It’s a backtracking search program looking for integer solutions to a specific problem. I’ve tuned a series of bloom filters to fill my L2 cache so that I rarely have to touch main memory (this alone took my IPC from 0.1-0.3 to 3.0 per thread). Without SMT it’s 4.6 IPC/core.
I think it’s only able to exceed 4.0/thread with SMT off because of a uops cache? From what I’ve read the Zen5 front end only had a 4-wide instruction decode per thread.
> Doesn't really make sense to me. The whole "sovereign" thing is to basically say you won't be screwed if you go to war with a country tomorrow because you are dependent on their supply chain.
I don't think it is that basic. If anything it's probably more about having protection from trade and sanction threats, and insulation from shortages, supply chain problems.
> Nothing about this changes that IMO. They're still getting the chips fabbed somewhere else and run the same risks.
It certainly changes it. Chip design is of enormous importance and the technology leaders there are concentrated in the USA.
Silicon manufacturing is a different thing. Japan is also trying to bring up sovereign leading(ish) edge manufacturing. Obviously you keep peeling the onion on everything though - to do digital logic they need EDA tools and quite possibly IP blocks for PCI and DDR and so on. To do manufacturing they need scanners from ASML and a process design from IBM and so on... It doesn't mean that because they can't do everything from the ground up that it's pointless. At some point they need to import coal and iron ore and bauxite.
> Researchers asked frontier models to play chess. Have a look at the MAR rates in Table 3. When not explicitly told which moves were legal, no model identified legal moves at a rate better than 80%. Many asked for more illegal moves than legal moves. And even when explicitly told which moves were legal, the models continued to ask for illegal moves. With illegal asks discarded, none of the bots could beat a chess model calibrated to 1100 ELO.
It can be very interesting and even entertaining to know where models don't do well. I don't find something like chess to be very instructive about anything though, nobody is going to pay for AI to play chess at any significant scale even if it could do it perfectly.
> The author of the originating post says that "current frontier models need laborious oversight and guardrails on even the simplest tasks", and he's absolutely correct.
I have found that not to be the case, and I am not an AI power user or cheerleader by any means. There's probably a bunch of even "simplest" tasks where AI doesn't do well and might never. That doesn't take away from the cases where it works well and is a productivity booster. It doesn't even have to be solving millennium prizes or any other breakthrough creativity or reasearch, it's still very useful in places.
You don't need to be some kind of super genius to understand this, if you have not been sucked in to the populist left wing propaganda about petroleum. We can't even replace petroleum for energy right now, let alone all its other uses.
It's not just oil we'll have issues with either, lots of minerals could easily see shortages and rapidly increasing costs to extract over this century.
> We are doomed.
More propaganda. We don't need infinite economic growth, infinite population growth, or even continued quality of life growth. Nothing bad even happens when you have degrowth, all other life forms have periods of growth and degrowth. It's all crazy propaganda that people repeat we need to constantly keep boosting up population numbers.
A few generations ago nobody on earth ever saw a motor car for the entire history of humanity. Or electricity. Antibiotics. Synthetic fertilizers. Any kind of plastic.
Things will change, people will adapt, conflict and famine, etc. Much the same as the rest of human history. The people of the future might no lot be able to expect to be clothed, fed, housed, and entertained in exchange for sitting around an air-conditioned office eating a birthday cake before putting in their solid 2.5 hours of work for the day drawing up the next draft of policies around who can use which bathroom before scrolling tiktok for the rest of the work day. But I wouldn't call that "doomed".
In human history, when there has been a period of protracted degrowth something bad has very much happened - usually civilizational collapse or severe prolonged economic downturn.
> We don't need infinite economic growth, infinite population growth, or even continued quality of life growth
I agree on the first two, but I believe that an expectation of the third is foundational to the post-Enlightenment social order. It has waxed and waned, and there have obviously been incredibly dark periods, but we've usually returned to "things will be better in the future" as a setpoint.
I read "Degrowth" by Kohei Saito recently. I found it challenging (in that it challenged a lot of my preconceived notions), but also a fairly positive approach to implementing degrowth. Sadly, at least in the US, I think it is alien to the American mind and only has a hope of being implemented at the local level in sporadic places.
> In human history, when there has been a period of protracted degrowth something bad has very much happened - usually civilizational collapse or severe prolonged economic downturn.
Bad things have caused degrowth, that doesn't refute what I said though. The Edo Period in Japan is said to be a canonical example of degrowth that was not catastrophic. Japan today has had GDP per capita pretty steady for the past 35 years and so far nothing horrific, just the hilarious wailing and gnashing of teeth from western "experts". So I think those are sufficient counter-examples.
> I agree on the first two, but I believe that an expectation of the third is foundational to the post-Enlightenment social order. It has waxed and waned, and there have obviously been incredibly dark periods, but we've usually returned to "things will be better in the future" as a setpoint.
1. Have we, really?
2. Various objective and subjective quality of life metrics have been going backwards in some western countries for a generation now.
3. What about it is it that you think is foundational to social order? What happens if we don't have it.?
In any case it's an unsustainable expectation that is going to have to change one way or another.
If we're going by feelings and anecdotes, my feeling is that people in general don't actually need that at all. They want some amount of fairness, freedom, self-determination, and the ability to personally get ahead. They don't need to be ahead of where their parents were at the same age, they want to be able to be ahead of where they were yesterday (at least in economic terms).
> But if Meta is truly interested in prioritizing the safety of users, the board should request the resignation of Mark Zuckerberg and restructure to put safety above profit.
It is singularly idiotic that people think people, companies, boards, governments, or anyone else should be "good", and that society will progress by rooting out "evil" ones, as adjudicated by their own standards.
People and by extension companies of people and by extension corporations, NGOs, governments, and everything else in society behaves according to incentives. "Safety of users" or "approval of The Guardian employees" does not rank particularly highly in the incentive scale next to money for anybody, least of all Facebook shareholders, lol.
The factors that influence behaviors are pretty simple. Money, negative money, threat of judicial or ultrajudicial retribution, etc. The people as a whole have levers to pull in order to shape society as they would like. Wishing for good and praying for evil to be dispelled are not among them, that's just like any other superstition.
The far right won office and made it manifest. Just like they complain on government inefficiency and that it can never accomplish anything while explicitly making those things worse while in power to prove their point.
Yes possibly so I'd have to see the evidence for that, but my point is it's great that such a large majority of people are now well aware of the horrific and disgusting corruption, inefficiency, and incompetence in governments.
For some quick and easy evidence, you could look at the history of the Post Office.
USPS has always been the poster child for efficient and effective government agencies. Despite claims otherwise, the agency still manages to deliver letters and other postage to areas private industry like FedEx, UPS, et. al. would never touch.
Unfortunately, over the years, Republican Congresses and Presidents have repeatedly attempted to sabotage the agency. Some recent(-ish) examples:
* The 2006 Postal Accountability and Enhancement Act, which required USPS to create a fund to pay the cost of post-retirement health care costs 75 years into the future.
* The Republican-appointed postmaster general Louis DeJoy overseeing the removal of 711 sorting machines as well as a number of public mailboxes in the lead-up to the 2020 election, which, due to Covid, was expected to have larger-than-normal use of mail-in ballots. Not only did it cause significant concern regarding mail-in ballots, but it also resulted in significant delays and backlogs in general mail delivery.
You were not clear at all, I am asking for clarification. I thought that `atmavatar's question was clearly in response to your statement and you are saying it is not.
Assume I am ignorant to the context of this conversation if need be. What were you referencing when you made your claim to needing evidence?
> Whilst processing in memory is clearly the future
How clear is that? The idea has been around for about 60 years, and many attempts made by people who thought the same thing. Maybe this time it'll be the future.
Maybe. The use cases have always been around FP arithmetic over arrays, because that's what is easy to parallelize. I saw a standalone systolic array box attached to a MicroVAX circa 1990. Lots of PIM approaches in the mid-90s, too, but mostly what survived from that era are shared memory NUMA multiprocessors and using GPUs for general purpose computing.
This is really just restating the assertion I'm asking about. Previous work also thought their use cases fit the needs.
Everything changes all the time in computing, so there are lots of differences. I'm not asking if things are different or claiming it won't work this time either. I'm asking what it is this time that makes it different / obvious. Not rhetorical, I'm interested if someone can actually explain. Preferably with trends and numbers that have $ and flops and picojoules and gates and bits in their units.
> The western world has more or less come together to reduce emissions. If you look at US for instance emissions are down in abs terms. Not relative to gdp or per capital but total emissions. This holds even if you adjust for trade.
It certainly is reduced per-production. China has double the carbon emission intensity of production that the USA has for example.
Hah. Australia's economy is built entirely around two things. 1. Extracting fossil fuels and minerals. 2. "services".
During a global catastrophe Australia would run out of fuel within a few months, fertilizer not long after that, and soon afterwards lose the ability to grow food let alone distribute it, or keep basic services running. The lights would go out because it has no capability or even wishful thinking to manufacture advanced electronics, precision parts, much in the way of plastics and chemicals to keep airplanes, trucks, cars, computers, communication, phones, power plants, transformers, railways, or tractors running. Basic antibiotics, anaesthetics, insulin, any kind of medical or surgical tools or technologies? Don't even bother. Most people who survived the initial waves of panic and looting and destruction would soon starve and freeze to death in the dark surrounded by coal mines and fields of wheat rotting where it lies.
Also, Australia is a strategic point in the southern Pacific and Indian oceans with submarine bases and communications, air bases, signals and radar stations. It was believed to be one of the countries targeted first by the Soviets after USA if it came to a nuclear exchange. The calculus might be a bit different with China today but Australia is not some neutral party that would escape unnoticed in the event of a world war.
I don't know what would be the safest place, but I know Australia is not it. Somewhere where people can actually build and make things, and whose people might be inclined to cooperate and be capable of enduring some hardship, would fare a much better chance.
Does it use vector? What can you hit with SMT disabled?
reply