Posted by GrumpySciGuy 3 hours ago
People laughed at Gas Town, and at that "Fences, not Sandboxes" article yesterday.[1] They're both AIs set up as an organization, not an emulated human. They're both expensive to operate, because the AI members of the organization spend a lot of time talking with each other. They're both applied to frivolous problems.
Now we have the same concept, applied to a more useful problem. It's time to pay attention. Something important is happening here.
Everybody has been talking about "human-level AI", or "superhuman-level AI". This is a different kind of AI - corporate-level AI. It works and thinks like a company's management. It watches a wide range of data sources and responds to some of them with action, and some with internal discussion within the AIs. It can potentially pay attention to far more inputs than any human. That overcomes a basic problem of corporate organization.
A classic management failure is not paying enough attention to something. This is usually not a problem for computer systems. Humans are bandwidth-limited. Teams of AIs are far less bandwidth-limited. They may soon be outperforming humans for that reason alone.
Think about this for a while. These team AI systems may be very useful.
You're probably totally innocent, but this comes across as being AI-written, especially in the context of an AI problem and when your response is pro-AI. For the next 3-12 months (maybe more, we'll see), I suspect you'll have a lot more success if you shy away from those speech patterns.
But it turns out it's better to do it the other way around. A good plan is easy, implementing it is not.
So the responses will be even more incoherent than Opus 5?
I hope legislation is never manipulated enough to allow an AI to form a corporation or become it's legal director. With whom would accountability rest? Who would you throw in prison if it acts egregiously enough to warrant such a measure? If it's bots all the way down.
Corporations already exist to limit liability and accountability. This would be an extension of their current purpose not some weird deviation from.
Prompts like: you are an activist in XYZ, you wrote this book. Not tell us if this decision is ethical, and if you approve it.
Or running entire HR department in AI...
> A classic management failure is not paying enough attention to something. This is usually not a problem for computer systems.
This is actually my #1 problem with frontier LLMs. They are plenty smart, but they have no true long term memory, and their context window is not nearly enough for a complex system like a corporation. When my project gets big enough and I'm using a lot of /compact, the LLMs always have the same failure point: they fail to pay attention to something.
When they figure this out, I think it's going to be a very, very big moment.
I’m adopting this stack under me (CTO) and integrating it into agent zero
CEO jobs are safe until the humanoid robots have mastered this skill.
> One of the miners, rejected on account of his size, offered to fight any man there as proof of his suitability as a soldier, and six men were eventually called upon to remove him.
Sometimes the claim is "set vision" - which AI can extract from the collective.
Someimes the claim is to prioritize the work (with bias), AI can extract that from the collective too, right?
Making resources available to the team? AI.
Coordinating between teams? AI.
Can the AI do this in a more un-biased way? Can it refine collective intelligence and desire into objective goals?
More than a software factory id love to see a manager-factory.
One thing, AI cannot do (yet) is bring the network. Like when you hire folk from Booth or Sloan so you can sell to places that hired Booth and Sloan grads.
Not yet. Maybe soon. I'm not placing bets on timelines.
> What do CEOs do
No clue. I'm not there yet. Maybe it's critical. Maybe nothing.
> and other managers
There are a lot of bad managers. I'm halfway not terrible. A significant part of my job is setting technical direction in the same way I did as a staff+ engineer but without the overhead of asking somebody else's permission. Another large part is the "shit shield" function -- somebody thinks they want something, and they either find themselves blocked by me or immediately satisfied with whatever code I'm able to cook up between meetings, not breaking the team's concentration. A large part of it is navigating organizational nonsense (for "reasons," I'm rambling, ask if you're curious). And... whatever. My job didn't get easier going into management. Not all of it should have to exist in the abstract, but in any specific company it's still an important function.
Be friends with the board. Negotiate high pay and benefits (for themselves), and a golden parachute in case things go south (which they do most often). Take credit for things that were going to happen anyways. Blame xyz for things that didn't work out.
These are all impossible tasks for AI. AI can't play golf.
> The job of an executive is: to define and enforce culture and values for their whole organization, and to ratify good decisions.
(And that's the whole list! Nowhere do you find set tasks or make decisions)
To me, that means that they're still pretty essential - on both fronts, AI needs values and AI needs a system that can tell whether decisions are good or not.
Well, so do most CEOs.
For example an open-source project could/would still have a BDFL but, the AI-CEO could do the operational busywork while the BDFL is out talking to humans and presenting at PGCon or some other productive thing.
The way he phrases it, if I remember correctly (and this is by no means the only way to executive, I just really like his framing), is that the way you make decisions is by getting your immediate subordinates to agree, and they iterate. Decisions should be made at the lowest level that they can be made at, and if that's the CEO level, you've made mistakes.
So you make the values list, you make sure that people live by it (up to and including firing people who don't and hiring people who do, which he basically thinks is the whole day to day), and then you bring your subordinates together to agree on decisions. If they don't agree, you send them off to work through it together and come back with consensus (or, less great, compromise). If you have to put your thumb on the scale to get a good decision, again, you're not CEOing well.
Basically if you have to be a dictator, there's no way you can actually effectively do that. Proper leadership is getting people to act the way you would, if you had all the knowledge and perspective they do, but without having to make them do it. Lift them up, empower them, give them the responsibility and tools and values, and then endorse the result.
This is not a competitive advantage
Most times it's pretty simple, in my experience. A CEO either runs a business as a (1) Sales focus [best deals and vendors] or (2) Marketing focus [best reputation and trends] or (3) Investor focus [best boardroom friends].
Lots of other opinions are usually thrown out until there's really only one facet used for running the company.
The best products are usually from single minds with a strong vision. "Too many cooks in the kitchen" is a phrase for a reason.
Your least favorite products are designed by committee or a vision less person following trends.
So they need someone to review AI output huh?
At my previous 50k person company I never talked to anyone within 4 levels of the CEO and thought he was useless. But realistically he probably had a somewhat similar job - talking with the board, talking with politicians, and talking with the CEOs of other companies in the effort to lower costs (tax breaks, partnerships, etc) and generate sales (contracts, partnerships, etc).
If that kind of thing is easy for you (the honest ones or the clandestine ones) then you might make a really strong entrepreneur/business leader.
At least the good ones do.
To be an effective CEO, you have to make judgement calls that the collective can’t always have wisdom and context on, is confidential, or requires fast decision making during which there are interactive discussions or negotiations!
My personal take, somewhere down the line people feel scared of taking responsibility for more than themselves and their family. This limits the ability for most people to take on such ventures. Not everyone falls in the same bucket but a lot of leadership in todays date is about taking the role without responsibility. They happen to be in the role with the help of network affect and somehow that just amplifies.
I believe that the average are average.
I don’t believe what I wrote suggests that. I was attempting to be as neutral as possible.
Without a CEO to fire when the business goes south, how is the board supposed to respond? Taking accountability? lmao
That is what the CEO is for – to be on the receiving end of beatings and flogging if the business they are at the masthead of effs something up.
Unbounded liability means nobody starts a business, or they do it under the table without responsibility or taxes.
Turns out we want successful private businesses much more than we want to hold a few nasty CEOs accountable.
I deliberately stopped short of invoking golden parachutes and the numerous other escape hatches available at that altitude, as I am perfectly aware they exist. I am equally aware of chief executives who have departed with remarkable agility from organisations they themselves had first steered off the cliff.
My point, however, remains the same: as the letter of the law presently stands, AI can't be made accountable – at least not today.
https://archives.fbi.gov/archives/oklahomacity/press-release...
The answer is to be responsible to the business. The buck stops at the CEO. Therefore, their responsibilities are technically everything in the entire company. And I think many people who have only worked for other people underestimate how valuable that responsibility is.
The thought experiment, to be clear, is building an artificially intelligent CEO or company founder with the intent of generating profit. If this AI can turn a profit, it has passed this new test
Of course, by the time this new AI starts turning a profit, it will be maximized to the extent that it loses its competitive advantage, and then we’re back to square 1. That is, when everyone uses the same AI, nobody has a competitive advantage over anyone else
My current theory is that because of this, it is not possible to build an AI that generates a profit perpetually. Only temporarily. Then human intervention would be required to adjust things to make the AI competitive again. But at that point, it’s no longer an AI driving the profit, it’s its human tinkerer
1. https://en.wikipedia.org/wiki/Tendency_of_the_rate_of_profit...
https://github.com/SenteLabsAI/OpenExecutive/blob/main/packa...
“Walking away costs you the deal. Signing costs you the company. The math is obvious in retrospect; harder in the moment when there's quota pressure”
such valuable, should I say load-bearing insights
Other gem from my new CFO:
“The financial statements are the result. The variance analysis is where the operating insight lives.”
And my new C(3-)PO just told me
“If we removed all sales involvement, what % of revenue would we still get? #5 is the honest PLG test. If the answer is <30%, PLG is supplementary at best. If the answer is >60%, PLG is the actual motion and the rest is overlay.”
But oh ho there is no human ceo, must be a good place to work!!
Interested folks can start by putting in shareholder resolutions with the proposal and the prompts that would replace CEO. SKILLs.md?
See you can add anything there and the sentence will make sense.
I will never vote for an AI system to actually make laws lol wtf are you even saying
Let the C-suite do it all with AI alone and see where they get.
A warning: Don't burn Chrome.