Maybe someone will make a way to use DDR3 in a DDR4 system. :-) Make the impossible possible.
tal
joined 2 years ago
I shouldn't have thrown out all that DDR3 memory that I did.
Well, China would rather have AI in China than in the US. I don't think that that part is what's dumb about this.
I think that if I were the current American far-right, I'd stop dragging the pope into this stuff. Like, there is an anti-Catholic strain in the far-right that's been around for a long time, but as I've commented before, if you start getting those anti-Catholic Protestants riled up, you're also probably going to antagonize the Catholics, and a Catholic-Protestant schism in the US would be really bad for the Republican Party, since they're trying to keep social conservatives as a bloc in their coalition.
I guess it's more-understandable with Thiel than when Trump was having a spat with the pope, since Thiel isn't as directly a political operator, but still.
If the pope says something that you don't like, and you're a prominent right-wing figure in the US, even if you really have it in for Catholics, you're probably better-off biting your tongue and just addressing the thing they said, not turning it into a conflict with the person. Like, if I were Thiel, I'd have said "AI is really important to make the US competitive on the international stage" or something. Whatever, but avoid referencing the pope in particular.
Trump gets OpenAI to offer US 5% stake
https://www.schwab.com/learn/story/some-indexes-accelerate-entry-massive-ipos
Some Indexes Accelerate Entry for Massive IPOs
FTSE Russell: FTSE Russell announced the results of its index consultation on May 26, 2026, with changes effective as of the same date. Russell's U.S. Equity indexes will now allow fast entry for very large IPOs on the fifth trading day, instead of waiting for the next quarterly rebalance. However, the weight of the IPO in the index will be based on investable market capitalization. In other words, only shares available to public investors will be considered when calculating the company's size. Plus, IPOs must have at least 5% of shares available to public investors, either on the IPO date or within the next 12 months.
Uh. So does that 5% count towards the minimum float amount that triggers automatic buying by some index funds?
https://knowyourmeme.com/memes/clippy
Clippit, better known as Clippy, is the default animated character in the English Windows version of Microsoft Office Assistant, an interactive user's guide that came pre-installed with Microsoft Office bundles from 1997-2003. Due to its impractical and intrusive nature, Clippy quickly became a subject of mockery among Office users, inspiring a series of satirical images and parodies addressing its overall incompetence.
I guess we'll see if Round 2 goes better than Round 1.
On the negatives:
- Online abuse: AI is fueling the spread of sexual abuse material and sexually explicit deepfakes, with women and children most at risk.
Eh. I don't see this as that fundamental. I mean, Photoshop lowered the bar too to making synthetic pornography, to slapping a head on someone else's body. The ability to do image warping allows resizing body parts in a convincing-at-first-glance look, and I remember when that was a fad. It just doesn't seem to have changed society all that much in the past. It may be that people just stop caring much about pornographic images of a particular person. I'm not saying that it won't have an impact, but I have a hard time seeing a scenario where it really deeply alters society.
- Disinformation: AI can generate false information that is as convincing as the truth, undermining trust in public debate and democracy.
Yeah, that's a bigger issue.
It is not one that is impossible to deal with. It was a situation that we had to deal with prior to recording technology. And there were problems
like accusing politicians of going to one place and saying one thing in their speeches, and then to another and saying something else was a real thing in the US back when the only record we have was from newspapers printing summaries of what they said. We had solutions for that era, like people who would put their reputation on the line to attest to various facts. We could do it again. But we've benefited from having easy technology that made it pretty easy to make a credible record cheaply and easily of all sorts of things, audio recording, photographs, and video recording, and the bar for that might rise.
It was always going to happen one day, whether-or-not neural networks were involved. Computer graphics and audio synthesis have only gotten more-accessible and convincing over the years.
I think that one new issue is the ability to synthesize propaganda using the bandwagon effect on social media. That is, it used to be just financially impractical to have a zillion people online trying to influence people. But...if chatbots drastically lower the cost to that, that could be a real issue, and one that we haven't had to deal with before, ever.
Crime: Criminals are using AI to carry out cyberattacks, fraud and social engineering scams.
Yup. Some of this we can fix, and is because we have outdated authentication mechanisms (like recognizing someone's voice on a phone because the phone network has basically no authentication mechanism). Some of it is going to be harder. I think that this is largely not fundamental problems, but some things are going to have to change.
- Mental health: Some AI systems can reinforce harmful beliefs or behaviours, leading to mental health crises, including suicide.
Ehhh. I mean, yes, but so can all sorts of other things. Sitting alone watching TV all the time. I remember distress when the Internet became more widely-available about how people could enter into harmful forums. And we've always had crackpots with weird ideas out there.
https://en.wikipedia.org/wiki/Time_Cube
Ray didn't need AI to develop his theories.
Honestly, I'm kinda more concerned about stuff like cults and scams, cases where someone is actively attempting to maliciously manipulate people than "random person goes and spends time talking to a chatbot and feels that it reinforces his crackpot views".
- Loss of control: As AI becomes more autonomous, experts warn it could become harder to monitor and govern without stronger safeguards.
Yeah, this is a big one. The Friendly AI problem is a hard one. I don't know if there are practical solutions, not for self-improving advanced intelligences, which we are certainly going to try to build.
- Environmental impact: The energy-hungry data centres which power AI are contributing to greenhouse gas emissions which leads to global warming.
Eh, yeah, but this is basically the same as our existing energy problems. Like, we were already emitting an unsustainable level of carbon dioxide emissions. The answer has to be shifting to different generation methods. The answer was never going to be "we keep burning coal and whatnot and everyone just reduces energy usage enough on a per capita basis to keep human-driven carbon dioxide emissions at a sustainable level".
But you’re not selling that as a service.
No, but...is your concern the ability to run a commercial service involving AI compute? I mean, there are certainly a ton of startup companies doing that. Heck, there are people selling access to their GPUs for parallel compute on vast.ai.
I'm just saying that I don't believe that the ability to run neural nets on parallel compute hardware is something that is going to be terribly exclusive over time, and certainly isn't today something limited to someone with a net worth of a billion dollars.
As I've commented before, we'd need a lot more RAM than exists in the world today if everyone's going to do it. Like, AI companies are buying more 2026 RAM production than the rest of the world combined, on the order of two-thirds of global production. If they get something like 100% capacity utilization of their hardware, and a typical user doing local AI compute would get something like 1% capacity utilization of their hardware, then we'd need about a hundred times that much memory to let everyone run comparable stuff locally. That's a pretty stupendously large amount of memory. But if, over time, there's demand for it, I expect that it'll happen. We've scaled up parts of the computer industry by orders of magnitude in the past.
The United Nations is supporting a new international architecture to help countries make informed decisions about AI.
In 2025, the UN General Assembly established the Independent International Scientific Panel on Artificial Intelligence, made up of 40 experts from every region of the world serving in their personal capacity.
The panel's role is scientific rather than regulatory. It assesses, on a regular basis, the latest evidence on AI's opportunities, risks and impacts and produces independent reports that governments can use when developing policy.
This is approximately what I recommended some time back on Reddit. Can't have Sam Altman or advocacy groups driving it
need reasonably objective, unexcitable observers. Also, not just to governments but to the public.
There are a lot of unknowns, but at least we can start building consensus on some points and identifying the known unknowns.
It's also not just computer, technical unknowns. Like, one of the big questions is how much existing stuff done by humans can we realistically automate (Sam Altman has an incentive to say "everything, tomorrow"), and what. Economists can't predict the impact without technical data on capabilities, and right now, we just don't know a lot of that, and we don't know where we'll be in a decade. But they can firm up a set of questions to ask and to watch for. "If we can achieve capability X, then this is what the likely global impact is."
Those predictions will probably be bad at first. But...that's science. You start with a model and start refining and revising.
I'm also interested in coming to some kind of firmer consensus as to superintelligence risks. Right now, we have a lot of ideas flying around that run the gamut from If Anyone Builds It, Everyone Dies to "it's impractically hard to build and is thus a non-issue" and little consensus as to what lines we might cross that would expose us to superintelligence risk. Needless to say, this is is basically useless if you're trying to do anything from a policy standpoint.
I don't think that we should kick the can down the road on that. The problem is that our major advantage, as humanity, in dealing with superintelligence, is time. We get to make the first move, structure the environment. So we don't want to throw that away. If we need to figure out how to contain, how to control superintelligent systems, we likely need to do so before we have people actually building them. And there's a lot of potential economic and military and so forth benefit in building a superintelligent system, so people are probably going to try to build such systems. But there aren't similar incentives associated with safety surrounding superintelligent systems. People may not just go out and do that on their lonesome. So if we need to fund them or whatever, we should be doing so without waiting for those systems to exist.
You mean to say, the thing that only billionaires can own
I own a dedicated AI compute machine, and I'm certainly not a billionaire.
I don't know if I'd get one now, with RAM prices where they are, but down the line, I expect that they'll have better hardware and bang for the buck, as hardware designed from scratch for running neural net compute instead of being repurposed from doing 3D rendering spreads.
I wonder how to tell if an instance supports it apart from just blindly trying.
They don't have to stick it at the same prefix, so there's no reliable way to check by just guessing the name.
It looks like p.lemmy.world is running the Photon frontend. On lemmy.today, we have it at photon.lemmy.today.
Lemmy.today lists the alternate frontends in the instance sidebar, but there's no requirement for an instance to do that.
The breaking point? A price jump of 16%.
When shoppers do make the switch, groceries are the first to go with 76% dropping a brand they were previously loyal to
I guess that's why shrinkflation is such a thing specifically with food.
I seriously think that they should push it back to 2028.
I think that Valve's making a mistake by not pushing back the Steam Machine to 2028 too, though at least for Valve, a hardware platform flopping isn't a big deal, since they don't rely on it alone to make sales.
Or maybe I'll be wrong, and gamers will be significantly less price sensitive than they have been in the past. But my guess is that they aren't gonna be jumping on consoles with a pricetag that's that high. As I said before, the only console to be successful in the past that cost nearly that much in inflation-adjusted terms was the Atari 2600. Everything else failed.