Call us now:
One persons ‘world model’ is another’s ‘domain specific model’.
In general LLMs are exactly what they say on the tin, they are general models. They have, and do, consume allot of data from domains they can access. They find it difficult to consume data from domains, subjects or topics they cannot access. In short todays LLMs don’t ‘watch the world’, they consume what they can see, but that is a fraction of the data, information and knowledge that exist.
For example I have a watch maker living near me. He is one of the few people left in the country that can repair specific watches and clocks from the 17, 18 and 1900s. When he dies his knowledge and experience goes with him for ever. It will never reside in an LLM.
For AI to effectively enable experts in specific fields of work, AI needs to consume data that is specific to those fields. AI needs to be domain specific Dan Miles, it needs to ‘watch the world’; gather information from the world, and learn from the world. AI models needs to be domain specific, we need domain specific models.
That’s what we created, domain specific models.
Faisal Khan Michael G. Jacobides Yann LeCun Tom McLeod
https://lnkd.in/ep5xXXjZ
Aegis Labs
#artificalintelligence #ai #domainai
View Original Post
https://www.linkedin.com/feed/update/urn%3Ali%3Ashare%3A7353311597399842816