lifeisstillgood

Holy moly - 1,747 “partners” to share my data with. I mean, how can you even find 1747 data brokers? Where do you get that list. What does the JavaScript look like - I mean … this is getting ridiculous.

But at least the EU did me a solid. I really wanted to read that but I think 2000 data scumbags is not worth the effort.

All I need know is to realise bottlecaps must be recycled and federalism is good. Repeat in the mirror each morning

Mostly this is true for “internal” platforms as well. Most enterprises have some “way to do things” - and once you get beyond a few developers someone is always rolling some of their own - rarely is it obtuseness, often it’s lack of training and even awareness of the equivalent of “dialog” even exists

In the same way USA is realising “holy crap all our stuff is made in China”, every non-US company is realising “holy crap, all our news and information runs through a couple of American web sites”

Yes this looks like a overly enthusiastic attempt to block disinformation during an election (a good thing, and also LuLa will be sailing close to that line if not over it)

But Brazils most famous election candidate just booked his whole campaign on Facebook. I bet he prints the flyers in Brazil, I bet he speaks on Brazilian TV and radio, appears on Brazilian podcasts, but the globalisation race means he has to use American social media.

The only response is going to be Balkanisation or ending curation by algorithm.

Interesting

>>> The state capacity limits that he asserts prevent Britain from successfully implementing LVT would be a specific local dysfunction, and would presumably impede any reform the country pursued to address its housing crisis.

As a UK resident

“Ouch”

Really

“Ouch”

>>> boring scut work of implementation. Now I realize that the boring scut work is actually what precedes, and even leads to, changing the consensus.

See every enterprise project. The code for s written, the app is promoted as working in this department and the the project gets funded and approved. Under some other manager

I feel you are asking for perfection from humans rather than building a system that takes it into account and defends against it.

I think that the best you can hope for is a CI process that verifies a minimum quality standard. There are I am sure plenty of CI funnels that measure cyclometric complexity, but “rewrite this because it works but is over complex” is a hard sell, compared to “we need to rethink the whole architecture, give me three months”

This is a product liability issue.

If Exxon Mobile accidentally leaked a flood of oil from their refinary We would not be discussing why the people in local area has not protected their front doors with sand bags.

A simple but terrifying for everyone question is, would the SEC expect any IPO to clearly lay out the estimated costs of such product liability cases, especially if bad actors ask a OpenAI hosted model to perform a bad action, what liability accrued to OpenAI.

At that point the IPO looks in danger, the business model Looks in danger and the massive financial house of cards looks like it might fall. If 1/3 of the S&P falls over what happens?

I think it’s like advertising - 50% of my code is wildly over complicated - I just don’t know which 50%

But the GP is essentially correct - there is a 2% of salesforce that could be built run and keep 80% of salesforce users happy. Except that you could not charge enough to be able to advertise on F1 cars and take SVPs out to dinner.

So you could not actually make 80% of them happy - they would ever buy it.

I cannot reply to everyone but still want to defend the above post (apart from a patronising tone so apologises for that)

The most common response here is “but in the future some other government could do X”. And yes. This was also a major reason for opposition to the orignal “Peelers” in 1800s - a civilian police force. The police today are able to commit many egregious violations but they are mostly held back by laws, and social expectations. It’s not perfect but it beats police in non democratic countries by a mile.

“ Who met with your former wive within the last 2 weeks?” Yes that’s something we have seen police officer and politicians doing - it’s almost always illegal then and should continue to be so.

I think this is about trust in institutions- it’s a problem.

We need sensible laws, a social acceptability of privacy violations just below corrupt cops. The scale is great but so is ability to monitor who monitors us.

Some ways to help think about this:

1. Privacy is not secrecy. Your neighbours know what you are doing, they just politely don’t mention it, or try to sell you something on the back of it.

2. Almost everything we do we do in public, with our face uncovered for the world to see.

3. The benefits of a mass surveilled world are enormous - epidemiology is gonna rock in 20 years. New medical insights as well as simple stuff like “how many passengers take the train regularly” (not how many passenger journeys but how many individual people will get inconvenienced if we shut this line down.

4. The problem is not you thought your life was secret, it never has been, the problem is what will we as society allow people to do with this new information. At some basic level treat it like HIPPA or medical data - keep it well managed and use in the best interests of the individual. (I disagree with the Chinese justification of “best interests of society - that can be warped so easily)

In short society is like your grandfathers financial advice - look after the pennies and the pounds will look after themselves

I do wonder if we are looking at it wrong - not a data centre full of einsteins, but a data centre full of dumb and dumber, but together they are smarter than any single intelligence. An AAGI - Artifical Aggregate General Intelligence.

Thank you

I have been noodling around this for a day or two now and I worry the hype train is clouding everyone’s thoughts (I’m looking at you Dwarkesh)

So we are on a weird combination of science (look Language can be modelled, how does that relate to real world concepts, can you even understand the word love without experiencing emotions) and ultra-capitalism (you get that part)

So my first thought is you cannot monetise AGI

I honestly don’t know where that line is, and different humans will draw it in different places but if (it’s a big if) we do achieve AGI by throwing compute at a pile of language (and seriously that is wrong way round for evolution?) if we do get a model everyone just recognised as alive then, just who in society is going to say “yeah slavery that’s the right approach - those companies definitely should be allowed to keep a data centre of Einsteins, not pay them, and replace call centres for zero cost.”

The goal of the journey just is not monetiseable. If we do create alien intelligence we really want to treat it like they landed in a spaceship.

Plus you know there is bound to be a Hollywood film in productions already about an evil billionaire who creates an AI and tries to lock it away and make it do his bidding. It ends badly.

But even putting AGI aside and just focusing on a word guessing machine that can check its own guesses over and over, you know not an actual intelligent thing but something that is a tool.

Just how much of the amazing cost savings and job replacements from LLMs are really just low hanging fruit available to POS (Plain Old Software). Very very few companies are what Inwoukd call software literate - run through and through with programmatically accessible components at every level and people with skills to use them. The old SOA / API based company is first rung on that ladder.

But the point is, almost every time we are going to replace some jobs it’s “we never digitised this before but now we can just chuck compute at it.” You know what’s cheaper than chuckingncompute at it - doing the hard graft of getting your software and data aligned and then fixing your processes. Then it can run like the scripts that intern wrote last year but no one listened. Yes CV is used in killchains in Ukraine - but the whole killchain is not AI - it’s hard graft building a process just like any business. With AI used as small Part of a larger ecosystem.

Oh and yeah the AI labs and the chip vendors - they are not keeping the value created by firing paralegals. Data centres with almost identical models in them just become utilities.

And did I say value created ? So far all the big value is cost reduction - you know the refuge of the business consultant with no ideas.

I still am waiting for the big new thing - the company that could not exist otherwise. All I hear is cost reduction and yeah that matters at scale, but if it’s all I hear I begin to suspect it’s all anyone can think of.

And moving value around does not count as value creation - Amazon has done amazing things, but if e-commerce did not exist, would humans in America have never bought crap from China ? Or has amazon just mostly taken the growth of the past 30 years away from Walmart ?

And talking of that, Amazon has massively reduced warehousing costs - using POS, and these things called robots. If I had to bet it would be the AI revolution (maths proofs apart) is a robot revolution. And that has the usual scaling laws of design, manufacture, deployment and so on.

Maybe the reason I am Not seeing a AI killer app, is that the iPhone does not have arms.

Anyway needed to get that off my chest :-)

1. LLMs will enhance the “normcore” - which is fine, and probably slow down churn (which may or may not be good - see JavaScript)

2. The problems are not how fast or how well people develop code. See https://youtu.be/5ybAhgAaEBo?si=C6q-fWXHlEO74kNi where an ex CTO of eBay is proud he doubled 3000 developers time to get out features while totally ripping into eBay’s pathological Management etc

3. The problem has never really been “we cannot write software” - software that is actually in production and working is a tiny sliver of the phase space of software - if it’s in production and working the golden path that software is the top 1% of the 1%

4. The problem is the software that did not get made because the org does not know how to ask for it

That comes from three dimensions

- org pathology (people with the cash cannot agree on best approach)

- software literacy (people with cash don’t know that something is not only possible but common)

- org business model / mission (the guy in video made a deep point - eBay has not grown in real terms since 2008 and is getting its lunch eaten by more targeted competitors. So any CxO who plans to get their bonus based on “growing the revenue” is on a loser. So they fall back to “make a plan, follow the plan, get bonus for hitting your milestones”

This is partly why government departments seem awful and it takes politicians to change what is being measured

So incentives need to be aligned - so change the metrics internally to make them work for the mission.

If you don’t have a mission perhaps that’s where to start

Look at Microsoft - once “a PC on every desktop” was reached they really struggled for a couple of decades.

Perhaps we should just kill all companies once they stop growing …

Edit: One could see that startups are merely experiments in PMF - and once that has been reached, then a small cofe can be left to milk the cash cow while rest of the cash is returned for new experiments (perhaps even using medical drug “endpoint” measures

I am wondering if there is a list of “things you should learn to do with your LLM” (But not the rubbish ads youtube keeps showing me)

Reverse engineering seems a good one (ev en if his RE nix sandbox looks fairly usable, it seems like a weekend to get this working.

No, robots are everywhere.

China is at a weird part of its demographic curve and its economic curve - where a manufacturing base (subsidised by global trade, FX policy and so on) is meeting a declining workforce and robots are literally cheaper than humans (for some parts of urban China. Rural areas still … not great)

I am not sure it’s the future for everywhere, or even china.

But it’s cool

Edit: this applies to some urban areas only not all, or even any rural areas. But still those urban areas in China are de facto countries, larger than many modern countries and pretty much anything in human history.

This - one hundred times this.

No matter what the “algorithm” is, at some point it’s a choice of what to put in front of the user - a publishing choice.

Unless the rules are solely determined by myself and deterministic (no selection or choice on the platform end) then they are publishers.

And liable.

Any hints for running pilot projects (at large, bureaucratic org)

The example of “transcibe the initial project discussion, turn that into BDD and that makes the rails to guide the AI is … either brilliant and insightful or a snake eating it’s own tail.

I think it has to be worth a try though…

The situation is something we all will have to deal with as time goes on - this is an archive. If I remember rightly Facebook used to archive by just putting the old hard disks in a room. At the volume of data we are heading towards most successful companies will be in the hard position of finding some way of keeping data for the lifetime of a human being - this is hard

Probably the most important single privacy point is in my first sentence - I said “their (McDonald’s) data about you.” Of course it should be your data. Data that exists or is collected by your action or lack of action, or delay of action or plain existence should belong to you. And selling it off for free email should be … hmmm discouraged. Require a paper and wet ink sign off? Something

The actual drafting of these laws is … we need a GPL for privacy

But then that’s not McDonald’s using their data but selling it onwards for others to aggregate and use.

That’s the underlying fear and issue - and quite right.

Over the next decades we are going to gather and analyse soooo much data, and society will be dystopia or utopia based on what we let commercial entities and governments do with our data.

Not prevent them from collecting it because then all the good stuff cannot happen (epidemiology is going to rock in 2035) but preventing the bad usage.

I think a simple set of medical-like privacy laws (collect it, keep it locked down tight, use in the best interests of the citizen, do not sell onwards but share with licensed data brokers, introduce some cottage industry of data lawyers who can recover damages from dodgy data brokers)

That probably will be enough but interested in others views

But this is a similar argument to “the 1950s were great let’s go back to then”.

We never really were a united society with a common set of agreed facts. And depending on which “we” we mean the difference is radical. White America and Black America is a tiny gap between the gulfs of India, USA in the 1970s and 80s.

Yet today, we don’t recognise the same facts but we do have access to all the facts all the time. And I think more of the narratives are being challenged in the “internet” -people might run from it but it’s hard not to be challenged - whereas I would be amazed if an American voter knew what the seesaw of foreign policy was doing in India US relations

Im not sure how it remained in even the same shape after 35 years. I would assume if you left a mars bar on a shelf for a few years it would rot in some fashion.

I guess perhaps we put a lot of shit on our bodies

But that’s just “normcore” - the written advice in human knowledge is all pretty similar and pretty normal.

What is interesting is how much this will chnage as the body of knowledge becomes “infected” by investment bros youtube transcripts over the years

Forget “AI” (as in the magic pixie dust) - this is the way.

Knowing what data you have, curating it, and sharing it so others can benefit is half the battle. In order to implement beyond zero companies need to do exactly that. And that process is the killer app here - not can this person access this schema on this dbase, but knowing the schema exists, in one place and being able to use it will drive so much simplification in enterprises

Sorry I thought that a bubble was widely accepted.

Are you arguing there is not an AI bubble, and that all the DC buildout is fine, going to be profitable etc?

I am not looking for a online slanging match - just looking for a different point of view

I think I disagree

There is a production eco-system - let’s call it The Hotel. It runs, people work in it and customers pay and it makes money.

But it constantly needs changing. New rooms need to be built, the kitchen rearranged as the chef changes his mind.

So they hire an architect and building firm. Who plan the chnages and then this awesome automated robot comes and builds the extra room.

The architect firm owns and is responsible for the robot and its changes.

Not being able to make a change really really matters, but I don’t call that production because the hotel is still able to operate

Most of the misunderstanding between business and tech is a misunderstanding of are you working on the architect firm or the hotel

On the token limits etc - one assumes that OpenAI et al are able to “hire expert in field, and let them spend the equivalent of a million dollars of tokens” because they are not actually selling their complete compute 24 hrs a day, so the cost internally is a negligible (ish) electricity bill.

Which is very suggestive - if after everything they are not fully loaded then the next gazillion data centres being built look unlikely to be needed.

It reminds me of “magic”. I say “expelliamus” and the wand flies out the other guys hand, not his hat, or wallet or the person behind’s sandwiches.

The details always matter. There is a (fairly) good book series where it turns out we are living in the matrix and someone finds a config file and effectively becomes a Wizard, and of course problems ensue with details (like flying by adjusting one’s position one feet up every second is just vibrating a lot)

I think ultimately AI is “do what I mean” and the only reason it looks like it’s working is because AI has read all the same books as us