I've been busy. I haven't watched a lot of news or spent much time on LinkedIn. I just found out about Puffy/P Diddy a couple weeks ago and that J-Lo made a sci fi thriller. Both of those are things I never would've predicted and I still can't decide which is worse. The areas where I did prognosticate seem to be doing better than those 90s/Y2K icons.
Apple announced at their developer conference the release of the next OS for all devices will have AI, Apple Intelligence, integrated into the phone as only Apple can do it. Your AI experience is personal to you and secure. Here's some of the capabilities:
You'll be able to have Apple Intelligence integrated into other applications enabling Siri to take action inside of apps for you. They also mentioned improvements for Siri (maybe now she will stop calling me, Dan) taking it beyond a novelty to something actually useful.
Hey Siri, could you find that receipt from last week at the convention in Orlando for dinner that was between five and six hundred dollars?
Large Language Model is becoming a misnomer. With more models being multimodal (image to text, image to audio and text to whatever) they are more like large data models. Some are getting some smaller. Google released Gemini 1.5 Flash using less parameters with a context window of 2 million tokens that could accurately respond to questions with amazing accuracy (>99.7% for text, 100% for video and 99.1% for audio). That's like feeding it 22 hours audio, 2 hours video or 1,400,000+ words. This is important for a few reasons:
- Insurers often deal with claims involving complex narratives, medical reports, and legal documentation. The long context window allows you analyze the entire claim history, identifying patterns, inconsistencies, or potential fraud indicators that might be missed with shorter context models.
- You can analyze extensive medical records, including doctor's notes, lab results, and treatment plans. This can help in assessing the validity of claims, identifying potential pre-existing conditions, and flagging any discrepancies between the claim and medical documentation.
- Some claims involve legal contract or agreements. The ability to process large volumes of legal text can help insurers quickly assess the relevant clauses, terms, and conditions, leading to faster and more accurate claims settlements.
- Fraud cost billions each year. You can quickly analyze a vast amount of data related to a claim, including the claimant's history, the nature of the loss, and any supporting documentation to identify subtle patterns and anomalies that could indicate fraudulent activity.
- Underwriters can use this to analyze vast amounts of unstructured data like customer interactions, social media posts, news articles, and even satellite imagery. This holistic view can uncover hidden risks and patterns that might be missed with traditional methods, leading to more accurate risk pricing.
- By analyzing a customer's comprehensive profile, including their claims history, financial background, and lifestyle choices you can create highly individualized risk profiles. This allows you to tailor premiums more accurately, ensuring fair pricing for individual customers and minimizing adverse selection.
- Examining all historical claims data, combined with external factors like weather patterns, economic conditions, and even social sentiment, to develop predictive models for future claims. These models can help insurers anticipate potential losses and adjust premiums accordingly, ensuring they have sufficient reserves to cover claims and maintain financial stability.
Nvidia recently released a digital twin of the Earth to predict weather patterns more accurately. It generates 12.5x higher resolution images than current numerical models 1,000x faster and 3,000x more energy efficiently.
There's a lot 2024 still to go. I expect we see more companies thinking outside of the atiquated insurance box, creating a strategy to implement AI and automation in ways that will improve the customer experience, increase employee engagement, and impact their bottom line.
