[giancarlostoro]

In terms of SpaceX (the space portion of it) they've produced the cheapest way to get any payload into space. If you pay anybody else, you will overpay drastically depending on who you want to take your payload into space.

In terms of AI, we've seen even here on HN everything from mathematical problems that remaind unsolved, being solved, mathematical proofs being used to disprove theories, heck we even learned more about alzheimers, new antibiotics, precision targeting in oncology, using AI to flag healthcare anomalies in imaging. The benefits are easy to miss, but they're snowballing into place, there's definitely an explosion of useless crap, but you have to look for the real things and you will come to find, that AI is giving us things we otherwise either might not have discovered or wouldn't have within our lifetimes.

[zwarag]

I have yet to see an application outside of harnesses and LLMs itself where adaptation has happened on a larger scale. Devs are fine with babysitting their LLMs. People like to use LLMs to improve their mails and so on. But outside of that, the adaptation is not there yet.

Don't get me wrong. I love LLMs and use them myself. But the biggest gain for me is easier context switch and text manipulation. It's not the: replace X with a bunch of LLMs every CEO is dreaming of. So yes, you have higher productivity, but is the eval of those companies legit? x doubt.

[ijidak]

Markets value future cash flows, not today's cash flows.

By the time you see the applications, the market will have moved on to value the next set of future cash flows.

If the market only valued the obvious, investors would jump in to buy the price up, until it met the average expectations.

The market might be wrong, but the question is not: "Have you yet to see?", but rather, "What do you see in the next three to five years?"

Otherwise, how could investors ever invest in a startup?

Startups never have revenues to justify their initial valuations.

It's a bet on the future.

Investors are future looking.

Consumers are present looking.

We didn't see LLM harnesses coming even two years ago. Now they generate billions per month.

Investors can't wait until reality materializes to make their estimations of the future.

That's why investing is hard.

You have to try to predict the future.

[mullingitover]

> Markets value future cash flows, not today's cash flows.

Sir, this is a casino/Keynesian Beauty Contest.

Markets value what the market participants think the other participants will value. On occasion, this intersects with reality.

[giancarlostoro]

I just gave you several non-dev examples of how AI is used or has achieved things we could only dream about. Not to mention how Qualia is using it within the Real Estate sector.

https://cloud.google.com/customers/qualia

[joquarky]

> Real Estate

Now there's an honest sector that is sure to use these tools to benefit humanity!

[tristanj]

One third of all software code is written by AI. At the frontier AI labs it's 80%+. It has completely upended the software industry. How is that not a massive adaption?

[davemp]

> One third of all software code is written by AI.

I find it interesting that using lines of code as a metric is making a comeback.

[joquarky]

We are devolving due to cronyism. The gold rush has replaced the nerds with brogrammers.

[tristanj]

The number of lines of code doesn't matter any more. A better metric is AI-assisted commits, which has the same statistics.

[rustystump]

I mean if there are more bugs there are more commits no?

[camillomiller]

You couldn’t have picked a better argument to show how this bias is exactly what’s making tech people think this shift is ubiquitous. “It works extremely well for coding, in which I am a domain expert, so why wouldn’t it work for all the other domains I absolutely know nothing about?”

[tristanj]

ChatGPT has 700 million users. What do you think they are using it for? It's upended entire industries.

How many millions of emails do you think are composed using ChatGPT? How many legal briefs were reviewed by AI? How many businesses use AI generated art? How many kids do their homework using ChatGPT?

The GP is arguing that AI has struggled to replace humans, but in so many roles AI is doing the heavy lifting and humans are copying its output.

[FlyingSnake]

Which “entire industries” has ChatGPT upended?

[tristanj]

The entire education industry, so many students ask ChatGPT to help with their homework, and lots pass it off as their own work. Many students have quit going to office hours & tutoring, and ask AI for personalized answers instead. The education industry has no answer, and is struggling to deal with this.

The homework "help" industry (i.e. paying for answers) is dead. Chegg stock fell 99% because of ChatGPT.

Stock photography is rapidly dying, nobody will pay for shutterstock when ChatGPT can generate a passable image for free.

ChatGPT is killing studio photography, it can generate great looking studio photos for free.

Same with basic graphic design / custom art commissions.

SEO / copywriting has been almost fully replaced by AI. Companies no longer pay writers to churn out SEO slop, and now the web is full of AI generated SEO spam.

Customer service as a job is dying and is rapidly being replaced by AI chatbots.

I can go on, but these are the major ones.

[soperj]

It doesn't even work that well for coding.

Otherwise Github wouldn't have 14% down time in the last 3 months.

[jwpapi]

Because it doesn’t work without me

[hypendev]

What is a larger scaler for you? What is "outside harness an LLM"?

What is _the proof_ if all the proofs are not _proofs_?

I don't babysit my LLM based services which are used by coaches and clients around the world. One of my LLM based solution get 30-4k daily hits and I have users coming back on the regular to use it. without babysitting, doing things that would take them hours of manual work and research.

I don't babysit the developers I work with and our clients, which both use LLM's themselves and at scale with their clients, serving all kinds of LLM powered services to millions of users worldwide.

You are not "seeing" the large adoption because:

- The technology is "a few years old" in its usable state - The corporate adoption cycle is slow - You have to understand the technology to use it in a good way, which most corporate devs and PM's do not

So it will take a bit for the "obvious" adaptation on large scale.

But you won't "know" when the large adoption happens.

Silent inference is growing every day, and that is what real adoption looks like - not an LLM being in your face chatbox, but running in the background, sorting, finding, fixing things, aligning data, figuring out analytics, tuning the ads, cleaning the datasets.

[ozgrakkurt]

> In terms of AI, we've seen even here on HN everything from mathematical problems that remaind unsolved, being solved, mathematical proofs being used to disprove theories, heck we even learned more about alzheimers

What a story this is

[CookieCrisp]

What part is wrong?

[rockskon]

Isn't AI routinely making significant mistakes in analyzing medical imaging?

[stingraycharles]

My understanding is that it’s better than doctors themselves. But it’s probably the same as with autonomous driving: the bar isn’t just “be as good as humans”, it’s “be flawless”.

[haldujai]

It’s actually quite a lot worse than even doctors in training except for highly constrained experimental settings and a few very nice applications that are mostly too tedious/impractical for a human to do or are very basic detection tasks.

I am a radiologist and researcher predominately focused on AI.

[yread]

I work with pathologists and radiology is way ahead of us with AI use in clinical setting (but still not very far). Only things that get serious use are lab-developed (ie not commercial) image analysis algorithms for very limited (tedious, error-prone and ultimately not that often used) biomarkers. Don't believe the hype.

You could also look at the market, one of the biggest players, Paige, was acquired for about 30% of the money they raised.

[marcus_holmes]

Thanks for the informed take :)

Do you think this will result in more routine/boring/tedious tests? Is the bottleneck on these things the human time to analyse them?

[haldujai]

I don’t think so, not beyond the current trend in medicine which is going up anyway.

For some things, like 3D volume segmentation of structure or disease (e.g. CVA/stroke volume, cardiac muscle mass, iron quantification) the bottleneck is the time it takes so we currently use approximations like single longest dimension, circular regions of interest, etc. AI will dramatically increase accuracy allowing for more accurate treatment and easier large scale research with quantitative endpoints.

Other things people think of like detection of aneurysms, fracture, lung nodules are not “hard” but AI has already added and will continue to add the second-reader benefit which will reduce detection errors. For this category the clinical benefit is as of yet unclear and we know that increased detection does not necessarily translate into improved patient outcomes and can in fact make them worse from over-diagnosis which means investigation related harms and over-treatment.

We were already in a phase of “over detection” in much of radiology with advances in imaging technology so the incremental benefit of current AI remains to be seen and I personally think is going to be much more limited. I had a case recently where a 2 mm brain aneurysm was missed on 3 CT scans over 10 years but was picked up by AI so now is being followed annually. This is too small to treat considering the risks and a serious argument could be made that 10 years of stability is proof enough that this is almost certainly clinically irrelevant for this patient.

Far more interesting areas of AI in imaging are in acquisition of acceleration (i.e. the medical equivalent of upscaling) which can dramatically decrease costs and increase accessibility as well as analyzing imperceptible features.

It may not be a popular take here but in my opinion the future of radiology is like what we see in software engineering today - a skilled human equipped with AI will outperform humans without AI and AI without humans, the latter of which we are still several years away from prototyping due to various technical hurdles.

[marcus_holmes]

Thanks :) Very interesting.

> in my opinion the future of radiology is like what we see in software engineering today - a skilled human equipped with AI will outperform humans without AI and AI without humans

I suspect this will be the case across the board. It's a useful tool, but it's just a tool. It's not a replacement.

[nesk_]

A friend of mine, a dermatologist, told me that LLMs are quite performant for melanoma analysis. Based on their own statistics, LLMs are able to beat humans with ~10 years of experience in the field.

They will never beat the human instinct tho, but they can be great tools sometimes. Unfortunately, LLMs mostly produce garbage.

[haldujai]

Whenever it comes to medical diagnosis I would caution anyone to be careful with what “beat humans” really means.

In real life pathology is a spectrum not a binary and physicians are not trained to be 100% accurate instead optimizing sensitivity and specificity considering pretest probability as well as the harms of overdiagnosis and under diagnosis for a given scenario.

For something like melanoma which is relatively easy to diagnose with a superficial, extremely low risk skin biopsy and where early staging dramatically improves outcomes you would want to design around overcalling (high sensitivity) rather than maximize accuracy given the significant harms with false negatives and minimal harms with false positives.

An AI may be more accurate at classifying melanoma/not melanoma but if it does not meaningfully improve on the clinical threshold of biopsy/no biopsy or result in less biopsies that accuracy is wasted and may even be detrimental.

Note: I am just using this as an example to illustrate the considerations.

[not_that_d]

I don't think your friend understands Large LANGUAGE models.

[stingraycharles]

You do realize that today’s multi-modal LLMs are actually able to understand images? They’re tokenized very much like language is.

[the_sleaze_]

First I'm hearing that is a lot worse - do you have sources? Genuinely curious

[haldujai]

The better question is are there any sources that AI is better than human readers? I haven’t heard anyone make this claim outside of single/few disease classification tasks and even those are mostly 2D.

Anecdotally, my practice has most FDA approved AI deployed as we are an evaluation site and very rarely is the AI result useful. Over the past few months we have been cancelling contracts as these cost quite a lot of money (in some cases eating >50% of the study interpretation cost) for little to no benefit and a LOT of noise.

[stalfie]

Last time I checked thoroughly (roughly two years ago), AI (in the form of small ML models) mostly outperformed radiologists in areas where the gold standard is "one level" above imagining wise. By that I mean that you train a model to detect on an X-ray what would normally need a CT. Or train it to see on a non-contrast CT what would normally need contrast or an MRI, or biopsy, and so on.

Essentially the cutting edge reaches up to 99% of human performance on the task it is trained, which is good enough for triage but not for a final diagnosis. However, magic sometimes happens when you train a model to detect something, which you already know is there, on an examination that is cheaper, faster or less invasive than the human"gold standard". Conveniently, this dataset exists since it's common to first do a cheap examination like an X-ray, and then escalate if nothing is found (or if something is found that you want to see better, or a number of other possibilities).

Examples of AI outperforming humans like this includes AI detecting sacral fractures on x-rays better than radiologists (who normally take a CT to conclusively exclude it), detecting potential precursors to pancreatic cancer on non-contrast CTs (where contrast or an MRI is usually required) and detecting an occluded coronary artery on an ECG without the archetypical "ST-elevation changes".

See the link below for references: https://pmc.ncbi.nlm.nih.gov/articles/PMC9478257/ https://www.nature.com/articles/s41591-023-02640-w https://rebelem.com/a-winning-hand-in-cardiology/

So AI, as a general rule, doesn't usually match or exceed the upper bound of the "gold standard" medical performance. But it tends to carry the quality of the upper bound downwards towards the faster, less expensive and invasive methods. In some cases, like in the case of EKGs, that's huge. In some cases it saves time, in some cases it decreases miss rates from tired radiologists or triages their review feed. And in some cases it's not very useful.

LLMs doesn't come close to specialized radiology models at the moment, because LLMs are more about applying knowledge than creating new correlations. Of course that's also hugely useful, but that's a bit of a different topic to unpack.

[goda90]

With these kinds of things, I want to see comparisons to trained, alert humans. Cut out all the distracted, stressed, tired, incompetent, intoxicated cases from the baseline. That includes rushed doctors at the end of a long shift.

A self driving car doing better than a drunk on the freeway doesn't reassure me that it'll do better than sober me in a snowstorm.

[rlt]

That would be a fine bar if you could ensure your doctors or nearby drivers aren't distracted, stressed, tired, incompetent, or intoxicated.

[fragmede]

How does sober you in a snowstorm cause the drunk on the freeway not to drive?

[darkwater]

Non sequitur. The core idea is that if you have just self-driving cars you won't be trained enough to drive properly next time you're caught in a blizzard, because you never drove for the last 5 years.

I also question if the kind of person who actually drives while drunk - knowing perfectly by thousand of society inputs and peer pressure that it is wrong - will care enough to buy a self-driving car.

[goda90]

It doesn't, but I'm not going to trust my own safety to a self driving car that can only be said to be better than the worst drivers. It's a bad baseline.

[natebc]

What are the premiums like for this AI radiologists malpractice insurance?

[BobbyTables2]

I’ve seen the same. But I don’t see that as a glowing beacon of progress.

A whole lot of doctors, if not most, didn’t pick their profession out of an interest in medicine…

[koolba]

It’s so good it even sees things that are not there!

[joquarky]

I don't need anything launched into space.

Math problems being solved doesn't make necessities easier to afford.

New healthcare discoveries seem moot when only the wealthy get access (Elysium anyone?).

I'm not seeing how a typical person is getting any benefit. The only visible change they are seeing is getting exposed to more elaborate scams.

[sharts]

Do those help starving or homeless people? I don’t think so.

[kridsdale1]

Those people are not in the economy.

And they can get their diseases cured by an AI discovered fix, via medicare.

[itake]

People want to buy SpaceX, not Twitter, Tesla, xAI. Unfortunately Elon has been conflating the three.

[moduspol]

Plenty of people want to buy Tesla. Not saying they're prescient, but the market cap represents that.

[joquarky]

I remember when the music industry would release albums with one good song and make you buy the crap instead of selling the single at a lower price.

[itake]

and the 2008 crash's mortgage backed securities...

[themafia]

> they've produced the cheapest way

Were we struggling to do this before? Was the overall percentage reduction in costs? Was some other achievement held back because we couldn't accomplish this? What is now enabled?

> to get any payload into space.

A limited set of payloads into space. No vehicle can get "any payload" to space at a fixed price.

> The benefits are easy to miss,

You've listed a bunch of reasons to publish papers. What is the actual ground level change that's occurred? Are those antibiotics produced? Do they actually work just as predicted? Why is that first world problems are exclusively listed but basic problems like world hunger are never even approached?

> or wouldn't have within our lifetimes.

And your life, your actual life, benefits, how?

[JumpCrisscross]

> Were we struggling to do this before?

We literally couldn't.

> Was the overall percentage reduction in costs?

Starship will bill NASA 1/20th what SLS does.

> What is now enabled?

LEO. Artemis. Out of all of these companies, being confused about SpaceX is super weird.

[duttish]

If SpaceX was only Starlink or only Starlink and rockets it would be an horrible circumvention of the rules.

But now he's also trying to get the indexes to pay for the giant cash fire called X.ai and the far right huddle Twitter too.

I have zero interest in owning anything of either of those companies.

[alpha_squared]

Granted, I only skimmed some high-line numbers, but isn't their only profitable project Starlink? SpaceX is functionally a satellite internet company that happens to make rockets.

[JumpCrisscross]

> isn't their only profitable project Starlink? SpaceX is functionally a satellite internet company that happens to make rockets

Yes. The thing that’s going public is almost entirely an AI play.

[HWR_14]

They seems to have decent revenue leasing compute to Anthropic.

[tech_ken]

> Starship will bill NASA 1/20th what SLS does

Is that before or after the program achieves profitability?

[harimau777]

I think you missed the core of their question: What has actually gotten better in practical terms for the average American?

[JumpCrisscross]

> What has actually gotten better in practical terms for the average American?

Starlink has made connectivity cheaper and more available. Earth imaging has made various food production processes more efficient. Weather forecasts have become more accurate.

If you’ve genuinely missed the massive economy that LEO has become, it will be a fun thing to catch up on.

[autoexec]

> Starlink has made connectivity cheaper and more available.

Yeah that's working out great for the average American isn't it (https://natlawreview.com/press-releases/2026-consumer-trust-...)

> Earth imaging has made various food production processes more efficient.

I'm not even going to bother sourcing the fact that food prices have only massively gone up negating any gains in productivity. The average American struggling to buy basics like eggs and meat aren't feasting on more efficient food production.

> Weather forecasts have become more accurate.

I'm sure the growing homeless population is happy to know they can better predict the weather they'll be sleeping in.

This is all totally worth supporting a nazi billionaire

[JumpCrisscross]

> that's working out great for the average American isn't it

Yeah. It did. My neighbour’s rates went up. He switched to Starlink.

> not even going to bother sourcing the fact that food prices have only massively gone up negating

This is like arguing fertilisers are useless because prices went up.

> homeless population

Not super relevant!

> all totally worth supporting a nazi billionaire

Nobody said that. But it doesn’t mean the benefits go away.

[usef-]

Do we apply a bar this high for any other company/job/business? Saving gov/tax money aka "billing NASA 1/20th what SLS does" doesn't count as worth it to you?

Reusing rockets reliably rather than "throwing them away" is a great achievement and I'm surprised people have to justify it on HN

[mandeepj]

> Reusing rockets reliably rather than "throwing them away" is a great achievement and I'm surprised people have to justify it on HN

You can milk a cow only a set number of times!

[usef-]

Yes, because you're not designing the cow. Progress on rocketry (and reusability) is not completed, btw, there's a lot still to improve.

[airstrike]

Stock prices indicate the present value of all future dividends, so it's not about what has happened but about the risk-adjusted expected value of all which is to come.

What probability you assign to arrive at that expected value and how you adjust for risk is on you.

[GPerson]

That proof has almost no significance because the average mathematician cannot spend 2 million dollars on inference.

[cowmix]

You'll overpay -- but not by trillions.

[waterheater]

On one order, correct, but it's still on the order of hundreds of millions to billions.

Also, keep in mind that a stock price discounts expected future cash flows. Is it likely that SpaceX will have a near-peer competitor within a few years? No, it's not, and that market share is being priced-in.

[hattmall]

Is it likely that SpaceX will have actual reasonable demand? Their major customer is Starlink. How legitimately confident are we in the numbers with regard to price reduction vs creative accounting to offload costs to Starlink and subsidize the launches to appear to offer huge cost reductions?

If there exists sufficient demand for the product of space launches then it's probably reasonable to expect their to be a near-peer competitor soon, but that's only if SpaceX were to be profitable, which it isn't, even with the subsidization by Starlink on the order of many billions.

[hvb2]

> If there exists sufficient demand for the product of space launches then it's probably reasonable to expect their to be a near-peer competitor soon

Space is not that easy. Even with unlimited money, it'll probably take 10 years to build a rocket like starship. Going from nothing to orbit needs a lot of money but more money doesn't make that faster.

[m4rtink]

There is about 3 chinese orbitallaunchers with some reusability support flying & about as much scheduled to debut this year.

But other than that, yeah - outside of China, progress has been horrendously slow & Blue Origin, the only other US company that demonstrated a partially reusable rocket just had a devastating pad explosion, destroying one of their 2 rockets and their only launchpad.

[giancarlostoro]

Sure but SpaceX can get you into orbit for $1400 per kilogram, and future projection and goal is $100 per kilogram. The competition is at $15,000 per kilogram. I think it's a no-brainer for anybody trying to get anything into orbit. Unless someone figures out superior tech that surpasses SpaceX, I'm just not seeing why anyone would spend more for less capable and costly rockets.

[Alpha3031]

Doesn't SpaceX charge 2 to 3 times their internal cost to external customers? ISRO is still more expensive, IIRC they charged ~US$60 million (roughly $6000/kg) for the OneWeb launches whereas after the recent price hikes SpaceX is supposedly charging ~US$74 million on a larger rocket (~$4200/kg), but that's far from an order of magnitude difference like your comment suggests, which I assume would be using the $25 million they charge Starlink internally (IIRC ISRO's internal cost is much higher, around $40 to 50 million, but that's still not anywhere near an order of magnitude). Using internal cost from one provider and external price for another is somewhat misleading.

[Planktonne]

> future projection and goal is $100 per kilogram

This can't be treated as meaningful, given other projections and goals (Mars colony, etc.).

[dh2022]

At the rates you quote, $1 T (the size of the market) is 714,285 tons of stuff in the space each year. I don’t think there is enough space in space for that much cargo.

[m4rtink]

Let me introduce you to the 30 km long rotating O'Neill Cylinder: https://en.wikipedia.org/wiki/O%27Neill_cylinder

Although realistically this will be built from lunar materials, you still need to lift a lot of mass to build the necessary industrial processing and mass drivers to launch it from the Moon to some Lagrange point.

And there are many other useful space megastructures that can be built in space from common materials, like giant solar arrays beaming power down via microwaves: https://en.wikipedia.org/wiki/Space-based_solar_power

Most of these proposals date from even 1980s.

[dh2022]

I guess for a trillion dollars vine can built Elysium. Generating solar power in space (vs. on the ground) makes as much sense as running AI inference in data centers in space.

[m4rtink]

Space solar power makes perfect sense once you have sufficient in-space industrial infrastructure available, so that all the actually heavy stuff (structural materials, solar panels, antennas) stuff is manufactured from local (Moon, asteroid) resources & only the most high tech stuff is launched from Earth (electronics, engineers).

The space solar arrays on geostationary orbit would get basically 24/7 sunlight with no weather effects & constant thermal environment. On the ground ad the microwave antenna level you get clean electricity to use for whatever, at any time of the day.

[pzo]

how many packages have you shipped so far to space? SpaceX could disappear tomorrow and most people wouldn't notice. Your satellite TV might get slightly more expensive. Those rare people that don't have LTE internet access and need starlink are exception.

[UniversalBlue]

Sorry but SpaceX has done absulutely nothing for space other than take billions in GOV funding and never delivering what they promised years ago. Hell the only cargo they ever shipped was a banana...

[adammarples]

They've shipped lots of satellites into low earth orbit, that's their starlink business and it's where all of the revenue in spacex comes from, and it is a good business in itself.