Thomas Burg’s Post

Great achievement of our AI team. 🎩

I think the notion, ⁠AI results are raw material (Cassie Kozyrkov), is a pretty good perspective on understanding what #gai (=generative ai) currently is, in particular w/ re to regulations and enterprise readiness … #generativeai #llm #riskmanagement #gpt

Just started today with this fine conversation, you should too. Covering #nlu #nlp #ai #knowledge #language #semantics

Our team looked into disambiguation and ChatGPT. The article discusses the progress made by transformer-based language models, such as ChatGPT, in addressing linguistic ambiguity in NLP. We conducted an empirical analysis to understand ChatGPT's strengths and weaknesses in disambiguation tasks, providing data and model outputs used in their experiments. The team considers this a promising starting point, but cautions that further improvements are needed to address any remaining flaws in the model. #nlp #chatgpt #gpt3 #dip #documentunderstanding

We're working on a series of blog posts about how #AI and #analytics can support #contractresearchorganization during the many stages of #drugdevelopment. There's a lot to gain (money, quality, efficiency, patient involvement) from dealing with lots of data (text, image and sensor data), automation, and analytics. Please reach out to me if you want to learn more or want to share your use case and/or comment.

Hat tip to my colleagues at dezzai. Checkout their Medical NER in Spanish tweets exercise. The team used token classification based on a Transformer model (XML RoBERTa) and a medical thesaurus (UMLS) which complements the TF. The results are great and make clear that this hybrid approach performs excellently. #ai for #health. #socialdisner

as usual, the better the data you input the better the output. Anything that helps getting order into the data mess increases the leverage…

Content curation is important. It's a time saver but it's also a relevant means of decision making – for the recipient. Now, we have literally tons of content (data), more so every hour. Don't ask those who curate manually. Instead of considering this as a challenge let's think of these tons as a gift. The more the better. But only if a machine does the legwork. Right, with the help of smart algorithms we ask the machine to come up with what's hidden behind all the noise. We want answers, we want advice. The decisions are still up to us. Curious, let us guide you in this regard…

#pharmacovigilance does benefit so much from utilizing smart technologies, even more so the people working in this field. It helps to stay above the waterline in a rapidly growing ocean of text, voice, and image which contains adverse events, outcomes etc. Check out this concise overview on the current situation of AI and pharmacovigilance > https://lnkd.in/d2_Xt3_h. And please, if you are in this business, make sure you don't miss our offering below. It helps both #pharma and #healthcare to take care of our lives.