Luke Mueller

The single MOST important news in AI for this month, perhaps this year & a new defining moment for this decade is here - Safe SuperIntelligence Inc. Ilya Sutskever - the brain behind some of the most advances within AI in past decade, who failed to oust his co-founder Sam Altman at OpenAI last year, started this new company ssi.inc where they plan to not only showcase that #AGI is within reach, but also how #SSI is within reach while keeping it "Safe". "Safe" - a word spoken far less than #AI & #LLM in past two years but is considered critical for safety of future humans - is becoming the basis of this start-up. This is the first tangible step anyone has taken in past 1.5 years towards AI safety. While many have raised alarms, none took any concrete steps. And I loved its website containing just one short webpage write-up. Worth a read. Shows the current state of world of AI & world. #Elon started another company #xAI - a competitor to OpenAI. Got one of the largest funding in Series B - $6bn. Has now setup the largest data center "Gigafactory of Compute" along with Dell Technologies with upto 100K NVIDIA Hoppers-100 which is nearly 4 times larger than largest GPU clusters till date, like Meta's. But he hasn't mentioned a thing about safety - which he vehemently spoke about last year to halt all work on AI development when #ChatGPT was released. Hopefully, he is also leading the work in this direction (apart from for-profit work of utilizing all the data on Twitter/X and leveraging it to build one of the largest financial institution of the future) #AISafety

18

5 Comments

I have an amazing team of data scientists at Labcorp. Many of them can work shoulder-to-shoulder with data scientists at the big tech companies, but they see our mission to improve health and improve lives and know we are onto big things. With that being said, the most interesting thing about the GenAI revolution is that: 1) the longest tenure of anyone with GenAI skills is 18 months. 2) the skillset is brand new, and NOTHING like previous gen AI where we spent our time training models using convoluted python code and linear algebra, not consuming models. 3) I like to think that teachers have the greatest pre-requisite skill set: being concise, clear, and setting objective expectations is the key to working with GenAI. And I'm not talking about using GenAI to write your emails or send a snarky response to an incompetent coworker (promise I've never done this!). GenAI seems scary to many - but the skills required to use daily are within reach of virtually anyone. My advice: find a course and get started. 😎

99

2 Comments

Interesting article. With the ever increasing size of context windows, massive prompts could potentially outperform fine-tuning for LLMs, based on a study from CMU. That means depending on situations, one can choose to front load their cost to fine-tune LLMs, or to have larger prompts and pay more per transaction to get to similar performance. https://lnkd.in/euFTP8QS

5

👉🏼 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

👉🏼 Predicting glaucoma before onset using a large language model chatbot 🤓 Xiaoqin Huang 👇🏻 https://lnkd.in/eehWZ8E5 🔍 Focus on data insights: - ChatGPT4.0 demonstrated an accuracy of 75% in predicting conversion to glaucoma one year before onset. - AUC of 0.67 and weighted F1 score of 0.77 were achieved by ChatGPT4.0. - ChatGPT3.5 provided an accuracy of 61% and a sensitivity of 64% in predicting glaucoma development. 💡 Main outcomes and implications: - The performance of ChatGPT4.0 was reasonable in forecasting glaucoma development before onset. - Large language models (LLMs) show promise in enhancing glaucoma research capabilities and clinical care. 📚 Field significance: - Future exploration of ophthalmology-specific LLMs utilizing multi-modal data and active learning is crucial for improved integration with clinical practice. 🗄️: [#glaucoma #language #model #data #insights]

1

Snowflake Dark Mode + Notebook + Cortex LLMs + Price Transparency data from Payerset = instant insights to help patients under the cost of care across various hospitals and surgical centers. Built in mere minutes with admittedly very limited knowledge of python - but Arctic and Copilot helped me write out the code. So fun to get equipped this week at Snowflake Data Cloud Summit and see so many friendly people on a mission to do more with data and AI. Safe travels home all - let's compare notes! #snowflake #aidatacloud #datacloud #snowflakesummit #datacloudsummit

59

5 Comments

Instantaneously access huge volumes of Price Transparency data for Aetna, United Healthcare, Cigna, Elevance, BCBS South Carolina, Kaiser Permanente, and Sentara (among others) on the Snowflake Marketplace. Here's one of the free datasets from Payerset: https://lnkd.in/ediBmkZr #pricetransparency #healthdata #datacloud #aidatacloud #payer #payor

9

1 Comment

In the current landscape, we've witnessed significant advancements in Large Language Models (LLMs). Precise assessment of these models is crucial for model selection, performance evaluation, and identifying and addressing biases. Let's talk about BLEU (Bilingual Evaluation Understudy), a widely used method for performance evaluation.  This metric, an extrinsic evaluation tool, compares model-generated outputs with one or more reference texts to gauge similarity. Scores range from 0 to 1, with higher scores indicating superior performance. This approach proves invaluable for evaluating the performance of translation models. Scores below 10 are negligible, while those exceeding 50 denote good translation quality. The BLEU formula comprises two key components:  1. Brevity Penalty: Penalizes overly short-generated translations compared to the reference length using an exponential decay 2. N-Gram Overlap: Assesses the match between unigrams, bigrams, trigrams, and four-grams in the generated output and the reference translations. To prevent overcounting, the n-gram counts are capped at the maximum occurring in the reference. This overlap is calculated as a precision-based metric, with the geometric mean of these four precision values being taken. Important considerations when utilizing BLEU for evaluation: 1. BLEU is corpus-based: It's not suitable for assessing individual sentences. 2. Lacks differentiation between content and function words: For instance, removing a function word like 'the' incurs the same penalty as replacing a significant term like 'Earth' with 'Mars'. 3. Limited in capturing sentence meaning: It only considers up to four-grams and overlooks long-range dependencies.

14

👉🏼 Assessing the Risk of Bias in Randomized Clinical Trials With Large Language Models 🤓 Honghao Lai https://lnkd.in/eBZ2cduS 🔍 Focus on data insights: - Large language models (LLMs) facilitate systematic reviews. - Methods and reliability of LLMs for assessing risk of bias (ROB) in randomized clinical trials (RCTs) are explored. 💡 Main outcomes and implications: - LLMs demonstrated high correct assessment rates for ROB in RCTs. - LLMs showed substantial accuracy and consistency in evaluating RCTs. - Potential of LLMs as supportive tools in systematic review processes. 📚 Field significance: - Enhancing efficiency and accuracy in assessing ROB in RCTs. - Advancing the use of LLMs in systematic review methodologies. 🗄️: #datainsights #riskofbias #RCTs #systematicreviews

1

Important development,  we cannot increase the context window to billions of tokens. At some point a data structure that can hold the context is going to become imminent. GraphRag is going to solve limitations of autoregressive systems. 

13

I'm at the Databricks AI and Data Summit in San Francisco this week. The question I am seeking to answer is "How can we make great decision-making more affordable and accessible to everyone?" Currently, it's really hard and expensive to produce great analytic insights. And those insights are super valuable. Let me explain with an illustration. How mass production has elevated humanity: ---------------------------------------------------- Over the past 300 years, the masses of humanity have elevated their quality of living through mass production. Take the example of the Green Revolution in agriculture: The introduction of high-yield crop varieties and advanced farming techniques dramatically increased food production. This helped reduce hunger and improve nutrition worldwide, showcasing the transformative power of mass production. By lowering the cost of goods and services, we’ve made them more affordable and available to more people. Why can't we translate this idea to the realm of insights?

4

🚀 Big Tech's sprint towards supremacy in LLMs has us on the edge of our seats! Snowflake's Arctic, Microsoft's Wizard LM2 & Phi-3, Apple's Open ELM, and Meta's Llama 3 are reshaping what we thought possible in AI. Each offers unique features, from open-source frameworks to small but mighty models. It's an arms race, but does everyone win? Great for AI advancements! 💥 💡 The top takeaway: Innovation never sleeps in the AI domain, with each new model pushing the boundaries further. It's exciting to see how these rapid developments will impact the future of AI and tech integration in our daily lives. 🤔This rapid pace of the LLM arms race raises several concerns: ▪ Bias and Ethics: As LLMs become more integrated into various sectors, the potential for biases encoded within these models increases, which can perpetuate discrimination or unfair practices. ▪ Privacy: With the increase in data processing capabilities of these models, there's a heightened risk of misuse of personal data and privacy breaches. ▪ Resource Consumption: The environmental impact of developing and running large-scale models is significant, given their substantial energy and computational requirements. These issues highlight the need for responsible AI development, ensuring ethical guidelines and sustainability practices are in place as this technology evolves.   🤔 What impact will these rapid developments have on the future of AI and tech integration in our daily lives? #GenAI HOPPR

30

5 Comments

Today, we're diving into the cutting edge of Retrieval-Augmented Generation (RAG), a paradigm shift that's turbocharging LLMs with the vast knowledge of your BigQuery data. Think about it: 🤯 - LLMs are powerful, but they're only as good as the data they're trained on. - BigQuery is a data powerhouse, holding the keys to your unique insights. Enter RAG! By seamlessly integrating BigQuery with your LLMs, we unlock a whole new level of intelligence. 🧠 We dive deep into the mechanics, in this blog post we cover: - Understanding RAG's potential: Explore how it transforms LLMs from generic models to highly specialized experts in your domain. - Practical implementation: Learn how to seamlessly connect BigQuery and your LLM, leveraging Google's tools for effortless integration. - Real-world use cases: Discover how RAG unlocks powerful applications, from personalized recommendations to cutting-edge research. Click the link below to read the blog post and discover how to unleash the full potential of your LLMs: https://lnkd.in/e8jacn9a Also see the other blogs in the series and watch this space for more: 1 - Multimodal input to Gemini to BigQuery Schema 2 - Creating marketing campaigns using BigQuery and Gemini models https://lnkd.in/enJ5rxec Watch the video https://lnkd.in/ekzeNcs9 hashtag#RAG hashtag#BigQuery hashtag#LLMs hashtag#AI hashtag#DataAnalytics hashtag#GoogleCloud hashtag#Innovation hashtag#FutureofAI hashtag#LLM hashtag#Gemini hashtag#BigQuery hashtag#datawarehousing hashtag#Cloud hashtag#marketing hashtag#cooldemos hashtag#bigdata hashtag#lakehouse hashtag#genai hashtag#ai hashtag#artificialintelligence hashtag#datamodelling hashtag#google hashtag#dataengineering hashtag#analytics hashtag#ml hashtag#GenerativeAI hashtag#BigQuery hashtag#DataAnalytics hashtag#AIInnovation hashtag#GoogleCloud hashtag#DataScience hashtag#MachineLearning hashtag#gemini hashtag#vertexai hashtag#agents

7

Check out this highly recommended course for those interested in optimizing their models: - Introduction to Quantization - Symmetric and Asymmetric Quantization Modes - Hands-on Implementation from Scratch using PyTorch - covering Per Tensor, Per Channel (Axis), and Per Group techniques - Weights Packing: Exploring 2 or 4 Bits Quantization Methods https://lnkd.in/gMAuWXHm

43