> ICU is obviously important in a different way. You can't really "experiment". Iteration needs data. So you need to go out and learn what good looks like from different disciplines, and then carefully plan the changes you want to make and get buy-in.
I would think test runs with simulated patients offer plenty of opportunity to experiment.
> Get it wrong, and people die. Best case scenario you're struck off, worst case you're going to prison for murder.
Get it right and people may still die. The whole reason for the improvement effort is that the current practice is excessively risky. No one is getting fired, nonetheless going to prison for trying a sensible improvement to reduce the odds of a child dying which they were approved to attempt.
Nope, it's a technological issue. Increasing immigration might address some of the symptoms of the issue, but it does nothing to address that a human being still needs to do this labor. Frankly even if you were to liberalize immigration laws, convincing people to upend their lives and move to a high cost of living country where cultural integration is difficult at best just to pick tomatoes is not exactly a trivial task. Even if you do get people to come for menial labor, as you say there are plenty of other areas like healthcare where labor is in high demand, so you're likely still going to be faced with labor shortages in less desirable fields. Immigration is a treatment, automation is a cure.
They are very clearly not. Humans are good at some things, like recognizing fruit in clusters. Robots are good at different things. Shifting from easy for a human to harvest to easy for a robot to harvest is both in theory and practice a radical change.
But the same things benefit both. And often "human" harvesting just means humans driving farm equipment, which are now basically totally automated.
Things like growing crops in rows. Things like grape vines being trained into rows. Things like nut trees bred so they shed their nuts when shaken by a mechanical shaker.
These things are the same, it's just how automated we go.
I've spent over a decade in the agtech/robotics space here in Norcal and everyone seems to have an opinion, until you actually go out to where our food is grown and find out it's already highly automated, robotic, etc. It's just not sexy in the VC tech way we need it to be to be cool to talk about.
Go to a modern farm, see how intertwined tech, farming, biology, all is. It didn't get this way overnight.
Put another way. What improvements to plants to benefit human harvestability can you think of wouldn't also improve robotic harvestability?
I don't think AI training on a dataset counts as a view in this context. The concern is predators getting off on what they've done, not developing tools to stop them.
Debating what counts as a view is irrelevant. Some child pornography subjects feel violated by any storage or use of their images. Government officials store and use them regardless.
AI doesn't know what either porn or children are. It finds correlations between aspects of inputs and the labels porn and children. Even if you did develop an advanced enough AI that could develop a good enough idea of what porn and children are, how would you ever verify that it is indeed capable of recognizing child porn without plugging in samples for it to flag?
LLMs don't "know" anything. But as you say, they can identify correlations between content "porn" and a target image; between content labeled "children" and a target image. If a target image scores high in both, then it can flag child porn, all without being trained on CSAM.
But things correlated with porn != porn and things correlated with children != children. For example, in our training set, no porn contains children, so the presence of children would mean it's not porn. Likewise all images of children are clothed, so no clothes means it's not a child. You know it's ridiculous because you know things, the AI does not.
Nevermind the importance of context, such as distinguishing a partially clothed child playing on a beach from a partially clothed child in a sexual situation.
So it is able to correlate an image as porn and also correlate an image as containing children. Seems like it should be able to apply an AND operation to this result and identify new images that are not part of the data set.
No, it found elements in an image that it tends to find in images labelled porn in the training data. It finds elements in an image it tends to find in images labelled child in the training data. If the training data is not representative, then the statistical inference is meaningless. Images that are unlike any in the training set may not trigger either category if they are lacking the things the AI expects to find, which may be quite irrelevant to what humans care about.
AI doesn’t understand context either — it can’t tell the difference between an innocent photo of a baby in a bathtub with a parent, a telehealth photo, or something malicious. Google is using AI in addition to hashing, and both systems can get it wrong. With AI you’re always dealing with confidence levels, not certainty. No model in history has ever had 100% confidence on anything.
A scanning system will never be perfect. But there is a better approach: what the FTC now requires Pornhub to do. Before an image is uploaded, the platform scans it; if it’s flagged as CSAM, it simply never enters the system. Platforms can set a low confidence threshold and block the upload entirely. If that creates too many false positives, you add an appeals process.
The key difference here is that upload-scanning stops distribution before it starts.
What Google is doing is scanning private cloud storage after upload and then destroying accounts when their AI misfires. That doesn’t prevent distribution — it just creates collateral damage.
It also floods NCMEC with automated false reports. Millions of photos get flagged, but only a tiny fraction lead to actual prosecutions. The system as it exists today isn’t working for platforms, law enforcement, or innocent users caught in the blast radius.
The issue is that when you make ignorance a valid defense, the optimal strategy is to deliberately turn a blind eye, as it reduces your risk exposure. It further gives refuge for those who can convincingly feign ignorance.
We should make tools readily available and user friendly so it is easier for people to detect CSAM that they have unintentionally interacted with. This both shields the innocent from being falsely accused, and makes it easier to stop bad actors as their activities are detected earlier.
No, it should be law enforcement job to determine intent, not a blanket you're guilty. This being Actus Reus is a huge mess that makes it easy to frame people and get in trouble with no guilty act.
Determining intent takes time, is often not possible, and encourages people to specifically avoid the work to check if something needs to be flagged. Not checking is at best negligent. Having everybody check and flag is the sensible option.
Everyone in this case meaning "people demonstrated to be in possession of child porn who took no action". And they are not assumed guilty, they are exactly as innocent as anyone with a dead body in their fridge that they also "had no idea about."
I spent years working in aerospace turbines. This is BS. Power generation turbines are designed to work at ambient sea level conditions. They don't rely on ambient air being especially cold for cooling, they can keep cool thanks to the large mass flow rate.
There is no technological difference between boom's engine and conventional jet turbines. It is still a subsonic turbine, it just happens to sit behind a diffuser that slows the air from supersonic to subsonic speeds. Genuine supersonic turbines are a radically different, and much less efficient, technology. Turbines for supersonic propulsion are actually more temperature sensitive and less efficient than those for subsonic applications specifically because they need to prevent more heating in the compression stages to keep their combustion chambers stable.
The other talking points are likewise bogus. The problem with aeroderivative turbines is maintenance - planes need to be high performance and don't stay up in the air for very long, so their engines are designed around frequent maintenance events. Powerplants, especially those for datacenters, need consistent uptime, not good power to weight ratios.
Boom isn't doing anything special in terms of materials or data monitoring. Yes, power turbines have been a thing for decades, and in those decades they have been arguably the most advanced machines humans have built industrially at any given time. Going back to the maintenance thing, turns out people really want to know if there's an issue before their $200 million machine fails.
I like Boom, I have friends working for Boom. I presume this is just an elaborate way to hop on the AI investment bandwagon. I get it, but it's still ugly to see. I hope this doesn't begin a string of hype-creep that causes their actual goal to fail.
> elaborate way to hop on the AI investment bandwagon ... hype-creep that causes their actual goal to fail.
their current goal might already be "failing" (as in, lack of real demand for hypersonic travel). Investment getting hard to obtain means they're looking for more/broader investment from other investors. Thus, the hopping on of the AI bandwagon.
That's what I was thinking, if they have an engine design, I can imagine (uninformed armchair opinion) it's easier to build a power generator than a plane around it.
Another red-flag to me is the diagram labeling "proprietary superalloys". This implies more than one new metal composition, which by definition is unproven? So a new metal, in a new engine, in a new airplane.
Add to that the "titanium" lable. Nothing is pure titanium. It is all alloys, most of which are "proprietary". Inconsistant technical terminology within a diagram is a huge red flag imho.
>> Legacy turbines need huge quantities of water for cooling to avoid thermal derate in hot environments
Yup. Turbine-powered helicopters famously haul big tanks of water around when flying in hot environments. Heat is a problem, not a game changer.
> I hope this doesn't begin a string of hype-creep that causes their actual goal to fail.
IDK, the first sign to me that Boom weren't likely to succeed was when Rolls Royce parted ways on engine development (1). Were the engines not technically feasible? Not economically feasible? Didn't believe in Boom's business model? We don't know Rolls Royce's thinking, but it's a vote of no confidence.
Taking it in house seemed like a last resort - designing "a new engine in a new airframe" is a known risk. A homebrew engine can't be benefit.
They don't even have the engine: Re-purposing the hypothetical jet engines as hypothetical LLM power plants seems like a nosedive, really.
I'd assume it was because Boom wanted more collaboration and offered less margins than the current customers of DoD super secret planes. Why work harder, longer hours on someone else's idea for less money?
Yeah 100% on the same page. I wrote a detailed comment here but I left out my intuition around supersonic being less efficient. Thanks for weighing in. TBO is a huge factor and the post relies on the readers naivete around that and other factors. But mainly replying because I agree that it's an ugly look to hop onto the bandwagon like this. Especially in aerospace.
I also notice they don't mention the noise profile of these plants. Putting a stationary engine like that will make an insane amount of noise, and even enclosed will change the soundscape for miles around, affecting everything from human well-being to animals' ability to communicate. Not cool.
But maybe they aren't really thinking about that because it is nowhere near being done.
Gas turbine power generation is already a thing. For example 20% of UK grid power right now comes from "combined cycle gas turbines" - which are a very efficient way to turn natural gas into electricity. They've solved the noise issue for those plants.
Um, noise mitigation is still an issue at existing gas plants (e.g., [0-3]) and often only partially dealt with by local regulations, and this group are claiming their engine is substantially different, supersonic vs subsonic, and crossing that threshold usually involves large noise generation effects.
The deeper BS is that there is no engine. People remember the XB-1 demonstrator flying and assume that Boom is farther along than they really are. The XB-1 had off-the-shelf GE engines.
As far as I am aware even China hasn't mass produced commercial jet engines yet so the idea that Boom have one that works and can be mass produced seems highly unlikely. We'll see though I would be interested in London to New York in 3.5h, but I'm guessing the flights will be for the richest people only.
> the idea that Boom have one that works and can be mass produced seems highly unlikely.
Booms claim is they're developing and will some day real soon have "one that works" and then it can be produced at a low-enough cost in sufficient quantity.
Rich people seem to spend a lot of times in helicopters and private planes, which is dramatically more dangerous than commercial air travel.
I could off the top of my head name a few rich people that died from it. Hell, the titan submersible, while a very different animal, is a pretty clear indicator that vast wealth doesn't preclude a willingness to risk one's life in highly experimental "travel"
And they want comfort. A 5-hour flight sleeping on a flat bed is a thousand times better than an economy seat on a 3-hour flight. Part of concorde's problem was the rising expectations of first class travel in the 90s. It was never going to be compatable with today's huge first class seats.
Maybe not the richest of people, but there's a significant amount of people who got their wealth due to their love/acceptance of risk. Climbing Everest is not cheap, is still very risky, and I presume it is more expensive than a cross-Atlantic trip on these jets.
Sure, but the main risk they'd be accepting here is that of spending an inordinate amount of time hanging around in an airport terminal waiting for a broken engine to be fixed.
It'd be hard to spin that as being anything like as heroic as the risk of being killed or maimed whilst climbing Everest!
> Power generation turbines are designed to work at ambient sea level conditions. They don't rely on ambient air being especially cold for cooling, they can keep cool thanks to the large mass flow rate
What could be contributing to this is recently Vertasium did a whole video on how jet engines operate at temperatures above their components melting point.
And how the cold air at altitude is what keeps it from melting.
You're telling us that data centers are more sensitive to downtime than airplanes??? That makes no sense.
All of the aeroderivatives were designed in the 70's before we had computer modeling to help optimize the designs. It's not that crazy to assume that we can design a better and more efficient turbine today with all of the help of modern technology.
Planes are quite sensitive to engine failure during flight. Data centers don’t tend to fly for three hours then sit idle for an hour, then sit idle overnight. They need to be up 24/7. When you’re talking 40 or 50 megawatts, you’re not going to necessarily buy triple or quadruple capacity. So it’d better be reliable without a lot of downtime for checks and maintenance.
> All of the aeroderivatives were designed in the 70's before we had computer modeling to help optimize the designs.
Not even remotely correct. The concept started in the 70s, and designs have been continuously improved, using the latest modelling techniques, for the last 50 years. Modern turbines are some of the most optimized machines humanity has ever produced.
I would think test runs with simulated patients offer plenty of opportunity to experiment.
> Get it wrong, and people die. Best case scenario you're struck off, worst case you're going to prison for murder.
Get it right and people may still die. The whole reason for the improvement effort is that the current practice is excessively risky. No one is getting fired, nonetheless going to prison for trying a sensible improvement to reduce the odds of a child dying which they were approved to attempt.
reply