Abhijeet Gulati

AI model providers are updating their usage policies. Anthropic has changed their policy to classify insurance - including underwriting, claims processing, and coverage decisions as "High Risk", with the policy mandating human review of all AI output. Like many such topics, there's some nuance. Human oversight is a good thing. But insurance is a heavily regulated industry that already has policies covering how AI is used. Restrictions from AI providers are another challenge to navigate for insurers, and evolving use policies create uncertainty for insuretechs who are building off of AI platforms. Philip Dawson and I discuss this further in the linked article. While it's not always practical for businesses, updates like this are one of the reasons I'm a big fan of running your own AI models and not depending on big providers. https://lnkd.in/e5fVs6df

10

👉🏼 How effective are large language models in product risk assessment? 🤓 Zachary A Collier 👇🏻 https://lnkd.in/eZy9nVrh 🔍 Focus on data insights: - Large language models, like ChatGPT, show promise in brainstorming potential failure modes and risk mitigations. - However, there were errors and inconsistencies in some results. - The guidance provided by the models was perceived as overly generic and occasionally outlandish. 💡 Main outcomes and implications: - ChatGPT performed better at divergent thinking tasks but lacked depth of knowledge compared to subject matter experts. - Similar patterns in strengths and weaknesses were observed when comparing ChatGPT to other large language models. - Despite challenges, there may be a role for large language models in assisting with ideation in product risk assessment. 📚 Field significance: - Product safety professionals can leverage large language models for brainstorming and idea generation in risk assessment processes. - Experts should focus on critically reviewing AI-generated content to ensure accuracy and relevance. 🗄️: [#productrisks #AI #riskassessment #ChatGPT]

1

1 Comment

How is #bigdata helping to tailor medical treatments? A nice read below! “Transitioning to the future of healthcare, we will be relying more and more on real world data and platforms, including population health platforms.” -- The OM1, Inc. PhenOM Platform generates a digital representation of patients’ phenotypic profiles powered by RWD and #AI. Using PhenOM, complex patterning in patient data is translated for real-world personalization. Built for both research applications and real-world, in-clinic deployment, OM1 PhenOM can; ✔Find underdiagnosed patients and under-appreciated variation in patient phenotypes ✔Identify rare subcohorts of patients within broader disease conditions ✔Predict health outcomes, including flares, progression, and serious events ✔Inform risk management ✔Track changes in disease progression ✔Support treatment decision-making ✔Enhance case management When dealing with a mis/underdiagnosed condition like Systemic Lupus Erythematosus (#SLE), finding and identifying patients can be an uphill battle. With OM1 PhenOM, companies can identify undiagnosed patients and under-appreciated variation in phenotypes. Watch the video below to learn more about our new product launches on the OM1 PhenOM Platform: 📍https://lnkd.in/d4QtTnXs https://lnkd.in/emKS5VuY #realworldevidence #patientjourney #rwd #realworldoutcomes #datanalytics #dataquality #medicalaffairs #access #bigdata #ai #predictiveanalytics #personalizedmedicine #preventativemedicine #patientoutcomes #heor #healtheconomics #outcomesresearch #rwe #digital #precisionmedicine #patient

9

I am pleased to report that the second edition of the Python Natural Language Processing Cookbook came out recently. This is a joint effort between Saurabh Chakravarty, Ph.D. and me, and has 3 new chapters written by Saurabh, as well as new content in the original chapters. We included lots of recipes with LLMs, both commercially available and open-source models. The format is the same as in the first edition and all Packt cookbooks, with code "recipes" that you can follow. The book is available on Amazon: https://lnkd.in/dx_kHT2u #nlp #ai

396

83 Comments

Frameworks and advancements like Dspy are exciting and do not get the attention they deserve. Future state on AI (especially Agentic AI) cannot over rely on human prompt engineering. Worth a watch if you are building production AI solutions. https://lnkd.in/expCwiBE

12

SEAL is a new model benchmarking service from Scale AI that fills a big gap: an additional objective, third-party, trustworthy way to compare models. People distrust many benchmarks because they worry that models are optimized to perform well on them—including by training on the underlying data (like cheating on a test by memorizing the questions and answers in advance). The LMSys Chatbot Arena (https://chat.lmsys.org) addresses this by having humans evaluate model output in blinded head-to-head tests. The new SEAL leaderboard (https://lnkd.in/gdCZzRnz) does something similar, but uses "thoroughly vetted experts using domain specific methodologies" to do the head-to-head evaluation. Sounds much more expensive, but they just raised $1 billion, so I think they can afford it.

1

“Gen AI: Strategizing for High ROI Use Cases in Insurance Process Flows” This is the topic of our workshop at #FOIUSA24 tomorrow 10 AM CST. We are all SOLD OUT, wish me luck! Akif Tarique Pavithra Byra Reddy Tauseef Muhammad Hormuz Gotla Steven Levine Shilpa Kona #GenAI

107

5 Comments

There are numerous benefits to semantic tagging, including governance and data quality. ADI specifically uses semantic tagging to materially automate entity resolution and data matching. Our benchmarking demonstrated that our platform can reduce data integration effort by 70% due to automation. Moreover, the more semantics we add to the platform, the more use cases we cover. This unlocks data integration for the data citizen and not just skilled engineers. ADI is making it push button easy to integrate and connect all of your data.

13

We updated our demo for diagram generation out of natural language specifications. In this short demo, an activity diagram is generated out of the textual description of that sequence. #ai #genai #requirements

19

1 Comment

Attention - I mean RNNs are all you need??!! Are new linear RNN based models going take over the top spot for the model of choice in the future? Specifically Mamba? Dynamic 'Linear' RNN's based models with compute of O(n log(n)) instead of O(n^2) (transformers) are starting to take off but are meeting some resistance in the peer review process. How did they overcome the two main issues with RNN's- lack of parallel computations and the training difficulty arising from only learning from recent inputs, making long sequence problems intractable and reserved for transformer and cnn based models)? Read about how they work in the papers below- Linear RNN's(state space models) : https://lnkd.in/gechNx9V Mamba LLM based on dynamic linear RNN's (no attention, no MLP needed): https://lnkd.in/g8dSmhZt

4

Don't let AI hype drain your budget 🙌   Healthcare costs are rising fast. Hospital and labor costs jumped over 20% from 2019 to 2023.   AI offers a promising solution, but "AI washing" — vendors exaggerating AI capabilities — complicates decision-making.   How to spot real AI solutions:   🔸 Look for a skilled team with healthcare knowledge. 🔸 Demand proof of success with case studies to verify AI effectiveness. 🔸 Ensure data fits and works with your setup. 🔸 Keep data safe and follow guidelines.   One of the best questions to ask yourself is ‘Could AI help solve this problem?’ Avoid starting with ‘What can I do with AI?’ as this might lead to creating something unnecessary.   Smart AI choices lead to savings 👀   What steps do you take to verify AI tools? Let’s discuss in the comments!   #Health #AI #HealthAI #MedTech #DigitalHealth #MedAI #HealthTech #AItrends

8

1 Comment

Want to catch up on webinars exploring key #AIgovernance issues? Kevin Fumai’s list has you covered.

3

At a Generative AI thought leadership dinner with other CTOs in NYC today and some interesting trends are emerging: * a consensus that LLMs are converging on a general set of features and basically comparable performance * a consensus that an org's data pipelines will matter more than the LLM chosen * and a group of people who read the TOS that are really scares about what these LLMs will do with access to all your data. Lastly, my prediction: the work of tech is moving from the keyboard to the whiteboard. Ai as microservice means structuring your data, thinking with ai, and chaining ai agents will supercharge your career and your cause -- but only if social change orgs platform their data. In my study of polymaths, Charles Babbage designed the basic idea for a unit of computing being something that takes an input, does something to it and gives an output -- the so called Babbage difference engine. In a real way, GenAI is approaching a world where chaining inputs ala a babbage difference engine is accessible to all. This can be an awesome opportunity if we get Non profits to work together!

11

3 Comments

Organizations have insufficient technical capabilities for the speed and scale of their ML model implementations. My colleague Sumit Agarwal just published a very helpful research with a capability view and sample vendors for defining a modular, flexible and scalable MLOps solution. https://lnkd.in/gnB_T9bq #MLOps #LLMOps #AI

149

17 Comments

Your Co-Pilot for Predictive Analytics: Dr. Mo Introducing Software-as-a-Service: AI Dr. Mo. Why choose AI Dr. Mo Software-as-a-Service over Machine Learning? 1.    Speed: Your prediction jobs are completed in minutes, hours not weeks 2.    Affordability: A Special Introductory Trial $50 per prediction job. 3. Any professionals can predict with Dr. Mo 4.    Accuracy: Dr. Mo Multi-model guarantees the highest prediction accuracy. 5.    Data: AI Dr. Mo uses numbers, words, and symbols. Contact us at: [email protected] Soft10 - Developers of Dr. Mo, the First Virtual Statistician (soft10ware.com)

1