cornholio

Naturalistic fallacies will only carry you so far. For example, my 12 year old car has none of the incredibly adapted limbs and muscles of a cheetah, but can still easily exceed the animal land speed.

The article makes a compelling case that a certain kind of sensory input and learning is necessary to crack robotic movement in general, it remains to be seen if such a fine array of sensors as the human hand is useful outside very specific use-cases. A robot that can stock shelves reliably would still be immensely useful and very generalizable, even if it can't thread the needle due to limited fine sensory abilities.

In any case, it seems like a "simple" problem to solve. An accelerometer chip costs a few cents and the data rates can be handled by a very light wiring harness, ex I2C.

So embedding such a sensor in every rigid component, wiring a single data line to all of them (using the chassis as electrical ground) and feeding the data back to the model seems a trivial way to work around this problem without any kind of real pressure sensitivity. The model knows the inputs it gives to the actuators/servos, so it will quickly learn to predict the free mechanical behavior of the body, and use any deviation to derive data equivalent to pressure and force feedback.

Another possible source of data is the driving current of the motors/actuators which is proportional to the mechanical resistance the limb encounters. All sorts of garbage sources of data that were almost useless noise in the classical approach become valuable with a model large enough.

What I'm saying that direct debit (the ability to authorize a merchant to draw money from your account repeatedly at a later date, for example to settle a monthly bill, comparable to giving them your credit card number) is a different payment workflow than a SEPA money transfer (initiated by the consumer), which would be presumably be used for an internet settlement protocol.

The customer can ask their bank anything, but the SEPA terms allow reversal for very strict scenarios, only if the payment was erroneous, it was duplicated by a human or technical errors, the wrong amount was sent etc. So the bank will refuse demands based on consumer protection reasons, they correctly fulfilled the customer request to send the money and further disputes are handled privately.

You might be referring to SEPA direct debit, which can be reversed unilaterally by the client at any point during the 8 weeks. These are payments initiated by the merchant, using an already established direct debit channel.

Open just means the implementers of the protocol can select the settlement methods. Cryptocurrency allows in band settlement, but there is nothing wrong in principle with trusting finance institutions that the money really reached your account. So the settlement can be done out of band, using non-blockchain methods by mutually trusted financial institutions, just like credit card payments. It's cheaper, faster, simpler etc.

It's an economic and business problem, not a technical one: the incumbents earn massive fees and have very little incentive to innovate, while the disruptors can't reach the minimum critical mass they require to get rid of the damn credit cards (aka static passwords you need to disclose on the internet).

A truly open payment protocol should handle any currency (not force you to adopt crypto coins) and would leave it to the merchant and buyer to agree on their settlement method or intermediary, the cancelation and dispute policy etc.

It's entirely possible then that some options would be available, for example in SEPA, that offer instant and irreversible settlement at zero cost.

>> The point is to get people in those contries to overthrow their leadership.

>That's the theory, but has it ever worked?

The point is not to (directly) instigate regime change, but to influence the actions of the existing regime, as well as other state actors not under sanctions, by demonstrating to them how bad it can get. Make an example and so on.

The suffering of the civilians is not the goal of sanctions but a consequence of the choices their - legitimate or not - leaders make, and which ultimately impacts their ability to engage in foreign trade. No country has an obligation to trade with any other, so if civilians suffer after foreign trade is limited, the agency and moral responsibility is with the regime that failed to secure friendly trade relations. Often, humanitarian exceptions are carved out to limit this.

It definitely "works", in the sense that it's often the only tool available, along with positive reinforcements such as aid and support and the threat of stopping them, which is just another flavor of the same. It's hard to have a benchmark for something that "works" better, since countries are sovereign and by definition have disputes and don't blindly follow any established rules or rulers.

Well, sure they exist, stealing and murder still exists too despite our best efforts to eliminate them. The point is that they are on the fringes of our society, I get maybe one spam email every few days or so and seldom bother to review the thousands trapped in the Spam folder. Robocalling never took of in my country either, and none of these scummy industries is receiving trillions of speculative investment like AI does. Social norms work, even imperfect as is their nature.

We will quickly evolve a social contract that AI are not allowed to directly contact humans and waste their time with input that was not reviewed by other humans, and any transgression should by swiftly penalized.

It's essentially spam, automatically generated content that is profitable in large volume because it offsets the real cost to the victims, by wasting their limited attention span.

If you wantme to read your text, you should have the common courtesy to at least put in a similar work beforehand and read it yourself at least once.

This is a totally predictable result, there are dozens of HPV strains and the vaccine will immunize against 9 of them that are high risk.

So, unless you are a sex worker or similar, it's unlikely you "have" all of them to the point where the vaccine is completely useless. You might later get infected with a strain that you didn't yet have, and it's precisely the one that kills you.

An alternate explanation is that, even if AI does not have any chance of entirely replacing an employee, not even junior hires, in the hands of competent seniors AI does substantially improve their productivity and eliminates whole classes of tasks that were traditionally where juniors cut their teeth.

So companies reduce junior hiring because their work is relatively less valuable, and they can meet their goals by shuffling existing resources. When they can't do that, they go for seniors since the immediate bang for the buck is higher (ofc, while depleting the global pipeline that actually produces seniors in the long run, in a typical tragedy of the commons)

> They want to play games, not operate a console.

I've gifted my decade old development laptop (after a beefy RAM+SSD upgrade to the best modern version it supports) to my 7 y/o nephew and he seems satisfied. It cold boots Windows 10 in less than 30 seconds and he can play Minecraft, Roblox, BeamNG, watch Youtube etc. in the living room where he can be supervised, without hoarding the family TV with their console.

Sure, a lower friction device is preferable, but the ultimate thing is that it plays the games they and their friends play.

That's the point I think, you can have some local page counters, broad region geolocation, referal stats etc. without selling your visitors privacy to some corporation, for some data that is not even that useful for you.

Apple still has quite atrocious performance per $. So it economically makes sense for a top end developer or designer, but perhaps not the entire workforce let alone the non-professional users, students etc.

The OP is comparing the SaaS offering with a vibe coded tool, and concludes that it's cheaper to spin your own. And I claim that, relatively speaking, any SaaS offering is much more "battle tested" then an automatically generated tool that only existed for 5 seconds.

Both can be true at the same time, the American market can subsidize research for the rest of the world, and one source of this subsidy could be US public programs leaving substantial cash on the table for big pharma.

Another perspective is that this extra spending is more closely correlated with global pharmaceutical profits and the spending by these companies for their US sales and lobby apparatus (which influences Medicare prescribing physicians just as well), and much less with actual fundamental research and clinical study spending required to bring these drugs to market. So it's relatively speaking "wastage" compared to direct medical research spending - perhaps through a global research institution funded by major economies. Hey, we can dream.

As to why would Medicare leave these untold billions on the table, I don't know the details of how they decide what to treat, but it's a very tricky political problem. Once you start to deal with things like DALYs, QALYs and micromorts, you are plowing straight into the "death panels" controversy. So all pharma has to do is refuse to play ball, continue to sell most of their products into the open market, and watch their Madicare centralized buyers quickly fold. You can only force Pharma to play if you institute a national monopsony; a single program, even a big one, can't possibly work since by acquiescing they would threaten profits in the rest of the market. For the same reason, a single US state could not go at it alone, pharma would just refuse to sell to that state at the imposed ceilings; it needs to be a nation wide program.

> most high earners that earn over 200k

Those incomes are by definition restricted to the top 5% of the general population - of course those top earners prefer not to be enrolled in a socialized medicine plan and prefer market insurance with a nominal price tag that is not proportional to income.

That's the entire point, to establish a social contract where the healthy and successful pay for the healthcare of the less fortunate, can benefit themselves from that social net if their fortune reverse, and deal with the risk they might never be net beneficiaries if they remain healthy and wealthy.

Given the option, high earners would argue that all public spending should be financed by flat nominal taxes, it's not like they personally require more work from the army or the police, or they drive longer distances on public infrastructure.

That's the point, spreadsheets get users to solve issues algorithmically without forcing them to learn programming, but allowing them to understand and control the solution.

Full stack vibe tools solve some 1st order approximation of the problem, but then leave the user clueless to how adapt and evolve the solution without learning programing - and in fact are harder to maintain and debug even for those who do know programming. Re-applying the same tool is widely understood as a guaranteed disaster in the long run, current vibe tools can't really maintain their own creations.

> Each of these could cost you $20-50/month in SaaS subscriptions.

Sounds to me that's a bargain, for half an hour in wages I get a battle tested solution built by real humans with real human feedback in the real world, instead of some unmaintainable oneshot AI approximation with zero support. I would easily waste more than that trying to wrestle some prompt into doing what I need, and even when it finally does it's still spaghetti code that will eventually fall behind surrounding reality and lock my precious data into a custom abomination.

Clearly the value proposition here is not replacing SaaS solutions but creating very customized solutions that simply don't exist as a product - just like spreadsheets in the previous era. A spreadsheet does not pretend it is a programming language, it just solves the data processing problem the user has in a visual and intuitive way.

This is to me the most glaring issue with vibe tools, they do seemingly advanced stuff but their users don't really understand the result they get, they are just firing their harpoon and get dragged wherever the whale will take them.

This part is very relevant:

> There is an income threshold—EUR 5,175 per month or EUR 62,100 (as of 2024), at which the percentage contribution is capped. What this means is that if your income increases, then this percentage remains fixed and doesn’t change with your salary increase.

Germany has a quite progressive tax scheme and the brackets are designed to be complementary, so that as income raises the pension and health benefits are capped while the general income tax grows more aggressively up to a total taxation rate around 45% for high incomes.

This is more than you would pay in the US but it includes both free healthcare for the rest of your life and a livable state pension after 67, and you get those benefits even if worse comes to worst and you suffer an accident, major disability etc. This is powerful social safety net that can't really be compared with what the American federal + state + private insurance offer you in return as benefits for your total tax+healthcare contribution.

I think there is credible research that US drug costs are primarily the result of a fragmented market and lack of collective bargaining. So a single buyer regime will by itself lower the prices since manufacturers will be forced to compete among themselves based on price and an objective metric of the therapeutic benefit - as opposed to marketing and the best sales network and "conference" benefits the prescribing physicians.

For example, if the generic option reduces the burden of disease by 10 DALYs and costs essentially ~0$ it will be fully covered, any other patented option with a 15 DALY average benefit will be covered up to a financial ceiling given by the total funding available and the extra 5DALY benefit it brings. A patented option that does not fit that ceiling will require patients to pay from their pocket, therefore cratering sales, therefore strongly incentivizing the pharmaceutical company to lower the price to fit the ceiling.

This is how other advanced countries lower spending on drugs without infringing on intellectual property - by forcing all manufacturers of all drugs for all diseases to compete on price, or risk selling basically nothing in that country until their patents expire. Since the marginal production cost of drugs is close to zero, it' always better to have some sales at some lower prices instead of no sales at a very profitable price.

"Medicare for all willing to pay 10% of taxable income" seems like a way to sneak universal healthcare through the political backdoor. It will initially have a funding shortfall due to adverse selection (it will attract primarily people with low incomes, high risks or both) but you would gradually move all public subsidies towards that program and remove all incentives, tax credits and rate protections from those who choose the free market option, and when just a minority still hangs on to the private providers, you make the 10% health tax mandatory to all.

The way I see it, you will never wrestle away the trillion dollar bone from the mouth of the US healthcare industry, but you can make it wither away indirectly by creating a better universal system and dismantling the status quo. It's not like Europeans are forbidden the option of private insurance, it's just that it can't take hold and grow to control a significant part of the spending because everyone can see it's not the solution that will save you when you will get really really sick.

The systemic problem is that those buttons exist to be be pushed, not that they are being pushed.

For example, all rules regarding single family occupancy - there is no gain society can possibility have from instituting such rules, and many countries do just fine without them.

But once you have them in place you have created a very strong incentive for incumbents to exclude others from diluting their access to public investments in the area, while also denying those people a political voice.

> not all 8 digit pins are "compromised"

Sure they have been, I can send you a text file with all of them. It's 850MB, but i expect it compresses very well.

The context after application of the algorithm is just text, something like 256k input tokens, each token representing a group of roughly 2-5 characters, encoded into 18-20 bits.

The active context during inference, inside the GPUs, explodes each token into a 12288 dimensions vector, so 4 orders of magnitude more VRAM, and is combined with the model weights, Gbytes in size, across multiple parallel attention heads. The final result are just more textual tokens, which you can easily ferry around main system RAM and send to the remote user.

Inference is essentially a very complex matrix algorithm run repeatedly on itself, each time the input matrix (context window) is shifted and the new generated tokens appended to the end. So, it's easy to multiplex all active sessions over limited hardware, a typical server can hold hundreds of thousands of active contexts in the main system ram, each less than 500KB and ferry them to the GPU nearly instantaneously as required.

You can knock off a zero or two just by time shifting the 700 million distinct users across a day/week and account for the mere minutes of compute time they will actually use in each interaction. So they might no see peaks higher than 10 million active inference session at the same time.

Conversely, you can't do the same thing as a self hosted user, you can't really bank your idle compute for a week and consume it all in a single serving, hence the much more expensive local hardware to reach the peak generation rate you need.

That's unnecessarily harsh. Patent pools exist to promote collaboration in a world with aggressive IP legislation, they are an answer to a specific environment and they incentivize participants to share their IP at a reasonable price to third parties. The incentive being that you will be left out of the pool, the other members will work around your patents while not licensing their own patents to you, so your own IP is now worthless since you can't work around theirs.

As long as IP law continues in the same form, the alternative to that is completely closed agreements among major companies that will push their own proprietary formats and aggressively enforce their patents.

The fair world where everyone is free to create a new thing, improve upon the frontier codecs, and get a fair reward for their efforts, is simply a fantasy without patent law reform. In the current geopolitical climate, it's very very unlikely for nations where these developments traditionally happened, such as US and western Europe, to weaken their IP laws.

Thank you for your very well informed perspective. Fusion proponents seem to be either industry insiders with vested interests, or less informed fanboys who are simply unaware of the technical and economic realities.

This issue of safety is particularly prone to handwaving; in reality, the combined effects of activation and proliferation risks and the substantial radionuclide release potential will make the operational realities, regulations, environmental litigation and associated costs very similar to fission.

That's not too say fusion is inherently dangerous, rather that modern fission projects are already very safe and fusion won't improve on that. Yet fision still failed. So if fusion can't improve on the economics - and they quite clearly can't for the foreseeable future - then they bring nothing to the table.

Solar energy has achieved such a cost reduction that nuclear can't compete even if the actual nuclear reactor part is free. Just the classical, steam turbine parts are becoming more expensive than solar, and this is evident for new natural gas plants, who don't even have any radiation concerns. Sure, fusion energy would be dispatchable, unlike solar, but momentum is building towards large scale interconnections, perhaps even at intercontinental levels, spanning many time-zones and climates and achieving highly reliable solar.

While it's unclear when all this will be achieved, nobody is ready to bet 10 billions that it won't happen in the next two decades they need to recover costs.