Pretty much. Open AI doing an IPO will be the canary. Unfortunately for them it's taking a while as they have to shake off the whole non-profit thing first.

It's like during one of the earlier bitcoin bubble, where the financial services industry was busy trying to get a crypto options market up and running. Like clockwork, once you could short BTC then it had an almighty crash.

There are a lot of people who want to exit Open AI to make a lot of money. They will exit, then make money on the crash, then make money when the Fed predictably starts pouring cheap money into the markets to try to prevent a wider crash.

It's hard to see how this is even fun for them anymore. They have more money than you can spend in a lifetime, and have rigged so much of the financial system that it's like playing a slot machine where you're choosing when it should pay out. Surely these guys are bored out of their minds by now.

I think Starlink just really messed up their roadmap with the satellite stuff. I know lots of people who do backcountry trails, and when the iPhone 14 came out with the satellite connectivity, a lot of them were quite excited. My guess is that after proving out the SoS functionality, they would have extended it to a paid service for messaging and then started upgrading it to handle more and more data. In the end you'd have a phone you could use anywhere for a hefty subscription fee. Combined with the whole rugged line (Ultra watches etc) they would have had another revenue earning service to lock in a lucrative customer group.

Starlink and their ability to do direct to cell basically destroys that model, so the satellite stuff ended up not being a key feature they could push.

Apple is in such a tricky place now. They have to drip feed features into the iPhone because they don't exactly have any block-buster things to add to it. It's a mature technology and people are just buying them on upgrade cycles now - which is absurdly lucrative for them, but won't generate massive growth.

I think Cook has been relatively at peace with this strategy - he's near the end of his tenure, and got to overseas massive growth as the market developed - but whoever takes over is unlikely to want to just drip feed out features to keep sales bouncing along. So it will be interesting to see what they do from a strategic point of view when that happens.

The whole quantum AI is pretty obviously going to be the next step in sustaining the insane AI valuations. I bet Elon and Altman will start a quantum computing company once the market gets wobbly enough.

They will then be able to say the **magical stuff will happen** as soon as they get the quantum machine running - which they're really close to doing. Even being able to do press releases around stringing more and more q-bits together, will allow them to keep the hype going for another five years or so until they find the next thing to push the bubble (probably something like making quantum computers in space).

They use compression ignition engines though (unless there are any gas turbine powered cargo ships...), so it is correct to refer to them as diesel ships.

One of the main motivations for ever larger cargo ships is that the square cube law gives you big savings in fuel efficiency the larger you go. It's why the largest ships are limited by the size of the panama canal - otherwise they'd go even bigger.

But if you don't have fuel costs then this ceases to be a motivation. Sure you get some savings in terms of reduced crew count with a bigger ship vs multiple small ones, but again, automation is dealing with that. There are also costs involved in having big ships, such as not being able to go to as many ports (so require more last mile transport), and the costs of updating the infrastructure they require.

Its entirely possible that a fleet of semi-autonomous sail powered vessels could be competitive with giant fuel powered cargo ships on many routes in the future.

My god, you can't see the wood for the trees. What do you think the point of life is? To contribute towards turning humanity into some beige corporate productivity machine that maximises year on year growth rates? Dude, you need to get outside. Go and enjoy some music, some food and drink. Meet some random people and chew the fat.

Life is for discovering. It's like we have access to the most incredible procedurally generated video game, and given about 70 years to explore it, and all you want to do is wipe out 99% of the participants and have everyone sitting in a windowless grey room with a bunch of people who are EXACTLY the same as themselves to wait things out until they're dead.

Look, MMT really annoys me in its attempt to rebrand Keynesian fiscal stimulus as some kind of magical new discovery, but COVID certainly demonstrated that you can boost an economy in freefall if you just pour enormous amount of fiscal stimulus into it.

I mean, even the Neoliberal prophet Friedman was unequivocal about how the govt can create any amount of inflation if it really really wants too (helicopter money).

Now, whether that's what you should do, or the exact details of how you do it, or what the end goal of doing it, are all very debatable. But make no mistake, if the govt wants to go on an 'MMT' money splurge to stimulate the economy they most certainly can.

Fundamentally, you have to realise that having an economy with stable 0% real interest rates is an enormous privilege - it allows you to extend and pretend forever. All those crazy valuations you can just kick down the road because the holding cost of maintaining your position is zero. Personally I think the deregulated financial system requires either a reset (massive crash) or for this situation, and returning to the moribund ZIRP economy is probably the best of the two worst options. So when the AI bubble burst I think you will see a quick return to ZIRP, QE and then they will be able to raise (through QE) whatever level of debt is required to prop things up.

The post GFC was the new normal, and that is where they will try to steer things.

I appreciate that you seem to really care about your children. But as someone who was homeschooled, what are you going to do when you kids eventually have to interact with the shitshow that is the real world? I for sure do not want my children exposed to the bullies and psychopaths of the world, but on the other hand those people are everywhere in life. You have to learn to deal with them. I got burnt extremely badly early in my career because I assumed everyone was inherently decent, and my ability to judge character was limited.

My children now go to a state school with a very diverse and fundamentally decent community. There are still a broad range of people there. I've taught my kids to be generous and share, and for the most part they have very healthy relationships with other children. But they have also already encountered the kid who just takes takes takes. And the kid that manipulates the group emotionally to try to get all the attention. They are learning how to recognise these types of personalities, and how to integrate them into their lives without allowing them to dominate. I think these are really important skills. Even learning that sometimes you have to follow stupid rules that make no sense I think is an important skill (I spent a long time being frustrated about all the idiotic things around working a job).

In the end though we just have to use our best judgement - I won't know for another decade whether I've made the right choice. Anyway, just some things that you might want to consider.

This makes no sense at all. Even if you don't want them, if Europe wants to continue running its economy on car exports it doesn't take a genius to see that developing your own driverless system might be useful. Further, there are plenty of uses in Europe beyond personal transport (e.g. long distance trucking) where a working self-driving system would be useful. There are loads of trucks criss-crossing Europe. The idea it is a public transport utopia is myopic.

Perhaps even more importantly, autonomous driving is just part of the development of automation stacks in general. The technology pioneered by Waymo etc will be useful for many other automation tasks. Things like SLAM/perception algorithms, simulation training etc would all be really useful for any next gen robotics (not taking stuff like humanoids, just all sorts of task automation). That's a good enough reason to pour money into the field even if you want to keep driving your own cars.

As someone living in the UK, honestly, the trouble is that Europe is so conservative when it comes to investment. To get money you have to find some rich aristocrats, and then plead with them to stop investing in expanding their stupid property empire and give you a bit of money. If you then fail to deliver, or they get cold feet, they will demand their money back and an official inquiry as to why you failed. Then they'll go straight back to stuffing their money into property investments again and tut tut about how risky all this fandangled tech stuff is.

Further there is just this persistent belief outside FAANG that someone who monkeys around on a computer all day couldn't possibly be worth paying banker sort of wages. So the smart tech people go into finance or work for a US company, while the average ones work for local companies for what are frankly insulting wages. This then reenforces the idea among the local companies that engineers aren't very good, so they refuse to pay more, further destroying their ability to get good engineers. If companies tripled engineering salaries here, they would suddenly find a whole lot of very competent people move across from finance.

I would have agreed with you if I hadn't ridden a bicycle around SF a few months back. I mean, sure the roads are wider, which means you just get delivery drivers double parking everywhere, people with gigantic cars everywhere, people driving faster, cars always parked on the side of the road which limits your visibility everywhere. I get you on the pedestrian thing, but I suspect that is because only crazy people try to walk (or ride a bike) around the downtown area. Then you have the street cars, and the tourist street cars, and the tourist in general, and the crazy four way stop thing. Add to that all that the potholes everywhere, faded road markings, and guys with shopping trolleys wandering in the streets (admittedly didn't see that until the mission). It's not this magical driving utopia. Any northern european city (e.g. Copenhagen, Berlin) would be way more suitable than SF.

I was honestly blown away with how the Waymos seemed to just deal with it all. I spent a good half hour just watching them drive through an intersection. They've absolutely nailed it.

Wayve is taking a VERY ambitious approach (end to end neural nets). This sounds great, but as with all probabilistic models, they are going to have a fun time trying to chase those 9's of reliability. Doesn't mean they won't catch up, but there are a lot of players (including Tesla) promising to catch up, but none have managed to do so yet and, frankly, after seeing those things in SF, Waymo is already there.

It's all the same bubble. It's caused by central banks intervening in the credit market to artificially suppress the price of credit. Combine that with a deregulation of credit creation (there is no concept of fractional reserve ratios etc now - credit is limited by the 'discipline' of private banks) and you just have gobs of cheap money going everywhere. When they did the COVID stimulus this just supercharged the process. If you have credit expanding beyond the capacity for real investment to absorb it (new machinery, building etc), it can only inflate asset prices and more than likely a bunch of Ponzi schemes.

If you could just flood the economy with cheap credit at any level to get whatever real economic growth you want then we would just do that and we'd all be living in gold palaces. But it doesn't work like that. Credit can only facilitate economic growth but it has to be part of an overall strategy to direct the broad economy where you want it to go. Our current system is like if someone proposed that the government just build more casinos to grow the economy. Everyone then goes to work either gambling at the casino (to earn money) or working in the casino itself. You have full employment and everyone's busy making money, except when they all go to find a doctor they can't because everyone is working in the casino. The problem is if you leave it to 'the markets' they quickly conclude that it's way easier to setup a casino then train a whole lot of doctors. The whole thing takes over your economy and that's what we have now.

When this current bubble burst all that will happen is that they will go back to ZIRP and QE faster than you can imagine. In fact, if the bubble destroys large parts of the real economy this will make their job a whole lot easier. We are just going to return to the moribund economy of 2012 to 2019 when they kept telling us who things were great while everyone had been slowly slipping backwards for nearly a decade.

Another thing to consider is that we are making rapid progress in terms of treating cancers. We are basically learning how to hack the human immune system to enhance its error correction abilities against genetic damage.

By the time we are regularly sending people to mars, it's not unreasonable that we will just have such good cancer treatments (or treatments for all sorts of genetic damage) that the effects of the radiation can be mitigated therapeutically.

It's like how explorers used to be at risk of scurvy, until science figured out how to mitigate that. Now nobody worries about getting scurvy when travelling on a ship.

I think the biggest concern I have with Starship right now is that the dry weight appears to be very high and they haven't proven rapid reuse. If they had either a good mass fraction, or had a stack with proven rapid reusability (at this point, just not having tiles falling off) then they'd have wiggle room to trade one for the other. But at the moment they don't have either. Of course there will be a lot of optimisations they can make, but countering that they might have to add a lot more weight to get rapid reuse.

The entire thing is such a knife edge. For missions like Starlink I think they will be okay, as even if the stack can only launch a small number of satellites, if it is truly reusable, then that's unlikely to matter vis a vis Falcon 9. But for the moon missions if you start cutting payload fraction it dramatically increases the number of tankers you need to send, which complicates your launch operations (must have a very rapid cadence), complicates the time you have to sit on orbit refuelling and just overall makes the mission much more complicated than originally planned.

My guess is it's very likely they can get something working that is viable for Starlink, but extending this to a platform that can do lunar missions might require a lot more work. Things like moving back to composites for parts of the space craft (realistically, a composite booster would likely be quite a bit lighter, but incredibly expensive and complicated to make compared to what they have) are all options but would add huge amounts of time to the project.

I mean, this is a moonshot, so those are exactly the sorts of questions they will be looking to answer. More than likely the answer will be the rather obvious 'not easily' but I guess if you have billions of dollars and can make a few extra billion on your stock price by doing press-releases like this, then it's not an entirely stupid idea to get some grads to look into it and write a report.

The other thing is that having some sort of edge compute in space is probably going to be a thing in the next few decades, probably through systems like Starlink. You can imagine at some point having sections of your CDN on the satellites might make sense for when they're out over the middle of the Pacific and you don't want to set up a bunch of relay stations out there. I'm sure there is also a lot of military interest in reducing the cost of on satellites processing through commercial avenues.

These sorts of studies are really about identifying the issues that need to be addressed and then they will be able to either invest some money into dealing with them, or monitor the landscape for technologies that will fix them. I don't think it's an unrealistic thing to consider at all for Google. But yeah, the fact they have to tie everything to AI is getting pretty tiresome.

Well, to be fair, all these AI companies are currently running on the old adage 'we lose money on every sale, but make it up in volume'.

It doesn't matter how much he spends on infrastructure, at some point he has to charge people more or cut his costs to run these models.