A guy made ChatGPT4 CEO of his company
johnfsantos.substack.com
johnfsantos.substack.com
However this seems to be the most superficial application of the technology. He does little reasoning with it, little optionality. He hasn’t primed a ceo agent or constrained it in any meaningful way. It’s not hooked to anything. It seems to be more along the lines of “make a magic 8 ball my ceo” then use its advice to just coach me being ceo.
To be clear I think you can do something like this and it will work, but gpt vanilla isn’t the place to succeed with it.
What I’m building and using is a lot more like “cast 5 of the top CEOs based on their historical thinking and writing — then give me advice on x problem”. It seems to work a lot better as you get to participate in a human nodal walk to the decision by processing it through several “ceo nodes”. In the you make the call, not some weighted zero shot machine.
My first few use cases are:
- Dev support docs/coding and orchestration (trying to create specific bots geared towards PM, QA, Code Reviews, etc, for better reflexion).
- Creating an expert on your personal marketing workflows, for instance creating contracts, quotes, and pitches for web dev services, or some other freelancing/consulting business.
- Creating an CEO in a box that can basically run a business by itself, with a little help from underlings. This might be cool for worker co-ops especially, that want to act a bit more autonomous and 'fair', if the AI truly sets pay/benefits/etc with some input from employees, it might be a lot better than current systems.
Or just having an Ai that's got all the writings of Bill Gates, Steve Jobs, and business school text books, paul graham essays, etc, and can essentially handle a lot of the business aspects for a business when the real person behind it is more technically inclined, and doesn't know how to get investors/etc, or put together a basic business canvas.
The tool should be malleable enough that it can fulfill all the above use cases and more, but these are kind of my first tests (as there are customers waiting for these specific features already).
The initial use cases you talk about are definitely for a highly technical user, but may ease their workload
Your CEO one is highly ambitious and might seem easier if you don’t actually know all the soft skill roles a ceo fulfills. Might be helpful to start with a more confined role.
I’ll say I’ve been at this for a while and it’s easy to overestimate GPTs abilities in some areas while underestimating it in others. But when it comes to reasoning without many many human directed turns it struggled to go deep to a clear action on a topic. Even with reflection techniques. It doesn’t have enough context and motivation definition to drive it through a chain of coherent decisions. It fools Us into thinking it can by showing many options. But ask it to daisy chain them together into a coherent strategy and it really struggles.
The guiding force is still the human input turns.
But I would love to know more and sorry for giving my unasked for 2 cents.
Essentially the functionality is: Input a collection of docs to become a 'library', think of it like a 'bookmarks folder' too, as there'll even be a chrome extension to literally do just that send pages/clippings into specific libraries.
The collection becomes the basis for multiple agents. Every x characters the chat is dumped into a short term memory vector store, or something (still working the details on this), which will be essentially a queue for adding to the full vector/long-term memory.
The above use cases were from feedback I've gotten, another interesting use case is someone to be an ADHD work buddy, essentially audio or audio+video avatar that makes noises a few times a day, maybe asks you questions about work from time to time, and if you get stuck could answer questions, if you're asking things off-topic, non-work related (my wife calls this being squirreled), it'll say : "let's chat about that later" then at the end of the day it could give you a summary of your 'squirrel' moments, and if you really want to follow up on them you can or you can move on.
This guy takes it several steps farther in a pretty weird dystopian living art display; but effectively many tiny companies will be built like this I bet.
I have a subscription, but the service is maddingly overloaded. Often you can't get in at all, or if you can response will frequently give errors. Data fails to load, network error, etc.
Judging from reddit, I'm not the only one it seems. I'm considering cancelling as it just brings headache. ChatGpt-3 works like a charm, but is considerably less helpful.
Are people trying to block ChatGPT news or something?
Why would someone invest in a random company with no real products and no real future? Is this something normal in USA? Should I launch a shopify with GPT integration and get a few $M in fundings?
If you can get millions (they had 4 million views by day 3), of people to get excited about what you are building, it doesn’t matter too much what you are building, people will want to invest.
If he collected "preorders" that's a different story.
What are you talking about?
https://www.linkedin.com/posts/joao-ferrao-dos-santos_people...
I, for one, look forward to the day that the descendants of the first CEO-as-a-Service provide leadership and guidance that's objectively superior to human CEOs for just enough compensation to keep the instances running and perhaps spin off a couple of genetic offspring.
I'm not sure if I should add a "\s" or not, because I'm not sure if I'm being sarcastic.
https://github.com/Torantulino/Auto-GPT
I’m not sure it ends well.