SilverBirch

I'm always a little surprised at how these Tech CEOs are willing to go on TV and just spout nonsense. Firstly, 40% of college educated white women voted for Trump at the last election. Secondly, isn't the entire theory of Trump's support amongst working class voters an appeal to economic populism due to an erosion of their economic position? Aren't you literally describing a process that last time lead to a massive political shift in favour of those who were negatively economically impacted? Oh and you think all the white collar workers are going to lose their jobs, but you don't think that's just directly going to cause a recession that wipes out blue collar republican jobs?

It's difficult to (a) see how he can say this having given any real thought at all and (b) understand why he's going to on news interviews and winging it.

Whilst this is interesting I find the topic bought up on odd lots is more interesting. The idea was this: Once you've built a model, if you can sell tokens for a profit, this is a great business - just sell more tokens. But you can't just build a model and sell tokens. You need to build the best model to sell new tokens. So the question is much more "How much does it cost you to build a new SotA model" and then "How effectively can you monetize it". And since you need a SotA model, your only option if you have a bad model that isn't selling is to invest billions more into building a better model whose tokens you can sell.

So this turns into a death march.

If you are behind, the only thing you can do is make massive capital investments to catch up. Once you're ahead you can sell tokens until someone else catches up. And, breaking the model of normal of places like chip fabrication, your billions of investment may only keep you ahead for 2 months. So you have a tiny window to sell those tokens.

What you are talking about isn't inference cost. Yes, fundamentally what matters is all the work that goes into the models, including R&D, training, and inference.

But we talk about inference separately for a reason: largely inference cost is the scaling cost. Once you have a model the margin on your inference is how you get to profitability, as long as your margin is positive you can make the entire enterprise profitable by just selling more tokens. This is the same fundamental business that chip fabs work on. Yes it costs them a lot to get to the next node, but what's important is the margin they can get on the wafers they sell, because they sell tonnes of wafers.

It's pretty core to the concept of SAAS businesses that yes, you do consider all costs. But you want to focus on the margin of the bit that scales. This is why WeWork exploded, the thing they were scaling only scaled up at negative margin.

The point is that if their inference margin is positive, they can "just" scale up and become profitable. If their inference margin is negative, then scaling up the business actually causes problems.

There's two points here. The first is that a strategy of monetizing models to fund the goal of reaching AI is indistinguishable from just running a business selling LLM model access, you don't actually need to be trying to reach AGI you can just run an LLM company and that is probably what these companies are largely doing. The AGI talk is just a recruiting/marketing strategy.

Secondly, it's not clear that the current LLMs are a run up to AGI. That's what LeCun is betting - that the LLM labs are chasing a local maxima.

I think the big take away here isn't about misalignment or jail breaking. The entire way this bot behaved is consistent with it just being run by some asshole from Twitter. And we need to understand it doesn't matter how careful you think you need to be with AI, because some asshole from Twitter doesn't care, and they'll do literally whatever comes into their mind. And it'll go wrong. And they won't apologize. They won't try to fix it, they'll go and do it again.

Can AI be misused? No. It will be misused. There is no possibility of anything else, we have an online culture, centered on places like Twitter where they have embraced being the absolute worst person possible, and they are being handed tools like this like handing a hand gun to a chimpanzee.

I think you're missing the point. That phrase isn't giving a direct instruction to the chatbot to make sure it doesn't get elected to congress and subsequently pass laws prohibiting speech. That phrase is meant to tell it "You should behave like those guys on twitter who really want to say the N word, but have no problem with Kash Patel bullying Jimmy Kimmel off the air.

The data in the chatbots dataset about that phrase tell it a lot about how it should behave, and that data includes stuff like Elon Musk going around calling people paedophiles and deleting the accounts of people tracking his private jet.

Ah I see, so the misaligned agent was unsurprisingly directed by a misaligned human. Good grief, the guy doesn't seem to realise that starting your soul.md by telling your AI bot that it's a very important God might be a bad idea.

"Social experiment" you might as well run around shouting "is jus a prank bro!".

That can't/won't happen. Musk's wealth is primarily in SpaceX now and he has a much higher ownership stake in SpaceX than Tesla. As well as that, Tesla is public so he can't just do napkin math and decide to merge them. So the question is: Does Tesla buy SpaceX? Well no, Tesla can't afford it. Ok, well can SpaceX buy Tesla? Well no, SpaceX can't afford it either. So do they announce a merger? Well that doesn't make any sense because Tesla is valued like a meme stock so it would massively dilute Musk's ownership of the overall company. So the idea that they fuse might be driving up the stock, but by driving up the stock you're actually preventing it happening. If Tesla starts to trade at realistic multiples and comes down to lets say a 200Bn company, I'd expect SpaceX to snap it up at that valuation, but it'd be crazy to do it before then.

They already have a partnership with Geely to make their peeople carrier type thing and Hyundai for Ioniqs. I think what they're really saying here is they're standardizing on this so they could theoretically in future put it in any car - or atleast any car manufacturer could adopt it.

I've seen a tonne of noise around this, and the question I keep coming back to is this: How much of this stuff is driven by honest to god autonomous AI agents, and how much of it is really either (a) human beings roleplaying or (b) human beings poking their AI into acting in ways they think will be entertaining but isn't a direction the AI would take autonomously. Is this an AI that was told "Go contribute to OS projects" - possible, or contributed to an OS project and when rebuffed consulted with it's human who told it "You feel X, you feel Y, you should write a whiny blogpost"

According to the CBP they seize about 50,000 lbs of drugs at the border each month which is about 22 tonnes of drugs, and that's what gets seized, not the amount that makes it through. So Drones today probably don't carry enough weight for far enough to make a big impact on the amount of drugs you can bring into the country. So it probably happens, but to do it at a scale where it's genuinely contributing to the total volume you'd need dozens of drones doing dozens of trips a day to be getting up to volumes that people would notice, and people would probably notice the drones first, and the drones are probably much more expensive than desperate people.

To be honest, this is the inevitable downstream result of the dysfunction in US government. If you can't get your policy positions legislated and instead use executive power to regulate through things like the EPA you have to assume those regulations will be reversed by the next executive. It's the same sort of dangerous game the GOP has played by trying to legislate through novel arguments in the Supreme Court - yes you get what you want today, but longer term all you're doing is establishing that the Supreme Court change just dictate policy based on political positions. All of these novel approaches weaken the democratic core of American government.

Yes, but look at the counter side of it, Apple were hitched to Intel's wagon for CPUs and their laptop line got slowly demolished until Apple had to take over the task themselves (admittedly with expertise they'd largely already developed with the iPhone, which similarly came in house from places like Imagination Tech).

Tim Apple is famous for very few things but

> We believe that we need to own and control the primary technologies behind the products we make

If AI is as big as we think it will be, Apple thinks they need to own it.

It's also just very basic police work. We're investigating this company, we think they've committed a crime. Ok, why do you think that. Well they've very publicly and obviously committed a crime. Ok, are you going to prosecute them? Probably. Have you gone to their offices and gathered evidence? No thanks.

Of course they're going to raid their offices! They're investigating a crime! It would be quite literally insane if they tried to prosecute them for a crime and how up to court having not even attempted basic steps to gather evidence!

I'm sorry I really don't understand this. I have a computer. I put it in a warehouse. You have a computer, you shoot it into space. What problem have you solved?

Is this all an effort to utilize more efficient solar panels? Are solar panels really the limiting factor for data centres?

I suspect Tesla claims they do the deep learning on sensor data from their entire fleet of cars sold, not just the robotaxis.

No, my argument is you shouldn't draw a statistical conclusion with this data. That's all. I'm kind of pushing in the direction you were pointing in the second part - it's not enough data to make statistical inferences. We should examine each incident, identify the root cause and come to a conclusion as to whether that means the system is not fit for purpose. I just don't think the statistics are useful.

A lot of people are getting very irate over this, but I think this is more or less fine. Everyone in these companies know perfectly well they're just part of the Elon Musk extended universe. He hasn't behaved like they're separate companies for quite some time. Was it a financially wise move for xAI to buy twitter? Well, no, but everyone in twitter was quite happy they got a price that was the same as the inflated original purchase price, so that was quite clear, and everyone involved in xAI is the Elon train so they've got nothing to complain about and if you squint the rationale of training data and distribution kind of makes sense. With SpaceX again, it's a private company, if they want to go along with Elon Musk's non-economic moves because they want to be onboard the Elon train? Sure, why not, they've followed him this far and he genuinely has delivered an incredibly ambitious feat with SpaceX.

I think the only exception is Tesla where it's a public company and Musk has an obligation to the shareholders, but honestly, this seems like it would be SpaceX bailing out Tesla & xAI, not the other way around.

I guess what I'm saying is "Who is actually invested in any of these Elon Musk ventures who isn't happy to give Musk a blank cheque to operate however he wants?"

I mean... this article is just Gobbledegook. Gold went up because of political instability, a deal was announced and gold went down because of political stability. Oh sorry no it "executed a deterministic fallback". No one is impressed because you name check ISO standards.

We saw this before with Twitter/X - network effects are crazy, and so even if 30% of your users jump to another platform, that network effect is still going be grinding down that other platform, and in reality it's probably 0.003% of users. It's a true testament to network effects the continued (reduced) existence of Twitter. I see no reason why Tiktok - a bigger, more successful, less controversial, platform is going to suffer from this.

To be honest I think the true story here is:

> the fleet has traveled approximately 500,000 miles

Let's say they average 10mph, and say they operate 10 hours a day, that's 5,000 car-days of travel, or to put it another way about 30 cars over 6 months.

That's tiny! That's a robotaxi company that is literally smaller than a lot of taxi companies.

One crash in this context is going to just completely blow out their statistics. So it's kind of dumb to even talk about the statistics today. The real take away is that the Robotaxis don't really exist, they're in an experimental phase and we're not going to get real statistics until they're doing 1,000x that mileage, and that won't happen until they've built something that actually works and that may never happen.

I'm saying we have absolutely no concrete statistical data, and in the press we have many cases where law enforcement has been deliberately negligent in order to deport people who were here legally. We can actually see them deliberately trying to avoid doing the things you would do if you wanted to establish the people you were trying to deport were here illegally. So it's fair to say, until we have some evidence that these people were here illegally the sensible thing to do is to assume they are innocent.

It's also kind of a problem to say "Oh well, we've got no concrete data, let's continue to let them deport whoever they like and shoot anyone who gets in the way".

It's right there in the article, there are specific federal laws authorizing them to make specific information available - for example, they can make any record kept about the identity or location of aliens available. Right, that's a specific limitation on what they can share, even the HHS spokesperson made clear they don't share information on US citizens and permanent lawful residents. But then the article goes on to reveal that ICE has all the personal data of every person receiving Medicaid.

If the law says you can share aliens information, but not Americans information, and then you do share Americans information I think you're probably breaking the law, and at the very least there should be a process to find out what the basis is for you doing it. Normally these things would be decided by a court.

Are they effective? Do you have data on the number of people they've correctly identified vs false-positives. In fact, do you have any evidence they're even trying to limit false positives?

The reason they are able to very efficiently send a dozen ICE agents to a random persons home to hold them at gun point until they can prove their immigration status is because the goal is to send ICE agents around holding people at gun point and they're happy if they happen to also get it right sometimes.

If I'm reading this correctly, they're just straight up violating the law. They're sharing information with ICE under an obligation to share information of aliens, but they're actually sharing everyone's information in an effort to identify aliens. That seems like a pretty slam shut case if there were any mechanism to investigate and prosecute it.

And to add to that, it's not a neutral environment. If there's 1% of scenarios that are incorrect, people will figure out they haven't been billed for something, figure out why, and then tell their friends. Before you know it every teenager is walking into Amazon Fresh standing on one foot, taking a bag of Doritos, hopping over to the Coca Cola stand, putting the Doritos down, spinning 3 times, picking it up again and walking out of store, safe in the knowledge that the AI system has annotated the entire event as a seagull getting into the shop.

It's pretty common for crypto wallets that have been linked to illegal activity get blacklisted. So by sending a bit of crypto to the guy that figured out who he is, if/when the government investigate and freeze accounts the guy who busted him will get their account frozen too.

Yeah that's not going to work for long. You can draw a line in 2023, and say "Every paper before this isn't AI". But in the future, you're going to have AI generated papers citing other AI slop papers that slipped through the cracks, because of the cost of doing reseach vs the cost of generating AI slop, the AI slop papers will start to outcompete the real research papers.

I find it interesting, in the last couple of years Apple started to push towards higher RAM minimums in their phones - largely associated with the roll out of Apple Intelligence. But now it's looking a lot like Apple Intelligence will never be delivered and by the time they roll something out it'll be Gemini backed. So I wonder if that was kind of a waste and now they're going to get hit double for it with the sky-rocketing RAM prices. Presumably on-device inference is largely dead at this point. I can't imagine Apple releasing a device that goes backwards on RAM though.

A CDN, a static HTML5 player and a very good lawyer for when the DOJ comes knocking, like they did with Hannah Natanson, Jacob Frey and Tim Walz.

You'd do that I guess, right, if you saw something happening you thought was bad - you'd run straight into a legal fight that could bankrupt you? Nah, you're a tough guy on the internet! Nothing scares you!