What makes another couple big steps like that inevitable in a short time frame?
Before the recent floodgates cracked open, AI research made only slow incremental progress for decades. Why couldn't we already be back near that rate of progress?
Funny, I put parachutes on my airplanes in kerbal space program (as a safety feature) but never considered what the real-life analogue to that would be. Turns out it's very similar!
Fair point. Humans experience reality and use words to reflect that. LLMs only have the words. And it's an open question how much of a limitation that is to understanding.
I like this anecdote because it gets at how words to an LLM have no connection to their real concepts. To an LLM, words are simply numbers arranged in a likely pattern.
I think one could make a hypothetical case that working at a cigarette company in 2026 is a more morally justifiable position than working at meta in 2026, because cigarettes as they're regulated today generally only affect adults, only affects those who opt in, and kills people on average near the end of their lives.
Meta's decisions affect everyone, even non-users, because of their outsized impact on society. But they also have way more users than cigarettes, and deliberately prey on children and teens in ways that could affect them their whole lives (see recent lawsuit).
I would struggle to judge anyone working for either of these companies. I think the blame lies at the top, or is shared by all of us for failing to build a better society which prevents such exploitation.
Measuring damage to society, or the degree of moral bankruptcy in a company's leadership, is a very difficult thing to quantify.
So, I agree with you that these are personal choices, and everyone will draw the line differently based on who they're comfortable working for and what they're comfortable contributing to.
Flipping the question like this was my instinct as well. I think the passion tax and supply and demand are definitely factors, but the real reason for the gap is that big tech has gotten increasingly exploitative and monopolistic, leading to larger margins and higher competition for talent.
I like classes for meeting new people for a few reasons. First, there's a structure around the activity which doesn't force you interact all the time. Second, you're not entering an established group as the one new person. And third, you're all going through a shared experience which naturally helps form bonds.
If you're really hard up for social skills, and you like being around silly fun people and are okay acting silly yourself, I recommend taking an improv class. Any introductory improv class is basically kindergarten games to help you realize how low stakes socializing can be.
But I also realize improv isn't for everyone. In that case I recommend finding an activity you might be interested in and taking a class in it. If you aren't sure what you're interested in, good news! That means you get to take all the classes until you find one you like.
Even if there's no classes for your activity, anything that can be done out of the house and with other people probably has a community built around it. Use your interest in the activity as leverage to expose yourself to that community.
The highly random and somewhat convoluted mechanisms remind me of something like character creation in the rpg classic traveller.
You do have a little choice there (unlike here), but I think the same appeal is present even without the choice.
The experience is a bit like reading fortunes from tea leaves.
The fun comes from assigning meaning to the outcomes. This happens, generally, automatically as a human instinct. In traveller, the process of character creation generates a kind of narrative in your head of who the character is.
I've been thinking about these kind of experiences a lot lately.
Is it a game? I don't think a discussion of definitions is very interesting, but I would call it a game by any casual meaning of the word. Certainty, in traveller's case, a roleplaying game. But I recognize the same appeal in these zero-player games.
You're playing something, just in your own mind. The primary game isn't what's on the table, but what's in your mind. You're not committing to choices in physical (or digital) space, you have no agency there -- but there's still a rich experience happening between your ears -- full of hopes, predictions, disappointments, elation, creativity. It's like reading but you're also part author. You're not reading from a book, but from pattern matching inside randomness.
I'm still trying to wrap my head around it myself. I have little more than the recognition of something interesting in this direction, but beyond that I can't articulate it.
I was reflecting the other day how discovering things online felt like being in on a secret. You had to just know about a chat room or BBS or website. Each one was like discovering a secret.
Now it's the opposite, anything special posted online will quickly get overrun. It's the parties and places not posted about online that feel like you're discovering hidden gems.
I remember before LLMs, someone on HN made a bot to program automatically by pulling the top rated answers from stackoverflow. To me agentic coding just feels like the next iteration of this.
And LLMs in general feel like an iteration on search.
The strengths and weakness of LLMs are already apparent, and in my opinion unlikely to change from here.
Not sure what you mean, can you help me understand?
Do you mean if free float goes from 5% to 100%, and weighting goes from 25% to 100%, it's more "extra supply" than "extra demand"? That's a good point I hadn't considered. I'm not sure the details of the lock-up period though, it might be staggered. So if it goes from 5% float to 50% float, that's 45% of additional shares available to buy, but an addition 75% of the market cap that indices now need to weight to. But it's true this would only happen once (when free float goes from below 20% to above). Then after that the extra supply would be more than the extra demand. Or do I misunderstand?
I'm not so sure the silent majority is positive on AI, I think the opposite is more likely. Let's not forget that national poll where it was less popular than ICE -- I think it was 26% positive vs 46% negative.
My view is AI is becoming a poster-child for the increasing wealth disparity. When people are negative on AI it's not just the technology but the entire idea around it. It's simply cool to hate AI and that's going to be a hard hill to overcome, I think.
That's true for the S&P but not nasdaq, nasdaq is market cap weighted. There used to be a limit that the available float couldn't go below something like 20%. (This is because 5% float available but 100% market cap would cause a huge supply/demand mismatch). But for spacex they changed the rule so there's no minimum, it's just that below 20% float, companies would be weighted at 5x the float instead of 100% of market cap. If spacex is planning on something like 5% float, it would be weighted around 25% of market cap with only 5% of float available to buy.
But it gets worse because when the lock-up period expires in 180 days after ipo (currently scheduled right before quarterly index rebalancing), it's possible that frees up more than 20% of float and it suddenly has to be weighted at the full 100% of market cap -- triggering additional automatic buying.
It certainly seems like it's set up for our retirement accounts to be the insider's exit liquidity.
In hindsight it maybe should have also been obvious from the language alone. "Richmond Hill" feels a bit like saying "Rich Hill Hill" which is basically like saying "Wealthy Desirable Area."