Each edition of Claude & Co. is based on huge amounts of existing knowledge.
The result may seem new, but it is not.
How should this be handled?
- The contents of the editions should always be viewed with a critical attitude, for example, the sources, the topicality of the knowledge used, the context.
- The relationships in the issues should be understood and comprehended.
- The content should be used consciously - after the content has been scrutinised.
Three underestimated fields of application for knowledge-based services
Legal system
I re-enacted a real medical malpractice case with Claude.
This revealed a partially hallucinatory language model:
- Court decisions were interpreted as law.
- Several paragraphs were taken from outdated legislation.
- The information on possible legal remedies was partly contradictory.
Only proven experts can categorise expenditure realistically and reliably.
Invention + Innovation
Language models combine what is already known.
Genuine inventions? Rare and random.
If you want to create something new, you need external experts, not just AI agents.
Democratisation of knowledge
Knowledge is available anytime and anywhere - but you should still understand it before you use it.
As a non-medical person, I can have a language model explain to me how an operation works.
But will I actually be able to operate? Of course not.
Language models give us insights into other areas of knowledge - but they don't make us more competent.
My conclusion
Even with AI, human judgement remains the key to quality, creativity and innovation.
Knowledge-based service providers are and will remain important - anyone who underestimates this may be overestimating the possibilities of artificial intelligence.
Responsibility cannot be delegated.