AI data centers sucking up all the digital oxygen - SSD prices leap 25% in a week! šŸ¤Æ

Watercooler
104

Iā€™ve been aware of it for a long time but have neither the need for that nor high-end hardware for working on professional projects. Iā€™m just a tinkerer, and stock Ubuntu has worked well for me for a long time, though Iā€™ve played with others (Ubuntu Budgie, Solus, KaOS, KDE neon, Trisquelā€¦). Depending on what Iā€™m trying to do at a given time, Iā€™m most likely to be using regular Ubuntu or Trisquel (for no real reason other than Iā€™ve played with different applications on different distros).

I like that thereā€™s so much variety available to try! :+1:

1 Like
105

Anyone in here experimenting with any of claw agents (eg. The infamous openclaw, or offshoots) or CLI agents?

This has been an insane month. Itā€™s hard to keep up. The entire world has shifted very fast and the majority of people donā€™t feel it yet. Just in the last month weā€™ve started the biggest paradigm shift since the Internet itself and essentially at zero marginal cost. Itā€™s blowing my mind.

Itā€™s insane what these things can do now within just the last few weeks. Essentially, anything a human can do on a computer, can now be done with AI agents. Literally anything. This new innovation of looping them into a heartbeat script for continual functionality along with access to anything and everything through apis and CLI, and other stuff. Then you add to that the recent updates to allow them them to have cache memory and long term memory, and stored instructions and project outlines in MD files.
And that anyone can run basically unlimited of them for free. Itā€™s mind boggling.

In just a couple short weeks itā€™s become the biggest leap in personal productivity since the invention of the PC. Weā€™ve gone from computers being ā€˜tools we useā€™ to ā€˜workers we manage.ā€™ Most people are still trying to figure out how to talk to the tool, while a few people are already managing basically unlimited agents.

(If anyone plays with these, please use them safely. Learn how to sandbox them in a container or similar. Donā€™t give them your credentials or API keys directly. Consider options like making them use N8N automations to gain API access so it canā€™t access or give away the actual key. Implement reasonable Human in the loop controls. Set up restrictions so it doesnā€™t have free access to the Internet but needs permission granted for specific things (white listed). Mint selected shared folders, some with read only access and setup backups. Audit logging.. Have python parse untrusted documentation for the relevant into instead of analyzing it directly and risking prompt injection. And be careful about plugins and skills in general and keep in mind that some creators will make a skill that is benign now with the intent to change it later and get an auto-update to infect people with a change, so please avoid third-party skills and plugins if you donā€™t know what youā€™re doing).

Seriously though. These things are blowing my mind and changing reality as much as the Internet did and itā€™s just now starting up with all the free open source agents and access to free open models.

On the downside, itā€™s been crashing the value of all the software companies globally because anyone can easily and quickly duplicate almost everything they do with customizations at almost no cost, but the new era of open source taking over is going to be insane.

Kind of sucks for me in some ways since Iā€™ve been involved in building a software startup for more than a year :sweat_smile: but despite that personal pain, this stuff is astonishing.

Related to Wyze, after Iā€™m done with some personal projects, I intend to aim my agent toward figuring out what improvements I can make with the home assistant HACS Wyze integration through the Wyze API and see what I can improve and add to that. Itā€™s a low priority though.

Did I mention the last couple of weeks have been blowing my mind? Iā€™m curious if anyone else here has tried any of these and if you use them for anything interesting?

1 Like
106

AI is supposed to be a tool not a substitute. All the younger generation will now have to ask their phone, ā€œHow do you tie she laces ? ā€œ

NO
NO249Ɨ214 20.1 KB

laughing

2 Likes
107

Made me think about this article i just read.

ā€œWe Are COOKED:ā€ Weā€™re Only Two Months Into 2026 And These 35 People Have Already Posted The Dumbest Things Iā€™ve Ever Seen In My Life"

2 Likes
108

Homework done while you sleep. :laughing:

2 Likes
109

Thanks for the chuckle :+1:

I feel sorry for generations that come :pensive_face:

1 Like
110

I am not sure how retirement works in Canada, but ā€œgenerations that comeā€ need to pay into the US system so I can continue to receive my Social Security.

111

Same here!

1 Like
112

ā€œBurger King will use AI to check if employees say ā€˜pleaseā€™ and ā€˜thank youā€™. AI chatbot ā€˜Pattyā€™ is going to live inside employeesā€™ headsets.ā€˜

113

At least theyā€™ll save on mystery shoppers now :laughing:

1 Like
114

When I first started using Google Assistant years ago, primarily via Google Home Mini smart speakers, I was pleased and amused to learn how the assistant voice would give encouraging and complimentary responses if I was intentionally polite. For instance, I could say, ā€œHey, Google, turn on the light,ā€ and the assistant would do that with whatever response it gave. If instead I would use more polite requests (ā€œHey, Google, would you please turn on the light?ā€ or ā€œHey, Google, would you please create a calendar item?ā€), then the assistant would occasionally generate a response thanking me for having good manners or being polite or whatever.

I donā€™t know when they stopped doing that, but I thought it was a cool feature, especially for someone just getting used to controlling a computer like that with my voice. It didnā€™t make me think that the assistant had any real feelings or that I had any sort of parasocial relationship with the machine, but I thought it was a thoughtful and interesting thing to have programmed into the assistantā€™s responses, and I could imagine someone making a case that this is a good way to model polite behavior in communication rather than just feeling entitled and commanding the machine. I havenā€™t heard it do that for several years, even when I intentionally try to be polite and ask it nicely for things. I kind of miss that feature.

2 Likes
115

Next time you talk to an AI Chat Bot, be sure to ask about its mother.

116

Or just ask it for a Bolognese recipe:

117

With the increase cost of memory and storage, you might be asking yourself, ā€œCan I run a computer without RAM?ā€ Spoiler Alert, you wonā€™t want to, but this video is still very interesting.

1 Like
118

I was particularly amused by his reference to using a brain[self-censored] interpreter at one point in his testing. :grin:

1 Like
119

ā€œRAM Prices Are Threatening the Viability of the Raspberry Pi and Single-Board Computingā€

1 Like
120

Hereā€™s my prediction on the RAM issue:

  • V2026: Prices remain high and volatile (another 30%-50% increase by end of year)
    • Note that there is a small chance that we could see a little bit of relief toward the end of this year if the rumors are true about some frontier companies reducing their RAM orders (For example, it is claimed that openai canceled Sora partially to reduce their RAM demands, and that they are trying to scale back on RAM needs from over purchasing, which may cause a sudden ā€œbullwhip effectā€ where prices drop as manufacturers scramble to sell their excess capacity back to consumersā€¦ I donā€™t yet believe this theory, as it has not been independently verified, but there is a possibility we could see something temporary like this by the end of the year.)
  • 2027ā€“2028: New fabs come online, supply stabilizes, and prices begin a slow descent. Prices should stop increasing in 2027, but take until late 2027 to slowly start to drop. 2028 Is when there will be real relief for when the supply starts to catch up with the demand.
  • 2030: Production efficiency finally catches up, and RAM might become ā€œaffordableā€ again, but likely never as cheap as it was in 2023.

The big three have been scaling up production and creating new manufacturing plants (including the new ones in New York and Idaho), but a semiconductor ā€œfabā€ takes 3 to 5 years to build and bring to full capacity. So we are not likely to really see those start to make a difference until late 2027 or 2028. So we are almost definitely going to remain with the supply shock for the rest of this year and most of next year, and then we will start to see some relief as the the fabs reach their full potential.

So yeah, we might see the end of rPi and other things for a few years, but itā€™s mostly a temporary issue and it hasnā€™t been as bad as I expected it would be when Micron announced dropping consumer production altogether. I think

So yeah, we are going to have a rough digital ride for a couple more years. Even iot stuff like Wyze will start to be more expensive.

1 Like
121

Data Centers better grab more RAM so they can make their ChatBots know how to spell.

2 Likes
122

:man_facepalming:

1 Like
123

I mean, it came up with a good example:

ā€¦They might say ā€œChatGPTā€™s really good at thisā€

And it is correct, in that sentence, ChatGPT DOES have an S in the word spelling when it is a contraction or possessive connotation. Soā€¦sometimes ChatGPT IS spelled with an S in those cases. 1 point to ChatGPT for coming up with a good hallucination gaslighting cover story. :+1:

1 Like