If your data is leaking to OpenRouter, how many companies would be OK with that? So I doubt OpenRouter are peeking inside the workloads that flow through their proxy. However, they do publish "Top Models by Task" ranking[1] so maybe you have a point.
It's not only going to be "connection maker". If and when robotics advance to a point where the LLMs are embodied, they can run experiments in the physical world and find new knowledge.
So applied math, math, applied math, applied math, and maybe some others.
That makes sense, although I'd argue that at least in the realm of HEP theoretical physics has extraordinarily expensive kit compared to what scientists make. See: The LHC.
I think it's telling that "Researchers cost more than any other aspect of research" seems to involve primarily math as inputs and outputs. Like I said the obvious outlier is theoretical physics because that tends to involve significant resources both to explore (supercomputers) and verify (colliders, telescopes, interferometers, etc). Most people involved in research in any number of fields do not find that their collective salaries are the majority of the cost required; math is an outlier in many cases.
Robots in labs already exist, but mercifully they're not hooked up to anything as unpredictable as an LLM. Robots tend to work best as specialists doing high-throughput, extremely repetitive tasks which nonetheless require a degree of precision. Giving a robot a "human" body makes very little sense if we're talking about the needs and productivity of a non-human; humanoid robots are marketing for humans.
Humanoid robots are obviously more than marketing. The entirety of human civilization is human shaped. Making robots that are human shaped is easier and more efficient than redesigning and rebuilding everything that exists.
> The entirety of human civilization is human shaped.
That's the marketing pitch.
A plumbing robot doesn't need to be humanoid, an octopus shape may well be better for all the awkward corners. A robot police officer could be the municipality itself for sensory nodes (essentially the sales pitch of Flock etc.), plus some drones or robot dogs to perform arrests*.
The robot vacuum cleaners and lawnmowers we already have are nothing like a human. A robot taxi driver can be just the car. Robot dogs are already used for maintenance and security sweeps.
If you've got wheelchair access, you've got wheeled robot access. If you've got guide dog access, you've got access for Boston Dynamics' Spot.
* this may be a bad idea with current robotics, but I aver it's not improved by making those robotics humanoid.
You are not understanding what humanoid robots are about. Those are specialist robots you are describing. The promise is of course one robot that can do the plumbing, clean your house, do the dishes, build a house, and basically every physical job a human can do. It's extreme lly likely that at least a somewhat humanoid shape is required for that.
Humans are really bad at everything we do compared to the specialized creatures in nature, it’s just that we can do many things that sets us apart.
A humanoid robot is incredibly dumb for that reason. At least add a set of arms and legs and 360 vision. And obviously dislocated joints than can move freely and hands should have two thumbs and more fingers.
It’s not hard to come up with a bunch of improvements for humans, it’s just that making robots in our image is a lot more trivial because you only have to solve for those same averages attributes that we have.
Adding arms and 360 vision is not free. Look at octopi with their insane nervous system required to support their tentacle. The one who wins is not the one that makes the best robot ever. The one who wins is the one who builds a robot that can do the most, while being the cheapest.
Vision is hard regardless of the angle, but scaling it from a normal camera to 360° doesn't add much cost or difficulty once you've done the hard part of turning pixels into a suitable latent space.
> Look at octopi with their insane nervous system required to support their tentacle.
500 million neurons across the whole animal, brain included. We'll only know the synapse count when someone does a full connectome scan of one, but based on the vague estimates I see with a quick search, their whole mind is less complex than a SotA LLM today.
> The one who wins is the one who builds a robot that can do the most, while being the cheapest.
This is a reason for specialists, not generalists.
The best container ship is a terrible pleasure yacht, and vice versa. You use container ships pretty much constantly, even though you (almost certainly) don't own one, by the power of indirection. You can rent a pleasure yacht as desired without owning own.
I will ignore the insane comparison between neurons and LLMs. But
> This is a reason for specialists, not generalists.
Precisely not. The amount of work a human can do is basically unenumerable. Requiring a specialist for each task will quickly run up costs to be completely unreasonable.
> I will ignore the insane comparison between neurons and LLMs. But
Please don't.
I mean, you yourself specifically said to "Look at octopi with their insane nervous system", and that's exactly what I'm doing and my conclusion is "doesn't look that bad compared to what else we do now".
Also note that I expressly noted that synapses, not neurons, are the important factor in parameter count.
> The amount of work a human can do is basically unenumerable.
> Requiring a specialist for each task will quickly run up costs to be completely unreasonable.
The existence of specialists in professional settings proves this claim false.
As does the existence of special-purpose robots, all the way down to injection moulds designed for one specific part, which overall make the end product cheaper, a process which probably started with whoever invented whichever came first of the wind or water powered grain mill.
Much of the economic gain of the industrial revolution was replacing the very expensive salaries of general-purpose humans with machines that, despite being extremely single-purpose and having high purchase and maintenance costs, were still cheaper than the wages of the huge number of humans needed for equal output.
Since the industrial revolution we have been the species that makes tools even though it sometimes also uses slavery, not the species that relies on slaves and has tools for decorative reasons.
Yeah. Can't wait to see three or four armed robots. Or five or six!
How do you train a robot to use three hands effectively when we only have two? Then again, why is the robot limited to being one robot? If two humanoid robots are in the same area, they don't have to be distinctly controlled. If they're both controlled by the same AI, a third arm on one body is the same as that arm being attached to another body.
Things you need to take into account that you've almost certainly glossed over:
Safety. Human-robot interactions are generally dangerous and avoided, unless the robot is specifically designed to interact with people. In those cases you often sacrifice speed, strength, and flexibility for safety and softness. Having someone come in with a specialized plumbing robot makes sense, you owning one probably doesn't, and you owning a generalize android capable of plumbing makes less sense still.
Cost. The more compact, complex, and interactive your robot is the more it costs. Make a strong, compact, complex robot safe for interactions with people in the wild is non-trivial and adds costs. The software required to do all of this is hypothetical, but obviously also costly.
Need. I understand the dream of a robot to do whatever you want is very much part of our culture, but when you consider the downsides do you really need it? I don't need a plumber living in my house any more than I need a carpenter or a landscaper to live on premises. At most these are services I would need occasionally or on a schedule. I also doubt my need for them will overlap much, unless we're talking about building a new dwelling.
So why do I need a generalist in my life that's going to cost more than you can imagine, when the means to hire existing human generalists is cheap, quick, and frankly less likely to accidentally punch a hole in you.
Generalist humanoid robots that are able to operate in unstructured environments weren't an option (they still aren't an option for the majority of operations). The humanity had no ways of building them.
Anyway, it's true that replacing an automated production line with a crowd of generalist robots doesn't make sense. Generalist humanoid robots are intended to replace the remaining human workers.
They weren't an option because they are much harder to build, that's my point. Fat from being the easy option, they are still the unlikely option, compared to purpose designed machines.
Historically this has never been correct. Turns out you get more efficient systems when designing them without how a human would accomplish a task in mind.
Outside of sci-fi, marketing proposals, and niches like "Elder care in Japan" there are very few humanoid robots. By contrast non-humanoid robots have been mass produced and used in industry for decades. Arms. Carts. Trollies.
No people. If you want something with fine motor control and dexterity, it's easier to make that the robot and then have another robot bring the workpiece to the arm than it is to build a single robot that can walk around and do it. There are compromises in human features because we're generalists.
Is it not dangerous to ignore text based on its color? For this particular case it might work but looks like you have a generic service that is supposed to work with any bank statement.
Yes it’s not a safe general rule. I can do it because I first classify the statement into a document_type and then run special code for specific document_types.
But I don’t have to use any apps from the store, in fact I’d say that at my company there aren’t any apps available on the store that we use. So the parent is right, it’s perfectly usable without ever touching iCloud. I don’t think I’ve ever even opened the settings for it on my work laptop
I'm saying it's bad because my new work laptop has Numbers instead of Excel ;_;
There are CSVs I had to import into GSheets because Numbers was choking.
I think it was when the LLM asked me a question at the end of its response. It felt like something other than a machine. Until then the pattern was me asking a question and ChatGPT giving me an answer, with or without hallucination. When it asked me a follow-up question it felt like talking to a being with agency. An entity that has thoughts or ideas or questions of its own.
I think this is probably related to which eye is more dominant for you. I've never skateboarded, but if I imagine myself doing it, it would also be facing left. And it's because my right eye is dominant and I would like that to be facing forward.
> I think this is probably related to which eye is more dominant for you.
I think it's more about a person's personal stability/biomechanics. The back foot is the stable one, forward is the "quick" lead. The preferred "plant" foot when kicking a ball is the stable one (though many people use both, so this is best used only when there is a strong preference), the one used to push off with when at the bottom of stairs or jumping is the stable one (the lead foot is the `quick` one). The best way I found to help determine footedness: have a person stand straight (feet together) walk around them (pretending to look at posture or something), once behind them push them forward (evenly with some force). Watch for which foot they catch themselves with. Thats the lead foot.
As for the eye dominance, in archery having the right eye dominant means your stance is regular (left foot forward). An archery open stance is near identical to a snowboard neutral stance (~ +15°, 0°). The 2 most important things to get right in (olympic recurve) archery is eye dominance and a proper open stance. As a goofy footed snowboarder and a right eye dominant archer, the archery stance took awhile for me to adapt too. It still feels weird.
Quick test to find out which is pick a point in the distance, make a triangle with thumbs and fingers to look through, and slowly bring it toward your face. Wherever it ends up is your dominant eye.
By "completely Nvidia-free" do you mean Nvidia wasn't used for training nor inference? Because if it's only inference, we know that Opus already can run on TPUs. Not to mention Gemini.
Yep but they don't run on Chinese hardware that is going to be available to everybody and will cost a lot less than NVIDIA stuff. So now you have a full non-US pipeline for AI, and soon they'll have the training GPUs as well.
Thank you for the link. I tried to opt out. They sent me this email:
Equifax Workforce Solutions (provider of The Work Number) has received your employee request communication, but additional information described below is required to fulfill this request.
We will be following up with a secure email to obtain the below requested documents:
Proof of Identity:
Provide a copy of one of the following (must include current/legal name):
- Driver's License (must be current)
- Paystubs (must be dated within 60 days)
- State or Government Identification Card (must be current)
- Social Security Card
- Military Identification Card
- Passport (must be issued from U.S.A. and be current)
- W-2 or 1099 Form (most current year)
- Birth Certificate
Proof of Address:
If you are requesting an Employment Data Report (EDR) or selected ‘Mail’ as your preferred method of contact,provide a copy of one of the following (must include current mailing address and be issued within the past 60 days)
- Driver's License (must be current)
- Paystub
- W-2 or 1099 Form (most current year)
- Utility Bill (phone, water, gas, electric, trash or sewer, etc.)
- Housing Rental Agreement or Mortgage document - your name must be listed on the document
For Identity Theft Block Requests, along with Proof of Identity and Proof of Address (if applicable), please provide your identity theft report and designation of items to be blocked:
- Identity Theft Report (police report, FTC Identity Theft Report, Police report, or United States Postal Inspection Service)
For Human Trafficking Victim Block Requests ONLY, along with Proof of Identity and Proof of Address (if applicable), please provide victim determination documentation (as described below), and designation of items to be blocked.
Victim Determination Documentation:
Provide a copy of one of the following victim determination documentation confirming that you were a victim of human trafficking, such as:
- Determinations made by federal, state, tribal, or local governments, government agencies, or law enforcement
- Determinations by non-governmental entities or task forces authorized by a governmental agency to make such a determination
- Self-attestation signed or certified by such governmental agency or non-governmental entity
- Determination by court in a case where a central issue is whether you are a victim of human trafficking. (Court documents can be made up of several documents from the court case that together show that the court accepted as true or finding no genuine dispute that you were a victim of human trafficking.)
We will be following up with a secure email to obtain the requested documents.
Data Investigation Team
Equifax Workforce Solutions
Obviously they have to be careful. What if they didn't check all this and someone went and tried to opt out on your behalf? That would be an incredible invasion of your privacy!
Obviously. On the other hand, your e ployer sharing your personal data against your knowledge or your will isn’t an incredible invasion of your privacy. Everything is fine citizen, move along and quit asking questions or thinking.
Mining isn't the only bottleneck with rare earths. There also the processing, which is an industry China has monopolized through sustained investments over decades. They have also improved processing efficiency through investments in technology. It's going to take a while for anyone else to catch up.
> There also the processing, which is an industry China has monopolized through sustained investments over decades.
I don't think this is the right way to characterize it. China invested when other countries didn't, but they didn't monopolize the market, they have no moat beyond expertise and some tech advancement that could be replicated easily enough. The only moat they have is related perseverance and other countries simply not wanting to put the work in.
I think they do have a moat because they dominate the supply chain not just in the raw material and processing but also in some of the actual technical experience, i.e. the experience of running such processing facilities, and also a monopoly on making the equipment that you need to build such a facility. They put export controls on those equipment and restricted their citizens who work in the rare earths industry from traveling aboard.
Basically, if we want to replicate what they did, we will have to do it mostly from scratch -- Japan and Australia has done some of the work already so it's not totally from scratch. It's obviously not impossible but it could take almost a decade for us to do that.
That said, I don't think this should be enough for Japan to stop investing in EVs. If Japanese car makers are really worried about this then they can build their plants in the US and leverage any deal the US has with China on real earths. They've already starting importing Japanese cars made in India and the US back to Japan so that's an established practice. Then once they've secured their own supplies they can make the EVs in Japan too. I think OP's point about the suppliers have more merit as a reason why Japan might not want to develop EVs.
I have worked with the Chinese REE industry, and we've often bumped heads and shared ideas together with them and I can confidently tell you, the Chinese don't use anything novel that has not been established in Western science already. What they do have is executing rarely-used techniques confidently at scale, but all of that is already often published in the West. The only reason the West hasn't done it is because these techniques are less profitable, and, surprise, the CCP actually forces processors to minimize ecological damage, which further bumps up the costs to the point only large-scale players can exist making such lower profits. You'll often find them using some obscure process alteration that was published minutely in the West.
As an addendum, companies in the REE Sinosphere are often encouraged by the CCP to exchange ideas with each other quite often, while Western companies often lock them behind proprietary patents and competition. While both systems have their pros and cons, the former allows for faster process proliferation (and a lower profit incentive for the innovator).
> the Chinese don't use anything novel that has not been established in Western science already
Like they say: in theory, theory and practice are the same; in practice, they aren't.
It's all well and good to have knowledge of the techniques, or to even have published or created them. But applying them successfully, working out all the kinks, and streamlining everything to become profitable doesn't happen overnight.
I have no doubt alternate sources can exist, but not without significant time and effort.
> my experience with theory is that it includes time and effort considerations
I would never disagree with you here. But the point is that the time and effort you spend on theory doesn't translate to time and effort spent on practice.
What I mean is that since the peak of American REE in the 1970s and 1980s(?) a lot of the engineers who have working knowledge are retired. There's nothing theoretical we can't dig up but I think there will need to be a number of years for the US to catch up in terms of craft knowledge or "metis" (as Dan Wang likes to call it) and processing equipment and plants.
Maybe I'm wrong. I gained my knowledge second-hand/third-hand from books and podcasts so I would defer to you to your actual experience and observations about Chinese REE. What is your estimate on how long it would take the West to catch to at least supply some of the rare earth components and what the real barriers might be? Would love to hear your take on this.
Thanks for sharing your observations. I had no idea about the minutiae of that industry, i.e. the ecological control and its effects on the industry.
No, you're right. China, and even India and Russia, also do not have the same talent problem of the West, in that there is an undersupply of engineers, especially in the geological, processing and chemical sectors. In the US, the average age of the chemical process engineer was touching 50 a few years back. The average age of a process safety engineer is well past 50. While Russia and India lose their technical talent to brain drain, the Chinese govt has done quite a lot in trying to reverse that.
The real barriers are talent and the regulation vs profit motive balance. What I mentioned in my previous comment was effectively an effect of the intersection of the two - you can't find novel ways of processing harmful substances without having the technical talent to find these out in the first place, nor without giving them a free reign after deprioritizing profit.
Let's take arsenic for instance, a substance that's a harmful byproduct arising out of most mining operations. We already have the technology in the West to lock away arsenic into glass, but a.) apart from the big ones, most companies are unaware of them, and b.) even if they were aware of it, the tech is a significant line item that shies investors and companies away from investing into it.
> What is your estimate on how long it would take the West to catch to at least supply some of the rare earth components and what the real barriers might be?
Never. Yes, there are a few companies still engaged in trying to secure REE supply (Glencore being the most notable), but due to Western regulatory and policy limbo, the answer is never. For this to change, you need regulators open to experimentations and a concerted effort by the government in trying to reestablish REE independence, both in extraction and in processing, but I have yet to see either happening. It's telling when frankly the US is the country in the West most likely to catch up still, but the gap is deeper than the Darien Gap .
>they have no moat beyond expertise and some tech advancement that could be replicated easily enough
Moat is decades of process / tactic knowledge built by disproportionate amount of talent on geologic formations others didn't invest in. Right now they generate 15x mining graduates, university of mining tech alone enrolls more than all US mining programs combined. Then you throw all that into a mining city like Batou with 3 million people running vertically integrated operation. That's ecosystem scale with compounded advantages beyond "wanting" to put work in, it maybe scale on PRC has demonstrated ability to produce.
Between shallow kiddy pool and Mariana Trencth in terms of ease of replication, I wouldn't lean towards kiddy pool. I don't think "right way to characterize" their lead is "no moat" beyond... all the things that are actually, in fact very deep moats, as if any country can persevere their way to replicate decades of work and execute industrial policy of a 3 million large city dedicated to mining/rees.
I surmise, PRC will build out EUV (technical problem) and produce them at scale before west+co meaningfully tackles HREEs supply chain (technical and regulatory and industrial problem).
> they have no moat beyond expertise and some tech advancement that could be replicated easily enough
Incorrect, de facto, the only firms invested heavily in the rare earth refineries technology are Chinese for the last 20-30 years. Their moats are as deep as TSMC moats so to say.
Processing is the thing china does, you don't really mine rare earths, they are in many areas. Sure there are substrates it's easier to extract from, but the massive pollution of the processing that china was willing to accept when others were not that allowed them to corner the market. It can be done more cleanly, the US has some processing for strategic reasons (not enough though), but doing it clean is _very_ expensive.
Lets hope the people modifying plants to concentrate elements make work.
As I understand it, some of these processes also require a sufficiently large industrial base to be even remotely economical due to a reliance on industrial 'byproduct' (for want of a better word). Because of this, some of these processes are not something that can be quickly stood up in isolation over a few years. It would take concerted large scale planning over a long time period - something the Chinese system of government is almost uniquely capable of.
[1] https://openrouter.ai/rankings#task-spend
reply