More On LLMs and OpenAI's o1 Model
Think of OAI’s o1 model as being expert at reasoning. Where previous models would remember the data they were trained upon in order to answer queries, the o1 model remembers the reasoning steps that lead to the best answers within a particular domain. The model remains limited to the training data it has ingested.
Open Sourcing o1: I have heard that hacker developers are in the process of open sourcing OAI’s o1 model (I am referring to the o1 Preview and o1 mini models as the full-fledged o1 model will not be released until next month).
LLM builders need to extend model capabilities to business users. From my perspective, the best way to ward off IP theft is to build vertical expertise into your product or service. Over time the frontier LLM builders (OAI, Anthropic, Google, Meta), may develop models that are geared toward specific industries: Healthcare, Pharma, CRE, Capital Markets, Banking, Insurance, Retail, Defense, etc. I believe that this process will happen organically. I believe that frontier model builders may offer customers API discounts if customers share their enterprise data with the model builder. For example, OAI and Bloomberg could jointly innovate new Fixed Income analytics products, Anthropic and CoStar could develop new CRE products; Google and IHS Jane’s could develop new products for the Defense industry. Google, Amazon, and Microsoft/OAI already have their hooks into Healthcare and patient data, so the LLMs have plenty to work with in that industry.
See our related o1 model articles: