AI 2 minute read
Training an AI System Really Matters
By Ciaran Power, COO, Ocelot
Artificial Intelligence (AI) makes sense to use in many contexts. My previous career was working with litigators. Litigation was an early industry to see significant use of AI. Attorneys are not natural early adopters of technology. They were forced to use it. Large litigations and investigations can often have hundreds of millions of documents to review. Limited time and budget made putting human eyes on all of them an impossibility. The solution was, and is, artificial intelligence.
To get great results from AI requires a well-trained system. In the legal context, an attorney very knowledgeable of case issues (the human knowledgebase) looks at randomly selected sample documents. She decides which are interesting and which are not. AI classifies similar documents as interesting or not. There are more steps, e.g., statistical validation, testing, classifications, etc. Only a small proportion of the total number of documents are looked at. Key is that the system needs training from somebody who is an expert. Good training of the system gets good results. Attorneys are often amazed at the high-quality consistency of responses that the system provides. Bad training often gets indifferent or disappointing results.
Results matter. For lawyers, the quality of outcomes can determine if clients win or lose. It can massively impact client costs and the reputation of attorneys. Consequences are similarly significant in higher education. A well-trained system can immediately provide far better results for students. They have better experiences. They may get timely information that can greatly impact their student careers. Financial aid and other student service offices get the happy students and efficiencies that they want.
AI-based chatbots are great solutions for today’s colleges and students. Deploying a chatbot that comes with a powerful knowledgebase (significant pre-training) will quickly steer a college’s AI investment towards success.