Видео

Copilot is exavtly the user interface that people will use if you integrade it everywhere.

Thats just the small pne, cant wait for the large one

What are the keyboard shortucts you are using in Obsidian to perform the functions you are using in the video?

Can't wait to get the Vision Pro for my workflow. Having "portable" giant virtual monitors with my laptop using Logic x is a game changer !!

Love the OG Men I Trust song

A lot of older systems from that time didn't even have a UI, just a terminal.

what about x ai

I appreciate this perspective. Good stuff, thanks for putting it out there.

Yawn…nothing new or interesting here.

I miss sunsets on Lake Union so much, glad you worked it into the shot.

Agree about Nvidia. Carl Sagans words about a technologically literate republic sound in the wind. OpenAI fundamentally benefits from investors and the public not understanding their technology and obfuscating it. And the investors found out the hard way. That's ultimately why the market tanked. I think it represents an irrational market. Whether 1 million tokens becomes free, it doesn't change the fact that soon every mcdonalds, walmart, and mom and pop store will have some sort of AI involved or running the store completely. That all requires chips. It's like saying hydroelectric turbines don't have value because all of a sudden some other guy can produce cheaper water now (silly analogy). And that applies to chips but also applications of AI. Whether or not AI is incredibly cheap doesn't change the fact we need people to create the AI that can diagnose and cure cancer before it even starts (for the time being anyway). Ultimately I don't think these investors actually understood the technology or their investment and that is because the powers that be are obfuscating how difficult it is to actually recreate these things. Someone let me know if this is entirely offbase would love feedback

I've been a commodity since I was conceived in the 1990's.

I find it hilarious that people take a number that was 6mil for the last training of deepseek 3 run and build their whole thesis on this. Anthropic last run of one of their models costs 10 mil. Not that huge difference. Also we know now that 250k gpus been sold to singapur and the assumption is that it didnt end there . I. Frustrated by the fact that you all youtube influencers dont even take 5 min to read up things but just behave like parrots

Everyone has tried deepseek now and can see how bad it is so this video is irrelevant. Like Deepseek

In the 90s I nearly got crushed by a beer truck in the park you are sitting in (broke loose by The Zoo)...

Android isn't just pixel. Android is an OS and it's the most used one by far.

well, it's a good thing all those silicon valley freakshows bought up all the water-rights to cool all their overpowered GPUs while LA burns so a robot can make cat pictures and create a recipe for a reuben sandwich for gramma.

Dumbest take I’ve heard yet, to have a new competitor appear and for you to say that the big winners are the companies getting undercut price wise, you an apple shill or something? You sound like a MBA that thinks they understand technology

R1 was developed on nvidia hardware, ran on nvidia hardware, until that changes, don’t sell.

Very succinct and impressive commentary. New subscriber.

awhh look at the duckiess

No need to be a prophet to predict that Nvidia is coming back soon as strong as ever... You don't address the Stargate issue, there was a whole gang organized around the promised 500B, the big deal is that these big alliances collapse.

Open source, efficient, and powerful, deepseek R1 is the dawn of hope on the horizon. I hope the sun of the new era will eventually rise.

i remember when you had like 300 views per video, crazy work ur doing

China out there getting headshots back to back to back

Meta's strategy of having AI make content is going to drive people away and destroy their platform. No one is on Facebook or Instagram to watch AI Generated crap. They are on there to connect to other human beings. When they start changing that model people will leave.

as above so below , as within so without this is the way of ALL things , indeed

Closed source AI is no longer a viable business model

Very rare on youtube talking about reality of technology not a motivational video with a very good scenery simple aesthethic lowkey hahaha

Hapa Alain Delon?

Where are you? It looks nice.

Maybe a good metaphor is to consider high end Nvidia GPUs as raw physical power. Larger, stronger people have more physical power. I see deepseek’s engineering despite hardware constraints being like BBJ used on a larger opponent that doesn’t know BJJ. But , a larger opponent can also learn BJJ, might already know BJJ and has kept it secret or has another undisclosed technique that’s even more effective. There’s nothing stopping big AI companies from also using chips more efficiently and with more chips and more efficient usage, there are even bigger gains. This doesn’t sound like something that would disincentivize buying more Nvidia chips. Am I missing something on this point?

Me and my boy just speaking tech🚤⛵️🌅⚓️

Who are you and why should anyone care what you think? You seem like a pompous mf

Just a few weeks ago I purchased Nividia stocks... I'm definitely not going to panic sell, even though yesterday's blow wasn't comfortable. We'll see how this pans out.

I have been very cynical about these smaller models that were refined by R1, since I've seen so much coverage making NO distinction between the 491B param actual R1 and R1-flavored Llama. R1 distills strike me as, like, the LA Croix of R1. However, I set up a home machine serving 14B R1 Llama distill, so I could have access to it on my phone, and it's really quite a lot better than raw Llama at that size. For context, I'm running it with a 3080 and an otherwise even lower power machine (don't ask me why it has a 13100 cpu), and while I can't remember token count off the top of my head, it's perfectly good at generating code in a reasonable amount of time. R1 and it's distills have reignited my curiosity in open source and particularly the future of local machines. The biggest issue I've had so far with these smaller models run locally isn't what I perceive to be it's intelligence - - it's moreso the much shorter context windows. If I give it a coding task and it gets it right first try, it's great. But if I need to go back and forth with it for a while, it will absolutely start forgetting aspects of it.

your assessment is a little weak here... nvidia has a monopoly over the chips, even deepseek used their earlier chips to train the ai. there's literally no other options because nvidia is good at what they do. theyre not simply just 'selling tools for the gold rush', give them a little more credit here. you just talk about the us companies, you don't talk about the chinese companies. it seems like deepseek came out of nowhere, but china has always been dominant in ai. people like yourself are just ignorant and do not give attention to RELEVANT topics. when you talk about AI you cannot remove china from the conversation. as a matter of fact china takes top priority and all these tech companies excluding the american ai companies and nvidia are less relevant. it's because of people like you who are clueless that's why you get shocked like this. let me teach you something boy, china has over 1.4b people, they use 5g and mroe technology than we do. in 1 day, they collect more data than we in the us collect in 1 year. ai is driven by data, so their technology is far superior to us, this isn't new.

meta is model is not open source he just calls it that.

NVIDIA just ran another "pump And dump" Sham on everyone.... They made billions on the way up. They were smart enough to sell Short - that will pay for additional Expenses.

Despite being chinese in origin (the online version of deepseek refuses to mention the tiannamen square massacre, for example), its good that deepseek hit the western AI giants right where it hurts, proving that it can all be done, open source, on a fraction of the absolute mountains of money nvidia, "open"AI and meta eat every day. The current trend in the market is that AI is moving on to be a paywalled service that only corpo giants would have true access to, and deepseek proves that such thing may not be the case.

Love the style of video here. Looking forward to more. Only thing I disagree with is Google's response. I think they have a pretty large base built out with there TPU infra. They're also the company that had (has?) the most RL talent, which Deepmind had been practicing a lot recently. I feel like Google will be bringing more surprises this year.

what the hell am i wathcing

DeepSeek likely used Nvidia cards but can not legally say so.

hey guy. deep seek is clubber lang. a new dev boom begins from here. big AI falls. AI has just been democratized. that's what happened man.

What a beautiful scenery I need to find a place like this in my city

I loved this, great video

Broligarchs spent billions, trillions of dollars on infrastructure to support their AI gold rush, at the expense of aggravating an already-bleak climate situation. I embrace the China Menace popping this bubble before it got out of hand. China figured out gun violence, healthcare and bullet trains while the US is dumping money to make a few bros with too much money even richer.

Someone probably already mentioned this but, these windows, your Obsidian notes, canvas, graph, etc, should all be independent from the mac's desktop. There's no reason why you can have your canvas physically behind you occupying floor to roof of space, your notes below your web searches, perhaps despite all this real estate you fill your space, so, assuming you want to stay put and not move around, you can organize these things in such a way that makes sense to you and have some easy and low-effort gesture to bring them up in some way. Can't wait to experience this kind of workflow some day.

I agree that Nvidia's position isn't changed by DeepSeek R1. I think Nvidia is overvalued. Sure they don't have strong competition today and they have a moat in that everything is built around their hardware, but at the same time if a competitor did show up then I think tha moat is gone in a few months. The AI companies are retraining often enough that the savings from building for a competitor makes sense long term.

Thank you Bo