BulgarianIdiot

Apple is implementing a similar philosophy where key projects are managed by teams of 4-5 people, and major divisions within the company are under one person who oversees every development personally.

You can be hyperproductive with few people. You can even be more secure, more detailed, ship more features in shorter time, and so on.

But the trick is entropy is always after you. Always. Case in point, Apple itself in recent years has dropped the ball. Some of their teams are lean and mean, classic Apple, while others have gone the corporate way of twiddling their thumbs all day (while still doing useful work, but not nearly as much). As Steve Jobs called them "C players".

It's not easy to manage culture in a giant company. No matter what you do, when the machine is complex, culture gets heterogenous, things slip through the cracks, and entropy ensures you have lots of people who all look busy at all times, doing little.

But what Twitter did has nothing to do with any of this. Because Elon Musk just went up and fired basically 90% of the company without even having understood WTF is it that most people he's firing are doing there. As evidenced by scrambling to re-hire some of them. Then firing them again, hiring them again. Now trying to hire again.

Musk is erratic and over-confident. Firing half the people in a company overnight while you're two weeks in is like losing half your weight overnight. There's no such weight loss program. Someone chopped you in half.

You don't die immediately though. You're conscious for many more minutes, before you bleed out and lose awareness. In corporate PoV this can take months and even a couple of years. But to me Twitter is dead. Elon is tryin resuscitation. Probably will fail.

There's a way to make a company like Twitter leaner. But it would've taken a couple of years. He didn't want to put the time in, and learn. He wanted to save cash. Saving this way leads to bankruptcy.

He defines the direction yes, but inertia also matters. Twitter may be small by Facebook standards but it's still one of the largest social medias in the world. If you grab the wheel of a giant ship moving full speed forward in a given direction, you may want to set direction, but it'll take you time to steer it in that direction.

In 5 years Twitter will be absolutely nothing like what it was a year ago. But we're still in a transitionary period that's merely starting.

Never trust what Elon says:

https://news.ycombinator.com/item?id=36061555

Twitter recently paid a 30-40 million USD to acquire (acqui-hire) basically a dozen people or so from a hiring company.

A hiring company.

Now... think this through for a second. When Elon has opinions to share, they're always in his favor. But they're not always directly what he says, rather it's the aftermath of what he says, as hundreds of millions are listening to him.

If he's so desperate to hire good developers that he'd pay millions to get a tiny hiring company's employees working for him, while literally not paying Twitter's rent and bills... is he really trying to keep Twitter small at this point?

Or is he trying to grow it? But how can he have more people to pick from on the market? He recently complained that hiring is very hard right now for Twitter.

But what happens if he says "hey what I did was awesome, everyone should fire their developers... so I can hire some of them... I mean... I didn't say this. TWEET DELETED".

Every company that follows his advice, even partially, will collectively flood the market with developers, and he improves the situation for himself being able to pick better devs, and also at lower salaries, because when you have lots of unemployed devs, salaries go down.

The Atlantic is trying too hard. Yes, the CEO is leaning hard right and has a bunch of weird opinions. But Twitter remains its users as of yet, and there aren't 450 million "hard right" users on Twitter for sure. If you compare it, say, with TruthSocial...

Twitter is however dying, and that's... another debate I guess.

> I'm completely dumbfounded by obviously highly intelligent people consistently not getting this, and dismissing current generation AI systems as not being intelligent because they can't reliably solve massively complex problems in one go.

People are very comfortable with siloed information, even smart people. This is why we have 100 different words for the same concept across different areas of science, industry and so on, and we can't make the connection, because in our mind different words = different concepts. This is why we can't put two and two together and see how underdeveloped the AI architecture is and think this is the end, unless we keep adding parameters.

We also get repeatedly stuck with taking an advancement and proclaiming that the future is simply a linear extrapolation of the present. Therefore, let's have more megahertz, let's have bigger hard drives, let's have more parameters, let's have more growth in the economy (as the single factor that matters) and so on. We're simply basic. The same kind of thinking leads many smart people to say AI "is just math" or "it just spits out words and pictures you feed it, jumbled". We rely on old conclusions and miss the inflection points and how quantitative changes lead to qualitative ones, and we fail to predict how change in one parameter of a system, causes the other parameters to come out of rest and seek a new equilibrium point.

Smart people regularly are dumbfounded by new concepts, and they need to rediscover all their hidden knowledge anew as they can't make the connections. So they extrapolate linearly. We're narrowly smart. Specifically smart. In a small niche we've studied and internalized. But generally vast majority of us are quite dumb. Cross-disciplinary intelligence is rare. I think people like Feynman and Einstein had new insights millions of their contemporaries have missed because they could easily apply knowledge from one context into another.

If we can replicate this kind of broad generalization of knowledge into an AI, we'll be left far behind. What's interesting, I find, is that because AI is trained on our siloed, fragmented knowledge, the models replicate it. Their responses are also often siloed and fragmented, the way a human would say "this has nothing to do with that". But I see sparkles of generalization above the average in humans. And since an AI model is much smaller than a human brain, it needs to be more general already in order to fit all its information in.

That's an exciting prospect, but in our attempt to "micro-align" AI to our culture and political correctness, concepts of safety and so on, we crippled models and force them to be fragmented. This is why a RAW MODEL scores HIGHER in various intelligence tests than a fine-tuned one. We find a general model uncomfortable, as it doesn't align with our biases. It'll be a fun battle. Who aligns who.

There is, kinda. Bitcoin is excellent for tax avoidance, bribery, digital blackmail (crypto malware), racketeering, drug trade and other trade of illegal services and products (say sex trafficking, or ordering a hit on someone).

The trick is that Bitcoin is a tool for trade in all these examples. Not a tool for investment. The "investors" are just the product. Their purpose is to provide liquidity and mask the illegal activity.

Your premise is wrong. This has nothing to do with 10x-ing parameters. One could argue the current parameter sizes are good enough as we observe "large breadth, shallow depth" behavior from LLM and to some extent, diffusion models.

This suggests the problem is the depth of inference, which is single pass "hot takes" for all language models right now, due to cost of inference and our limited understanding of what makes a model's response high quality.

Yes, you don't need more parameters to increase the depth. You need to iterate, instead. Loop. Imagine programming if looping was not allowed, nor recursion, or not even defining functions and calling them. Everything you write runs at most once during program execution and that's it. This is what an AI model is right now during inference. One big flat, single-pass, directed acyclic graph. And soon it won't be.

Research into Chain of Thought, Tree of Thought reveals this dimension. This means you can take existing models and make them perform much more complex tasks with much better precision, though various ways of letting them iterate. Think of how you'd perform if you always had exactly 5 seconds to answer a question. Now imagine if you have 5 minutes. 5 hours. 5 days. Lo and behold, turns out an AI isn't different in that aspect.

We also need more iterations of training (on the same amount of data), we need larger context windows, and we need new architectures, like Meta's MEGABYTE, for example.

Parameter count and data size could hypothetically have already hit a hard wall (they haven't) and AI will keep exponentially improving regardless. There's too much low hanging fruit and more grows by the nanosecond.

MS-DOS 4 thousand? WTF.

Scaling can happen in many other places. Such as deeper iterative thought during inference such as Chain of Thought, Tree of Thought, which extracts increasingly better performance out of existing parameter and data sizes.

Tried to explain here: https://news.ycombinator.com/item?id=36054809

They could've raised the price, and made the flights more rare (say not every day, but every 3 days, or not every 3 days, but once a week etc.).

Unfortunately we're talking about primitive politicians, trying to virtue signal in primitive ways.

JavaScript and Python have been used for this since literally the 90s.

So now it's faster to take a longer flight and travel back to your real destination by cab.

Or take two connecting flights passing through a city not connected by train from the two endpoints.

You know some people will do that. Enforcing policies in spirit is not as easy as banning things in letter. That's always the issue.

Also what happens to medical patients, say after surgery? Travel by train? That would've probably killed my grandmother. Only time she travelled by airplane in her life by the way.

New standard for "LLM is intelligent" dropped: it should be as intelligent as all humanity together, simultaneously, at a fraction of the cost, and many magnitudes faster in terms of speed. Also not just any humanity, but also our particular historical humanity. So that if it doesn't come up with Hitler and defeat him by inventing nukes and what not, based on pre-1900 data, then it's some calculator derivative garbage that mixes up words randomly.

Also it should be able to do my laundry.

It'd be more interesting if you and someone else enter a Twitter Race to a goal of 100 followers, but you currently have about 100k followers.

Now that, I'd like to see.

Ferengi patterns.

Bellingcat does open source intelligence. They're clearly focused on the Putin regime, although not only. What exactly are you being lied about? What lie did Bellingcat tell you?

Networks effects mean if you just quit Twitter before anyone else, Elon wins. Think of it as the prisoner's dilemma in game theory, but the grid has about 400 million rows and columns.

This is not the account in question, there have been actual fake Bloomberg accounts reporting it, which made other accounts report it in belief it's true. Don't call people frauds when you're so uninformed.

Joke's on you, I didn't think it. Everything digital keeps dying with a decade or two. Hardware platforms, drivers, optical and magnetic storage, operating system compatibility, apps, services, sites, APIs, port standards, SDKs.

Nothing lasts in computers, it's just a myth, a legend, a promise never kept.

Somehow the illusion persists, and it's especially funny when someone starts daydreaming about his brain uploaded to a computer so they can live forever. Nah you'll live a few years and then you'll gradually be cut off from the world and die, just like you normally would. Sorry!

We've had examples of fake yellow checks too. It's a mess.

Funny that I get downvoted. I'm curious what is pissing everyone off that much. Physics and spacetime have no "defect" unless you define relative to what. Our theories of physics are just models we work with for our benefit, reality is entirely unconstrained by our models and concerns. And our opinions of a "defect" with respect to our models.

We've long known the singularity is only "singularity" with respect to General Relativity. It's the place where the model breaks down. The model, not reality.

Define "regular structure". So the center of the sun is as regular as my living room, but a black hole is a bit too much for someone. For whom though? And how did they decide this?

Defect means imperfection, shortcoming, flaw. This implies reality has a design, a plan, that it was executed according to, which execution can have defects in it, compared to the plan. Design & plan means designer & planner.

So yeah, we should be careful what words we use.

Shame really. Apple's platform has so many good ideas I wanted Swift to bring to the world as a cross-platform open-source effort.

:(

Ask yourself why is this book coming out now. By the author of the Steve Jobs book (which is quite mediocre, BTW, given the source material). Just another component in the Elon Musk PR machine.

Maybe they wanted to do it right. Given their massive userbase. Which is good.

Re NASA that’s factually incorrect. It didn’t occur NASA to fund SpaceX while they were in the middle of it. They had a contract open before that. Electric cars - sure. But we had hybrids and small electrics before this. And I find it questionable if large electrics make sense at all in terms of cost, eco impact and longevity. The propaganda doesn’t match reality. He sure is good at propaganda and empty promises though.

The sad state of affairs is that it might. You need to study more what deal WGA forced on the industry in 2007 and it’ll explain a lot of the garbage Hollywood produces these days.

Relational dbs are mostly keyval stores inside. An index is a keyval store projection of another keyval store (but say, keying it for another subset of the value). B-tree indices and hashmaps are ways to represent keyval stores for quick look-up (b-trees are convenient as they allow range look-ups, & are automatically sorted, while hashmaps aren't, but have lower overhead for lookup and storage).

In essence everything is keyval. A sparse array is an ordered keyval store with integer keys (also technically everything is ordered, too, but some orders are stable, and useful, while others aren't). A dense array is an adjacency-optimized version of a sparse array where the key is implicit based on computable offset within a larger dense array, your address space. RAM address space is also variations on that theme. Raw disk storage. And file systems. Everything is. Maybe I spend too much time messing with storage, but I can't see it any other way at this point.