Rubio is the Secretary of State, he cannot escape being tied to Trump’s foreign policy disasters; foreign policy is the main job of his office.

It became somewhat inevitable once companies started buying mass amounts of Pis for every purpose because they were so cost-efficient.

It used to be an education nonprofit, but it effectively became an industrial SBC supplier.

There are a couple things that cement Steam at the top, imo.

The first is obviously familiarity and completeness - everyone knows Steam and it almost certainly has whatever game you are looking to buy.

The second is added value - very good controller support, mod support, (some) good community spaces, easy multiplayer sessions, proton integration, regular updates, etc. These are genuinely worth using the launcher for.

The third is the single-launcher desire (or the sunk cost effect depending on your point of view). People already have massive libraries on steam, keeping their collection there just makes it easier to browse and find whatever they want to play. Yes, there are other solutions for this, but that is an extra step - there will always be people who don’t take that step.

Familiarity can’t exactly be bought - Epic tried. Completeness can be bought (Epic, Gamepass) but it is hideously expensive. Added value is difficult to get to parity (mod authors aren’t going to instantly port to a new launcher even if there is support). The single launcher effect is a massive first-mover advantage.

I completely agree that Valve having a monopoly on PC game distribution is unhealthy long-term, they also just offer a better experience to me as a consumer than the competition. Even when games have GoG versions, for instance, they are often not updated as frequently and mod support is often more difficult, controller support is hit-or-miss, etc.

In the most charitable light, his statement had merit. Same way as there is a resource allocation issue with cars - the vast majority of them sit around for the vast majority of the time.

If tech companies hadn’t spent the last 30 years aggressively fucking everyone over for every penny they could get, maybe we would be more inclined to take the charitable view rather than see his statements as yet another attempt to add a fence and charge a toll.

While this is true, I will say there are some languages built for very specific problems that make poor tools for general use.

Trying to build a distributed web app in Matlab or Mathematica would either be a flex or a complete waste of time, for instance.

Voter fraud is incredibly rare in the US. The conspiracy theory is that elections are horrifically insecure and the result of every vote is fraudulent.

The proposed voter id laws take that as an inroad to erode the right to vote for all citizens.

And yes, you can train out biases. Rather easily, actually. You don’t even have to adjust weights. You post-train through rules files.

The assertion is not about whether ‘you’ can do it. It is that the model cannot do this itself.

A human recognizing a bias in the model and correcting it is not evidence robot intelligence, it is further evidence for human intelligence.

Training is absolute, it determines the set of relationships the model can represent between input and output. Context, rules files, system prompts, etc. operate only on the conditional probabilities within this set and are impermanent. This is ‘priming’ in human psychology, and why a person in a herpetology class will talk about Pythonic features differently than in a coding class.

As an example of how this is not real learning and is inadequate to solve real problems, here is a paper where the authors challenged multiple models to solve unfamiliar games where they need to discover the rules through play and feedback. Even the ‘best’ model performed pitifully compared to humans.

There’s some semantics here.

Large neural networks have been shown to build parallel models of the world that they learn through their training data. A chess model trained only on chess move notation does have an interior board and state that it tracks, despite never being told about a board, the number of pieces, how they move, etc.

LLMs do something similar with language, mapping tokens to ideas in ways that can produce meaningful and grammatically correct output without needing to be formally taught any of this.

Our brains also do things like this - we observe X and then try to intuit Y, the truth of the world and reason for X.

Where this breaks down is humans can then take our guess for Y and propose Z, and try to find it out in the world. If we can’t, then Y is probably wrong, and our new understanding of the world is Y* that is consistent both with X and the absence of Z.

I cannot stress enough how far we are from machines that can do this. They can learn to approximate Y through training, but they have no means to extend that to propose an unobserved Z, no means to update their internal state given the results of that observation (without fully retraining the model on new data - which makes it a new model), and no way to question X given Z - they can talk at great length about data biases but can literally never excise bad data from their training weights. There have been experiments with training models on time-boxed datasets (e.g. stuff up to AD 1880) and asking it about modern concepts - predictably hilarious results.

All the current ‘AI’ stuff is smoke and mirrors around this very hard problem. The machines can talk, but they can’t think (Z) and can’t learn (Y*).

The thing to lose is human expertise in finding vulnerabilities, which is needed to feed the machines. It seems likely fewer people will go into the field if it is viewed as ‘solved’ by a robot.

Doing something yourself always teaches you better than watching it, or just being told the answer.

A neat result from pedagogical studies is that being given the worked solutions to new problems to study from is much more effective for learning the material than simply practicing.

Students who were assigned practice problems performed worse on later tests than those given the same problems and all of the worked-out solutions for them. The explanation from the research was - when learning a new topic, immediate practice isn’t beneficial because the student doesn’t yet know what to practice. If they are lost when trying the problem, they can easily end up practicing the wrong skills by not knowing where to start. Those with the solutions have some of the cognitive burden removed, and are better able to internalize the task and generalize it to later problems.

I completely agree with you on LLMs and I think they’ll be a disaster for education. But I’ve always found it interesting that being given the answer can be useful for learning in certain circumstances.

50B is annualized revenue, accrued revenue will be almost certainly be quite a bit lower this year.

Tech companies love reporting ARR because it is always bigger than GAAP revenue (if the business is growing at all).

How the robot proves these statement is by writing LEAN code that represents the math. If the lean compiles, the math was valid. If it doesn’t, it tries again.

It will do this forever until it finds a solution.

Goedel predicted this in 1933, btw. From his work on formal systems, he ascertained that since every mathematical statement is trivially true following from the rigorous application of the axioms of mathematics or entirely false, theorems could be checked and proven automatically.

These systems are able to prove statements at a scale beyond humans, because they have effectively infinite time to exhaustively explore applying every rule.

Unfortunately, this isn’t all that helpful for the various fields of math as a whole. Effectively they have a list of statements and a second column with ‘true’ or ‘false’, but the actual route taken by the models to get that answer is often incredibly indirect and therefore difficult to gain insight from and apply to new areas. The approach taken to produce an answer is often more valuable to the field than the answer itself.

The field moves forward when the people in it understand the new results and techniques sufficiently to begin asking new questions, simply giving the answer isn’t all that useful. The robots cannot (yet) do this. Someone has to, at a minimum, pose the question in formal mathematical language. There are an infinite number of true statements in mathematics ( trivial example 1=1, 2=2, …) asking the interesting questions is the important part. The field will still only advance as fast as the people in it are able to digest results.

Further, results in pure math do not generalize to the more applied fields that directly affect our ability to do things. Most engineering formulae are NOT rigorously true - they are weak approximations with envelopes of applicability. The same is true even in base physics. So while a robot can crank through ten million lines of lean to prove an obscure theorem, it cannot think about physics for a week and come up with a theory of quantum gravity.

When peasants were able to copy ‘intellectual property’ for personal benefit and enjoyment, it was a massive crime. When the oligarchs can steal - everything - for the slight chance of paying less in wages, it is an innovation.

I think it is pretty much certain the model companies escape any penalty for their mass theft of IP. There will be some settlements and token penalties that sound big on paper, but are nothing compared to the market valuations.