The video is a kine of misleading, which should told you that the driving in the video was a vision of the future rather than a capability it already had or would soon have.
I am fascinated by how to train a model to do the math. As we've known that one crucial factor to make models so powerful is to conceive a deep question, like filling the missing words in sentences, for them. And what is the underlying question of the doing math ability?
ChatGPT trained by current methods as a search tool might have the "spamdexing" problem too. People can still find ways, which might not be as easy as it is now, to generate content and make them be included in the training data.
But I think LLMs like GPT can be used as a great weapon against spamdexing because of their ability to "understand" text, which can help improve nowadays search engines, like Google, a lot by applying them as a filter when the spiders are crawling the web content.
In the short term(2-3 years), I think, the hybrid(programmers give high level designs and instructions + AI models implements them) way of programming will become increaseingly popular. Nowaday's LLMs have shown some suprisingly abilities of reasoning, remembering, and imitating, but, for now, they are still not good enough to write competent code alone, especially when it comes to complex fields like system design. Things like system design need a lot more sophisicated thinking process than implementing some specific functions.
