A thread for unboxing AI
The rapid progression of AI chatbots made me think that we need a thread devoted to a discussion of the impact that AI i
I don't know if deepseek, qwen, etc... are china's "best" but it sure seems like most of china's public comparable stuff releases straight to open source. So their strategy may be as long as the US doesn't win, we win.
There is always going to be use cases that need more horsepower and energy but the stuff that is taking everyday people's jobs and tasks will get more and more efficient. Probably in not much time you'll be able to get the same performance from today's bleeding edge models running on your phone.
Yeah, that's not happening. These AI chips have 10K-100K more computational power than phone chips and that's just one chip; data centers are running on clusters of thousands of chips. And each one is pulling almost as much power as a household a/c unit. And that's not taking into account manufacturing capacity. We're literally at the point where it's more better chips for cars and phones or AI.
If there were a “ruling class” controlling everything, it’s hard to imagine they would’ve allowed the creation of an open, global internet in the first place, especially one that’s empowered dissidents, whistleblowers, independent media, and mass coordination. The internet has disrupted governments, corporations, intelligence agencies... pretty muc
It wasn't. The core design has its roots in making a network that could function even if single nodes went down. Partially funded by DARPA, the reason being rather obvious: A communication network that goes out with one node is very weak to hostile attacks. The work would be known become known as ARPAnet.
Anyways, the story of the internet is convoluted after that, but the short version is that a standard was established to allow networks to connect to each-other, with academia largely being the standardbearer in going forward. In. the late 80s you started seing private users, the World Wide Web standard came about in '93 a few years later the last restrictions on letting commercial interests partake was lifted. The snowball was rolling.
I think it would be wrong to say the internet was based on being opposed to control. However it was not designed with central authority in mind, a byproduct of being a network that does not rely on single nodes is that it becomes very difficult to control.
As for the reason why it became an effective tool for opposition to authoritarian government and why it was so effective at avoiding censorship I suspect is simply because the technology developed at a pace that was far faster than the ability of governments to respond to it.
These days, as governments and commercial interests have adapted, we see that the internet is increasingly used in the opposite direction: As a tool to spy on, monitor and control the behavior both of its users and the societies they inhabit.
Of course, lurking below the controllable internet is still that that old technological principle of a network that does not depend on single nodes. Here you can find a section of the world wide web that is not indexable like its upper portion. There you can have sites and servers that can't be found unless you know their address, known as the deep web. On that portion of the network, data can't be scraped by big data or government agencies and communication that is still extremely difficult for governments to control (to a point, signal analysis is still a thing). On it you can also find sites that requires specific configurations and ways to communicate, known as the dark web, which are even more resilient to such attempts because knowing they are there won't be enough.
It wasn't. The core design has its roots in making a network that could function even if single nodes went down. Partially funded by DARPA, the reason being rather obvious: A communication network that goes out with one node is very weak to hostile attacks. The work would be known become known as ARPAnet.Anyways, the story of the internet is convoluted after that, but the short
https://www.amazon.com/Hackers-Computer-...
Great read. Really gets into hacker ethics.
These days, as governments and commercial interests have adapted, we see that the internet is increasingly used in the opposite direction: As a tool to spy on, monitor and control the behavior both of its users and the societies they inhabit.
I agree. But not in the sense that it's control for control's sake, as some believe. I think it's all about control for wealth preservation's sake. Unlike back in Machavelli's day when the divide and rule adage applied to kings and such, the people just seem to auto divide on their own in democracies without much prompting.
ai tools are amazing, but they'll often give you terrible advice - doing some strategic planning trying to merge redundant codebases to streamline us for continued international expansion and the ai genuinely suggested a very complex and convoluted solution that ultimately was just calling it by a different name and changed nothing
i will say this though, a lot of people are going to be confidently bringing "we should build a monorepo" type of suggestion to their bosses in the near future - where they are taking credit for the amazing and innovative solution that ai provided them and it's going to expose a lot of idiots at work who aren't able to question the terrible ideas it presents but instead roll with it
soon they will be gone and the boss will be coming up with the AI generated idiotic answer for themselves.
Square lays off 40% of its workforce for no other reason than because of a belief those jobs can be done by AI
Hate to see what unemployment will be like in 2 years
i will say this though, a lot of people are going to be confidently bringing "we should build a monorepo" type of suggestion to their bosses in the near future - where they are taking credit for the amazing and innovative solution that ai provided them and it's going to expose a lot of idiots at work who aren't able to question the terrible ideas it presents but instead roll wi
I've built 2.5 home servers in the last year with ai and had basically zero knowledge except what I wanted my end product to be. Got in tons of arguments with ai over them doing stuff wrong just by using my common sense. As you know still a lot of hand holding to do but I can put stuff together in a day or two that would have otherwise been months of training myself reading stack overflow, Reddit, etc... but can only imagine it's getting leaps and bounds better as the months go by.
I've built 2.5 home servers in the last year with ai and had basically zero knowledge except what I wanted my end product to be. Got in tons of arguments with ai over them doing stuff wrong just by using my common sense. As you know still a lot of hand holding to do but I can put stuff together in a day or two that would have otherwise been months of training myself reading sta
oh yeah i agree 100%, it's an amazing tool that does amazing stuff if you have the skills to shepherd on the journey
what are the servers for? running sims?
i'm going to build my own personal ai server for home use running open source llms with it this year - super amped that i'll have my own in house alexa that i can also interface for for coding without having bezos listen in to our dinner conversations nor paying hundreds of dollars a year in subscription fees
southpark already figured it out
Finally set up my autonomous agent on a headless mac mini. Connect to it via signal. Look forward to unleashing the beast.
Finally set up my autonomous agent on a headless mac mini. Connect to it via signal. Look forward to unleashing the beast.
please please please share your build as i was considering the mac mini route myself
oh yeah i agree 100%, it's an amazing tool that does amazing stuff if you have the skills to shepherd on the journeywhat are the servers for? running sims?i'm going to build my own personal ai server for home use running open source llms with it this year - super amped that i'll have my own in house alexa that i can also interface for for coding without having bezos listen in t
Haha, no sims, normal homelab stuff. Arr stack, Google drive replacement, PDF editor, a personal wiki for my work stuff.
Did a couple of cool projects. Setup up arm (automatic ripping machine) and digitized my mom's whole CD collection to flac and alac. That thing is sick, just insert a CD to the USB CD drive and it automatically knows when a disc is inserted and starts ripping the music, goes out to the internet and grabs all the poster art, names the tracks to how you specified and then ejects the disc when done.
Also I was taking care of my mom for a few months and she needed her blood pressure taken everyday. Set up an N8N workflow where I would take a picture of the blood pressure reading, save it to a network folder which would then trigger it to upload to Google drive, chatgpt would ocr it and then populate a Google sheets spreadsheet with sys, dia, hr, date, time and a link to the original jpeg to confirm the ocr was accurate.
Was forced to use chatgpt for it as no open source ocr could read led screens accurately. Guess they're all trained for documents.
Square lays off 40% of its workforce for no other reason than because of a belief those jobs can be done by AI
Hate to see what unemployment will be like in 2 years
The big beautiful budget included a workforce Pell grant. The AI boot camps are popping up just like the front-end boot camps when the internet took off.
As Say's Law says: supply creates its own demand. Ie, the AI-driven 'supply' of code is creating a massive new 'demand' for human judgment.
OpenAI CEO Sam Altman said on Friday it has reached an agreement with the U.S. Department of War to deploy its AI models on classified cloud networks.
-Reuters
Rip all of us probably. I've never been an ai doomer but this is an accelerated schedule by accelerationist standards.
Isn't this the natural course of things? Every other country will be using this technology for military purposes. Surprised Raytheon doesn't have their own platform and hired as many top level engineers that would do their type of work.
Probably the world will come full circle and we all better learn how to live off grid with no power as every country/hacker group will be able to bring down any other country's power grid at will.
Surprised Raytheon doesn't have their own platform and hired as many top level engineers that would do their type of work.
Engineers are plentiful. You need GPUs.
They're not idiots. They're just so myopically focused on the trees they don't even know there's a forest. Unlike the micro economic world they know, the macro view and all it entails is hardly intuitive.
The Fed provides a far, far better birds eye view of things. And while their views range from cautious to optimistic, the caution is more to do with how fast things are changing and their lack of economic controls to deal with it. But even that's no where near doomer.
If it's a matter of national security the government is going to get theirs.
