HarHarVeryFunny

> Meanwhile, Wired found that Muse consistently creates detailed profiles of all your friends, family, colleagues, “collaborators,” and people you “follow.” Obviously much of that information is necessary for the agent to get to “know” you, but this being Meta, people are understandably uncomfortable with this sort of massive ramp up of data collection

This is pretty funny! What ramp up? Building detailed profiles of you and your friends/etc is Meta's core business. They even build profiles for people who are not signed up for FaceBook, waiting/hoping for the day you join.

> Being a doctor is one of the jobs expected to be least at risk from AI replacement

I wouldn't say that, but it's certainly one where the consequences are high and doctor-in-the-loop seems highly desirable for the foreseeable future, especially since most patients are not going to be knowledgeable enough to realize the shortcomings of an LLM and make an informed decision about whether they're OK taking an LLM-prescribed medication.

Sure, that's what language model pre-training gives you - a reactive model where any input elicits an automatic built-in response.

Of course you'd like an AI to be capable of system 2 thinking as well, proactive planning and reasoning, which is more where they fall short, other than "reactive reasoning" (coming from the pre-training data), and pre-trained narrow reasoning (coming from RL post-training).

So, sure, in as much as an LLM can do system 1 thinking out of the box, it can compete with Jev, and how well it scores will depend on the type/quantity of data each was trained on (which can change in a future release).

It seems fashionable to say "look, my LLM can do Jev too!", but IMO this is completely missing the point. Any LLM, however dumb, can do system 1 thinking, and if it's too dumb then just use a bigger or better trained model. The competitive edge of Jev (especially for the high volume business automation market it is targeting) is not to do something that couldn't be done with an LLM, but rather to do it a lot cheaper and faster based on a more specialized architecture (with the bonus of reliable structured output and calibrated responses).

Is this a first - an article that talks about consciousness and actually defines what it is talking about?!

Better yet, this "functional consciousness" really does get to the core of what consciousness is, and why it may have evolved, and is not only something measurable, but also something that can be gauged from architecture alone (even if testing for it is less error prone).

Naturally this is too mechanical for some, who will therefore switch the conversation to qualia and what it "feels like" to be functionally conscious, and the simple answer to that is to (once you've ascertained that the system really is functionally conscious) ask it.

Note that just asking an LLM if it is functionally conscious, or how it feels, without having first ascertained that it is (and to what degree, and in which ways), would be meaningless since if it is NOT functionally conscious but sounds as if it is, then that is just a language model parroting human feels.

Sounds more like Woody Allen's 'Zelig'.

It's what this article is saying they heard From the people they talked to.

But I've no idea what Russian state media is trying to tell people.

Individuals as well as companies, that have worldwide impact, do get noticed, whether via the results themselves or peer-nominated recognition such as as Nobel Prizes, or the Fields Medal.

However, intellectual geniuses are better measured as people who revolutionized or significantly advanced the fields they worked in, and may have received recognition for doing so, not necessarily those who had highly visible worldwide impact, which is often more a matter of luck and different character attributes.

The people we recognize and celebrate as geniuses also tend to be people who had many accomplishments (because they were geniuses, not just one-time lucky), sometime seeing success on whatever they focused their attention on.

We wouldn't consider Tim Berners-Lee as a genius even though he certainly had world changing impact. Ditto for Demis Hassabis - certainly very smart, but hardly a Feymann. Someone like William Shockley also changed the world, and got plenty of recognition despite working for a commercial lab, but perhaps better regarded just as a bright engineer, right time right place.

AI is an interesting case, certainly changing the world, but the people who invented the tech, primarily Jacob Uszkoreit and Noam Shazeer, are really more akin to Berners-Lee and Shockley.

People like Von Neumann and Feymann, who shocked other geniuses with their intellect, really are a very rare breed. For people like this I'm not sure that "initial conditions" make much difference - they play by their own rules, and the world comes to them.

I don't think you can compare few-times-per-generation people like Von Neumann and Feymann with run of the mill smart PhDs.

It seems the really smart people tend to gravitate to jobs/environments where they can use their intellect rather than just looking for money. You are not going to find an Ed Witten or Terrance Tao working for Jane Street, nor would Von Neumann or Feymann be found at such a place if they were alive today.

If we had a national Manhatten project for next-gen AI, recruiting the brightest and best, then a present day Von Neumann might be there, but I'm pretty sure he would not be sitting at his desk vibe coding RL training environments.

> We do not know the aims [of the war]!

> We will accept anything the higher-ups define as victory

It seems that realistically the way the war will eventually end is without Russian victory, but with Russia withdrawing and either claiming "mission accomplished", or shifting strategic interests such as better relations with Europe and/or the US.

The fact that most of the Russian population don't understand, or won't accept, the goal of the war, just makes it politically easier for someone to eventually say that it is time to end it, which is anyways going to be popular the more it is being felt at home in Russia.

How/when such at outcome may happen remains to be seem. With the Soviet war in Afghanistan it took new leadership in the form of Gorbachev to end the war in similar fashion, and that may be true here as well, although it seems it may require the Russian situation to deteriorate further, and for the West to consistently communicate and push for some "win-win" outcome before any such leadership change happens and becomes entrenched in their own position on the war.

I suppose you're pointing out that pre-RL models were less jagged hence more general (universally dumb)?

I came here to say the same thing.

AI is here to stay - will be around for hundreds/millions of years. Whether company A is a few years ahead of company B is irrelevant. In 10 years time company A, who started the industry, may be gone completely - also irrelevant.

Like it or not, AI has become part of modern warfare and well as part of the surveillance state (good for crime fighting, not for privacy), so it is an important part of national security, and you'd prefer to be as much self-sufficient for that as you can be.

Not all needs are going to be met by using or finetuning general purpose models, so ability to build your own SOTA ML models (LLMs or not) is also important.

It's interesting to read Von Neumann's perspective on CRT ("iconoscope") memory - saying how important it was not to lose the serial aspect of it since of course(!) you would not want to specify the address for every access, with most memory accesses being sequential (presumably meaning each bit of a 44-bit word), but it was advantageous to gain the speed of occasionally switching read location, which would be slower and more costly to do with delay lines (more, shorter, delay lines).

Note that Atanasoff's capacitor-based memory was based on rows of capacitors arranged around a mechanical rotating drum (a bit reminiscent of the later EDSAC's "initial orders" module!), so it wasn't true random access - there would be a rotational delay to get access to a given row. In a way it was as much a predecessor of rotating mass storage devices as it was of DRAM.

I just missed the core memory era myself, at least as a user, learning to program (while in highschool) on an IBM mainframe in the mid-late 70's, which was already using semiconductor memory. Still, years ago I bought a core memory module and a (very similar size) etched 6" silicon wafer, intending to frame them side-by-side as the defining story of my era - the switch from physical to solid state integrated electronics.

Feymann shared these qualities too, but seems he would have been more interesting to meet - more of a normal, yet brilliantly smart, person, unlike Von Neumann who while being a bon vivant also seems to have been a bit "off" with his inability to drive safely, not remembering where the drinking glasses in his own house were, etc.

Von Neumann's architecture (i.e. report on the EDVAC) didn't focus on random-access memory, and in fact his whole discussion on memory is rather rambling. The EDVAC itself, which the report is nominally describing, used serial-access mercury delay line memory. What really matters from an architectural POV is memory address-ability, not speed of access (random vs serial).

The whole document is, as the title promises, really more of an engineering one - a report on the EDVAC - rather than some clean abstract treatise on computer architecture.

https://web.archive.org/web/20130314123032/http://qss.stanfo...

It doesn't cheapen the product, but it may cheapen the experience of developing it. Part of the satisfaction of being a developer is overcoming the complexity and challenges of the task, learning something new, some amount of struggle. You're not going to get that same satisfaction from a task that is too easy, or has become too easy.

The current state of AI seems to be more about automation of well defined tasks rather than things like creativity and challenging design that bucks convention, or requires human sensibility, so I think there is still room to get satisfaction from that aspect of it (you do the design, AI does the coding).

People point out how computer chess hasn't killed human chess, but I don't see much satisfaction from doing any development work by hand that could equally well (at least as far as the resulting artifact goes) be done by one/few-shot prompting, even though part of the fun of being a developer is being able to conjure dreams into reality - the end result is part of the satisfaction, but the challenge in getting there also matters.

The big AI companies like OpenAI and Anthropic may have rushed into it too fast, and in effect wasted a lot of money on early model generations that never paid for themselves, early model architectures that were not cost-efficient, experiments that never panned out, etc.

The AI pure-play companies like OpenAI and Anthropic are presumably in worst financial position, since others like Google, Microsoft, Amazon and Meta (also Alibaba, Baidu, Bytedance, Tencent, Xiaomi) have generally been more conservative and are treating AI as an incremental revenue source not a highly leveraged all-in bet.

The LLM bubble may fizzle rather than pop in any dramatic fashion, but it's certainly not the last word in AI - much more the exact opposite.

We've had AI "bubbles" / hype-fests before (expert systems, Japanese 5th generation systems), and I don't expect this will be the last.

There will also be technology bubbles and hype-fests in the future outside of AI - it's human nature. Hard to predict what they will be of course (the future is like that).

Of course they will come down. Supply will rise to meet any sustained demand - it always does.

Depends on what kind of memory you are talking about.

For DRAM, CXMT were only 1% of the global supply 3 years ago, and are now already 10%. They are growing extremely rapidly.

If you are concerned about consumer memory - smartphones, laptops, desktops, then this is what CXMT are making. CXMT do make HBM too, for AI accelerators, but their volume there is held back by sanctions.

> I have been telling folks for the last 3 years that the price of AI compute

What does that even mean ?

The cost of AI itself (tokens/$ for a given level of intelligence) has been falling rapidly, and shows no sign of stopping.

It had mostly come down to 3 (Samsung, SK Hynix, Micron), but the Chinese manufacture's market share is growing rapidly, especially as they ramp up due to sanctions. CXMT is currently 10% of global DRAM, having seen previous YOY growth of 1%->4%->8%. YMTC is now global #3 in NAND.

It wouldn't be surprising to also see Japan make a comeback in memory - they are investing heavily (both privately & government) in semiconductor manufacturing.

I don't think the semiconductor business is so much like that - it's become more ingrained in them that you need to manage the business for long-term sustainability.

AL/ML isn't going away, but what that means for demand for various types of memory remains to be seen. LLM are clearly a commodity, and once they are "good enough" then it'll just be about price, meaning that smaller and less memory hungry models will win.

It's a bit like the Sun/etc workstation market vs PCs, and high end expensive PCs vs cheap ones. Once the cheap PCs became "good enough" then they naturally dominated.

The memory business has always been extremely cyclical, and I doubt this time will be any different. The manufacturers have learnt this, and therefore to avoid boom followed by bust they try build for what they perceive to be longer-term more sustainable demand, not just knee-jerk increase demand to meet an immediate shortage.

The risk to any of the manufacturers, especially those with most market share, would seem to be if they are not all acting in sync. Micron's consumer brand, Crucial, had a great reputation, but now they have exited that business in favor or devoting all their capacity to currently higher margin products... will this come back to bite them later if they want to re-enter the consumer business?

Regardless of how these large scale coordinated dynamics arose, initially functional and selected for, or just epiphenomena (coupled oscillators?), it seems that they HAVE to have become functional (presumably longer range inter-region communication and sequencing) since timing does matter.

If we assume that brain waves do have (have to have) some effect, then it seems most likely that the effect is either beneficial or detrimental compared to the alternative of uncorrelated brain activity, and therefore is being selected for or against, and as the production of millions of years of evolution, it seems the logical conclusion is that there is some functional benefit to it. The alternative is that the functional effect is merely benign, neither beneficial or detrimental, but this intuitively seems less likely.

Calling it a "mistake" vs just a lack of knowledge is a weird way of framing it. There was much discussion of "junk DNA" (now referred to as non-coding DNA) for years/decades - everyone realized it was an anomaly to be explained, but no-one knew the explanation. This comprises the vast majority (98%) of DNA - so clearly it was there for some reason.

Brain waves are due to feedback paths and dynamics - oscillations etc, but LLM connectivity is far simpler. In an LLM there is a strict bottom-up one-way flow of information, with no looping. There is no way for a higher up layer to affect a lower down layer, whether directly or indirectly.

At some point you just want something that works and is somewhat familiar, so sticking with a distro makes sense even when it becomes somewhat annoying.

I remember the early days of Linux when I'd be using a dial-up modem to overnight download Linus' hot-off-the-press kernel tarballs to build myself and update my system ... nowadays I just want a working computer with minimum fuss, and cost (else a MacBook might make more sense).

I think the poster just meant that whatever distro you choose, you don't have to stick with the default desktop and look and feel if you don't like it. If you are brave enough, then give your agentic AI CLI access. Personally I just ask AI on my phone and pick and choose from the advice.