many programmers expect that companies will just keep paying tens of thousands of euros/dollars every month per developer for minor feature changes or additions, while they are comfortably sipping their coffee from home office, working maybe a few hours per day, not paying nearly any attention to the big picture or showing any initiative, and writing mediocre code with very little design or true engineering effort. If we make software development prohibitively expensive for companies, and we become a nuisance to deal with because it costs more and more while changes take more and more time and the quality of the software is decreasing at the same time, then honestly, no wonder if trouble is brewing.

Not sure how I found your video on my home page, but I agree with you. For the time being it is a tool, and it will continue to be a tool if we continue to evolve and adapt to the AI's improving capabilities. But, if we programmers get complacent, we could be replaced by other programmers who are able to leverage AI more proficiently. I don't see a world where the AI is completely left to it's own devices for a long time. For now, we work alongside it and train it to be more accurate and helpful. Subscribed to your channel :)

Interesting name

Lol let's see how that goes for him

"You" people are misunderstanding the issue. Here's why: Supply and demand is a basic rule of nature, but people forget it applies/covers jobs as well. This applies to the AI situation not because it will literally replace people, ie all the people to be specific, but instead it will lower the reward/wages for jobs in all sectors… which means the people get replaced as-in they starve.

Overall I enjoyed this list. Seems like a nice first level look, although practically we need level 2 to really get anywhere.

Correction: the only thing Erlang and Ada have in common is they are designed for robust never-crash scenarios. In design, Ada is more like Pascal/Modula with concurrency features. Erlang is like Prolog with independent Actors, so it’s in the functional/logic space of languages.

Did ChatGPT write this?

Yall need to go ahead and bust out ubi now

Cool Video, I was thinking about this literally a couple of days ago, my 2 cents was: the word Agent is incorrect to use, "AI" "AGENTS" lack the same kind of SELF-directed intentionality that real human agents do. A real person can set their own GOALS, weigh moral or ethical consequences, and then consciously act based on personal DESIRES or beliefs. AI “agents,” by contrast, are bound by instructions and data from humans: they do not genuinely form INTENTIONS.

AI will reduce the need for humans in the way that automobiles reduce the need for horses. Extinction will be brought about by current divorce laws, effective birth control, education, and highly effective entertainment devices.

Because of capitalism all or most benefits of 'progress' are taken by those at the top. Capitalism is the issue here.

What's the evidence we're anywhere near AGI? LLMs? They just remix content created by humans.

I worked with Clojure. We had an "empty" template service that basically only had a health check. When starting a new project, we'd launch one in Kubernetes with the REPL's port exposed. You can then program the whole service *while it is running*, adding new logic, modules, endpoints etc. while programming in your local editor, even adding new dependencies etc. It's absolutely crazy productive.

Clojure also has Clojurescript that compiles to JavaScript with many interesting projects to speed web development, like figwheel that enables dynamic webpage updates as you write code.

Shell scripting: Babashka is a Clojure shell scripting language that replaces Bash, giving compiled speed like binaries do with a real programming language (has Java compiled in for speed up). I have made a number of scripts with it from Clojure code that are just like regular shell utilities.

Uh, learning Lisp will change your programming abilities once you understand it. It is based on lambda calculus . Composite functions from math are a good way to view it with focus on return values. This is equivalent (through some math magic) to the Turing approach that is at the base of Algol-like languages (e.g. C/C++/Java) or their forerunner Fortran, etc. Lisp makes you think in terms of higher order functions that return functions, code as data.

Haskell is more of an academic/research language.

3:40 you can use any language for anything if you have the expertise, time. That doesn't make it fun. A fun language will need to be powerful, dynamic, have simpler syntax, provide many functions to work on default data structures without having to be written, allow multiple programming paradigms, leverage many libraries, be practical... Guess which has all of that and more?

Haskell is static and difficult to grasp the syntax, etc. It cannot be described as fun. Clojure has some influence from it in its development.

Clojure is the most fun because of how dynamic it is. It is very easy to tinker, get fast feedback, among many other superior language features.

2:40 no, python is a Lisp ripoff. Clojure is a Lisp.

you have a lot to educate yourself about

zig

ASI will be a global public utility which governs and algorithmically observes worldwide traffic monitoring for subversive user and rogue AI behavior.

🍓❤️

As individual we exist because (a) we can earn our livelihood (b) in exchange we provide our labour which society appreciate (c) we feel proud, a purpose and achivement. If each of above is destroyed, I wonder why would I live and how does it matter to anyone if I live or not. And this is true at whole humanity level. An just for your info, if we think that somehow magically we will keep evolving and be relevant because machine can't do this or that, it's absolutely fool's paradise. Very soon coding will change that only AI can work with. Engineering and tools will change (or say evolve) that only AI and Robot machine can operate. Our work and intelligence will be as obsolete as today we consider a donkey and horses. Things will evolve so much that even if we want to work, we will not be able to contribute in good way just like a donkey can't sit on computer to code or horse can't do heart surgery on operation ICU bed. So think again!

kzbin.info/www/bejne/Y57Nf4J6gK90oac I wrote a song about similar topic from human art losing its value when generative art is so easy, fast, cheap and great looking. people won't care if the song took a single generation or 5 months to make.

What if, by capitalist forces, there's an AGI in every mobile? But seriously. I'm starting to question that anyone would even want to pay for AGI. I mean, you have a forklift, you want it to be able to operate in any store house, so computer vision, spatial understanding, etc., but do you need it to be able to critique Dostoevsky? What company would be like, oh yeah, let's add a couple of more hundred millions/billions to the project to get that as well! This is also where the dangers start popping up. I'm thinking a superintelligence would think the earth biosphere is rare enough to warrant protecting (after all, immoral/unethical super-intelligent AIs is probably more of a Hollywood product than a seriously real consequence here). The problems start when the AI is NOT super-intelligent but is given tasks beyond its understanding. We're already seeing that today...

3:45 well why do you think elon advocates the AfD (the far right party in GERMANY) Lebensraum that is the answer (Propably)

I think it will probably be something very similar to the movie Elysium (2013).

It is written on the Georgia stones, look it up

Pretty funny :D But I mean what did anyone expect? How is there not an emergency stop in those... Silly. Btw, your volume is way too low, about 12dB under the YT standard

I will never enter one of these vehicles...

I have two definitions of agentic workflow: the correct definition is the one Sam Altman says and the incorrect one which is whatever the user needs help with, wether it be defined or undefined but is dependent upon an initial interaction, or continuous interaction, or sparingly interaction etc. to get closer to the end goal which is either predefined by the user or a ‘collective defined’ goal as the agent aids to consider different perspectives etc. Still working on it but for now I’m sticking to the fearless leader whatever Dr. Altman says, because OpEnAi iS kInG

Look up Sam Altman's blog posts on this topic. He is very clear about the consequences and runs through different scenarios, describing how bad they could be if they don't mitigate the risk. One of the most serious issues here is that with no one having jobs, we'll _have_ to have a system like UBI (as mentioned) or populations will starve and revolt. You obviously can't maintain a country with 99% not working and 1% accumulating literally all the benefits. But then even with UBI, the complete destruction of social mobility that we still have to some extent today with skills is another disaster. I think you could have covered this topic in much more depth and seriousness with a bit of research and I would encourage you to look up these blog posts. He's not being flippant, the man was not worth all he's worth today when he wrote all this. One was titled "Moore's Law for Everything" and covers a lot of this, I highly recommend you read it and think of when it was written.