I had some meetings with this martech founder who builds AI agents for marketing, and over several meetings we sort of developed this hey bro vibe. And over a few drinks, i really started giving him a piece of my mind that AI agents are just wrappers, it's almost freaking hype. AI is too expensive to replace you, all big tech giants are just looking for excuses to fire employees in the name of AI, basically everything an anti-AI camp guy would say. No holds barred. again i was just curious and surprisingly he agreed his AI agents are just wrappers. But then, he told me something that completely blew my mind. a few days ago, he gave a demo of his AI agents to an HOD at JBL, who was basically a boomer when it came to AI agents and after the demo, this HOD guy was like, can you help me create a WhatsApp broadcast channel, i want it for my wife. And i was like fk, how did this guy even become an HOD at JBL. But that's what's so crazy, the AI adoption gap is something no one really understands. AI is being sold to the wrong people. I guess that's why every smart AI company is chasing folks from BFSI, manufacturing etc because they don't know much about AI or how to implement it in their workflows.
Comments (6)
I don't get this though. I've made several pieces of software that are actually useful. I've automated dozens of things around my home, started writing my own audiobooks with clones of my own voice, started a podcast of actual quality, and the list goes on and on. I think the hard part is creating actually good and useful things with AI which is actually completely possible, but it takes understanding and skill. If you just whatever it spits out it won't be good enough, but like anything else it is a tool and if you spend time learning, testing and optimizing you can use it to create really useful quality stuff.
I think the biggest issue is that it's so easy to create stuff that on the surface looks good but is actually a mess it that those types of things are the most commonly produced things. But it's still possible, with intention and care, to create actually good and useful things with AI as well. Those are probably fewer and farther between, but that's what it is. We'll see tons of crap for a while but there are people actually building quality stuff with AI. But it takes skill and understanding to do it and to separate the wheat from the chaff.
I don't get this though. I've made several pieces of software that are actually useful. I've automated dozens of things around my home, started writing my own audiobooks with clones of my own voice, started a podcast of actual quality, and the list goes on and on. I think the hard part is creating actually good and useful things with AI which is actually completely possible, but it takes understanding and skill. If you just whatever it spits out it won't be good enough, but like anything else it is a tool and if you spend time learning, testing and optimizing you can use it to create really useful quality stuff.
I think the biggest issue is that it's so easy to create stuff that on the surface looks good but is actually a mess it that those types of things are the most commonly produced things. But it's still possible, with intention and care, to create actually good and useful things with AI as well. Those are probably fewer and farther between, but that's what it is. We'll see tons of crap for a while but there are people actually building quality stuff with AI. But it takes skill and understanding to do it and to separate the wheat from the chaff.
I don't get this though. I've made several pieces of software that are actually useful. I've automated dozens of things around my home, started writing my own audiobooks with clones of my own voice, started a podcast of actual quality, and the list goes on and on. I think the hard part is creating actually good and useful things with AI which is actually completely possible, but it takes understanding and skill. If you just whatever it spits out it won't be good enough, but like anything else it is a tool and if you spend time learning, testing and optimizing you can use it to create really useful quality stuff.
I think the biggest issue is that it's so easy to create stuff that on the surface looks good but is actually a mess it that those types of things are the most commonly produced things. But it's still possible, with intention and care, to create actually good and useful things with AI as well. Those are probably fewer and farther between, but that's what it is. We'll see tons of crap for a while but there are people actually building quality stuff with AI. But it takes skill and understanding to do it and to separate the wheat from the chaff.
You did this because the technology is being subsidized at a loss. Literally thousands of dollars every time you ask an LLM to recommend a brand of toothpaste. It's way more expensive than people realize. There is no path to profitability. But the purpose of commercialization was adoption, so people would see it as less menacing than the next level of surveillance tech which it is, I.e. Flock cameras, Palantir, etc.
Here’s the path to profitability: all the frontier ai companies go bankrupt once the “cheap” models are good enough for most tasks since nobody will pay for high end stuff. No return on commoditized AI = bankrupt for all those companies.
Someone then buys the infrastructure / IP from bankrupted companies for cheap and sells frontier model access for the per query cost while having written off the r&d costs. At this point it will actually be profitable.
I agree with you, as it stands now, nobody is going to pay the actual cost for this stuff while the development cost is still on the books.
You have to realize a big part of this is marketing. You know all those hour long meetings that go no where, and could have been a quick email? That’s what AI is now, hour long demos that waste everyone’s time and produce very little gains. But like a lot of work it sometimes comes down to theatre and performative office politics.