With the introduction of GPT-io model this month, ChatGPT has succeeded in becoming the main subject of nearly everything, and it also has a huge presence in academic settings, since it performs best in general knowledge where lots of training data is available.
ChatGPT seems to be rapidly taking up to humanity catching up with olympiad level math students and being skillful enough in academic mathematics that it is a viable competiter to Post doctorate students.
Will this mean a decline to Google – the search giant? It might be true, but the blind spot that many fail to see is that ChatGPT can only perform well in some settings, and the most recent model, while there is much praisal addressed to it, actually has declined in complex knowledge that cannot be measured by testing, as it skews the original training data, and it also declined in the ability to produce relevant data from the prompt.
While the earlier models succeeded in being a delusional grandpa that has tons of experience, the current versions succeed in becoming a young prodigy that read(memorize) too much advanced material, but then started to attempt in thinking on his own and ended up being a odd mix of both ends – essentially useless.
Nevertheless, the current versions are scarily capable in providing necessary full assistance in colledge student level, low level knowledge. This will result in a rapid split between people who actively use ChatGPT, and people who struggle on their own. In my experience, many will fall in the former bin.
Moreover, even if ChatGPT replace every one of the human workforce, some human will still own the society, and as humans, they will want control over what they own, and therefore human analysis skills are essential skills that CANNOT be replace with any artificial intelligence technology.
That given, reading and writing skills – especially the ability to analyze data out of one’s own ability will be scarcer than ever, and will be more needed than ever.
Leave a Reply