Transcription
If you are hiring an AI company and they tell you that the AI is learning based off of the conversations that they're having and so it gets smarter and smarter, they're lying. That's not happening at all.
For that to happen, that would mean that their AI is learning, retraining itself on every single text message. It's not happening. You need an immense amount of data to do any sort of AI training. That's why these companies have billions and billions and billions of dollars because you need an insane amount of resources.
What's actually happening is it's not learning. It's just adding new information to the FAQs essentially to the filing cabinet. So this is called RAG, retrieval augmented generation. An AI learns statically and then they release a new model. It's not like learning on the fly. It's just storing data and referencing it.
That's like what makes AI so incredible. It can continue the conversation based off of information that you've given it so that it really feels like you're talking to like a person because it's like referencing things that in the data set, but it's not being trained on that data. It's just going into a folder. It's not actually like evolving as an AI model. It's just filling up the filing cabinet.
It's really no different than having a large set of FAQs that you keep adding to and keep adding to and you keep refining it. Keep adding more and more information so that the AI has more and more stuff to basically converse with and to make sense with. So it's not learning, it's just searching. It's searching for better information that it hypothetically now has.
When like ChatGPT or Claude or Gemini or any of these models are learning, like they're learning in closed doors. It's not learning on the fly as you use it. They're training it over at Google and OpenAI and Anthropic. They're training it. They're training it. They're training it with billions and billions of data sets and all sorts of things. And then they release a new model and that model is static. It's not learning after it's been released.
So, that's a common misconception that people have about, "Will this AI get smarter about my business the more we work with it?" And the answer is no, it's not going to get smarter. You're just going to give it more information. Competitors are always saying they can do things that they can't do and it's super annoying. So, we're just here trying to set the record straight about what is actually doable and what is not doable and what the truth is. So, that's it. Hopefully, that clears up a little bit of misconceptions.
You know what I also hate about AI? I hate about AI that it gives everything a title. The smoking gun question. Like, why does it do that? It's like all AIs have like the worst personality. So annoying. Like, you know how much work we have to do to make AIs not annoying? A lot. A lot of work. That's it. Later.