Ram Bala

As digital and AI transformations accelerate, the potential of hashtag #DataSharing is becoming more evident. Pioneers like Walmart and Procter & Gamble have, over the years, reaped significant benefits from embracing data sharing, yet many organizations remain hesitant. Thanks to new technology, data sharing has become more secure and efficient, easing operational hurdles and regulatory compliance. These advances also redefine the role of trust. In my most recent Fortune column, co-authored with Riccarda Joas, Guillaume Sajust de Bergues, and François Candelon, we explore how these technological advances are impacting the risk-reward calculus. We provide arguments that might encourage executives to reconsider their stance on data sharing. Read more on Fortune: https://lnkd.in/eWP2K6KF

27

1 Comment

82% of consumers worldwide have awareness of GenAI, with a quarter of them having already utilized these technologies. Notably, these figures are slightly lower in the US, with a 75% awareness and 23% using it.     Digging deeper, consumer reactions to GenAI are also varied, with emerging markets leading the way in AI enthusiasm and consumers most excited about its potential in the workplace, rather than daily lives. In the US, 30% of Americans feel excited about AI, a substantial 70% have mixed feelings or express concerns.    Strategic Recommendations for Leaders:    Build Trust: Openly communicate both the benefits and risks of AI to address consumer apprehensions effectively.  Manage Sentiments: Navigate through the fluctuations of misinformation, enthusiasm, and apprehension effectively.  Engage actively in markets that show a higher receptiveness to AI by piloting innovative ideas and products.    For detailed insights, explore the full report: https://on.bcg.com/3UtXhRM 

28

Excellent article by Isabelle Bousquette about spreadsheets and how their versions can quickly proliferate. Any leader focused on improving productivity has encountered this challenge - many versions of a spreadsheet and each has a slight variation. It also highlights the sharp contrast between what enterprises can achieve with AI and what the ground reality is with regards to data - thus what they are able to? And most importantly it brings forth the challenges with tabular data in enterprises. This problem does not occur with images - when has someone opened an image and changed a pixel value or a few of them and stored a version of it ? Tabular data is messy, complex and is the bread and butter of enterprise.

14

4 Comments

A decade ago, data was completely ignored. It was siloed, and largely ignored. In 2024, data governance, business intelligence (BI), and data analytics have taken center stage. We live in an era of digital revolution where data is the king, thanks to technologies like Generative AI. Now, every organization wants to focus on implementing data governance. You need to govern data well and make sure it's reliable. This means setting governance standards and keeping data quality high. One essential part of data governance is deciding who must be given access to what data and why. This calls for a solid data governance base. Whether it's for inside use, sharing information with people outside the company, or powering AI, clear rules on data access are a must. Looking ahead, I expect that AI will play a big role in data governance. As data and AI come together, organizations will need strong AI governance to use insights ethically. Ignoring these changes could be risky. At SignaTech, we know the importance of keeping up with new technology and focusing on data governance. We want to manage risks and find new chances for growth and innovation in the data world. #datagovernance

14

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

Just published my latest article: "Transforming Business Operations with LLMs: A Path to a Production Ecosystem." a breakdown of the key components necessary for integrating LLMs into core business operations, including Human In The Loop (HITL), Feedback, Evaluation, and Decision Executor subsystems. Learn how these elements work together to start building an AI platform.

37

13 Comments

Just a wee little article to help bring back Game Theory to the discussion of #artificialintelligence. Also, Phys.org is amazing. https://lnkd.in/gvM2niMK

11

🌟 Exciting News in AI Forecasting! 🌟 Google Research has unveiled a groundbreaking approach to time series forecasting. Their new decoder-only foundation model is setting a new standard in accuracy and flexibility, capable of leveraging vast datasets across multiple domains seamlessly. This innovative model bypasses the need for domain-specific adjustments, offering a robust solution that adapts to various applications from finance to weather forecasting. Imagine the possibilities when these predictions become more precise and reliable! For professionals involved in data science, AI, and predictive analytics, this could mean more accurate forecasts, efficient resource management, and ultimately, better decision-making. Eager to get my hands dirty with it. Do you think that the future of forecasting will be in tied to the evolution of foundation models vs current approaches? #ArtificialIntelligence #Innovation #GoogleResearch #Forecasting #FoudnationModels

76

5 Comments

What an amazing episode of the Super Data Science Podcast with Jon Krohn! Sol Rashidi, a data executive and author which I admire a lot, shared her approach to executing successful enterprise AI projects. She emphasized the importance of balancing offensive and defensive strategies, building a cohesive data ecosystem aligned with business needs, and overcoming challenges like role confusion and data governance. Here are some highlights: 1. Build a data ecosystem that aligns with business needs ⚙️ 2. Prioritize AI projects based on impact and feasibility 🚀 3. Focus on high criticality, low complexity AI use cases 📈 4. Maintain clear data engineering and data science roles 🧩 Her book, "Your AI Survival Guide," offers valuable strategies to maximize impact while avoiding common pitfalls. Don't miss this insightful conversation! 🎯📊 #AI #DataScience #SuperDataScience #Leadership

7

3 Comments

If the P Value is the probability of observing a difference as large or larger than the one you observed assuming the null hypothesis is true - where does that probability come from? How is it actually calculated? What I am trying to answer in this post is - how do we go from data to a probability of how likely we are to have gathered that data and calculated the results we did? If the above statement is not clear, I recommend you check out my earlier post explaining what the P value is: https://lnkd.in/gtRQWdD6 Assuming you have a good understanding of what the P Value represents, we now need to explain what a probability distribution is in order to understand how we calculate the actual probabity (P Value) from data. A probability distribution is a function that describes how likely we are to sample an outcome from an experiment. You can visualize this in the graphic below - the X axis are the different values you can sample (as an example if you were to measure systolic blood pressure the mmHg values would be plotted on the X axis and the Y axis would represent how "common" or likely you would be to measure any give possible value - in our blood pressure example we would see the tallest point on the graph around 120 mmHG (assuming a healthy population) and the lowest point on the graph at the > 180 mmHG mark on the X axis. To calculate a P Value we need to assume how the population is distributed - we need to assume which probability distribution function we should use - more times than not we assume the normal distribution because of large data and the Central Limit Theorem. From there, it is as simple as finding the our calculated test statistic value (usually the average) on the assumed distribution and we check to see how far away (how many standard deviations) our calculated test value is from the assumed mean (this is indicated in the picture below by the various lines and bars with the different shading of blue). The further our calculated value is from the assumed mean, the lower the probability becomes that we are to sample that value. THIS VALUE IS THE P VALUE! The table below the graph in this post shows you the various P values (running across the top rows) for a t-distribution. You would find your degrees of freedom (not covered in this post but coming soon :)) and you would then find how many standard deviations your test value is from the assumed population mean. From there you simply look at the top row to identify just how likely you would be to have calculated the test statistic given the null hypothesis. I know this is a lot and a fairly complicated topic. If you have any questions about this or would like to learn more about the P Value please do not hesitate to reach out and please check out my website for a lot of free biostatistics materail :). Happy Monday

100

6 Comments

Is RoI the holy grail when implementing AI? Reading through Deloitte‘s State of Generative AI in the Enterprise Quarter two report (https://lnkd.in/dSYrcjnT) I came across a quote of a Fortune 500 company, that they have established a very strict rule, that if they don’t see value form Gen AI solutions, they would not scale it. Of course, demonstrating value is important, but on the other hand data without utilization/monetization is also useless. Businesses not starting to use Gen AI run into two major problems. 1.       They will fall behind those who step into Gen AI usage already today and risk to lose contact. 2.       It is kind of naïve to think that your employees are not using Gen AI for their benefit already today: Without offering a managed solution the risk of data issues (data protection, data quality, …) is extremely high. AI and Gen AI must get part of a business’s tools portfolio today. To be as safe as possible companies should follow certain “rules”. ·     Any use case must follow the business strategy. If it does not, it is very likely to not produce value for the business. ·     Install a C-level authority taking care of data and AI. Unlike IT, AI is about content creation that helps driving your business through innovation or boosting efficiency. ·     Chose to act ethically. The discussion about job-killings, data protection, … will not seize. ·     Skill-up to scale-up. Plan for educating your workforce. How are you scaling (Gen) AI in your organizations?

9

Friends, we often struggle to choose the right metrics to evaluate our machine learning models. I want to share a perhaps little-known metric - the ksaai correlation coefficient - for identifying complex relationships. The original article you can find here: https://lnkd.in/d8HyZhnb The Python code and the brief introduction you can find here: https://lnkd.in/dtzuifTa. Thank you Tim!  

7

In this blog post, we break down these two essential concepts in data analysis with a touch of fun: "Correlation and Causation" You’ll learn how two variables can be related positively, negatively, or not at all through our explanation of correlation. We also delve into what it truly means when one event causes another, and how to establish this connection, helping you grasp the concept of causation. Discover why distinguishing between correlation and causation is crucial for making informed decisions, developing effective strategies, and minimizing risks. Start reading now: https://lnkd.in/dK9uAHMX

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

Excited to have Vrushali Sawant, a data scientist with SAS's Data Ethics Practice (DEP), on the latest episode of the ODSC’s Ai X Podcast! In this episode, Vrushali Sawant discusses trustworthy AI, including an overview of what it is, what a career path in trustworthy AI looks like, how you can implement trustworthy AI, and what’s already being done to promote ethical AI. We'll also explore: 🔹 Real-world examples of the effects of bias 🔹 The AI Incidents Database 🔹 The importance of data management for trustworthy AI 🔹 The hurdles to achieving trustworthy AI 🔹 Practical strategies for identifying and mitigating bias 🔹 The importance of asking the right questions at the beginning of the AI lifecycle This episode is a must-listen for anyone interested in building AI applications! 🎧Listen Now: https://hubs.ly/Q02vKPhx0 #AI #Podcast

6

If you use large language models (LLMs) in your research, register for this seminar to hear Elucidata’s Dr. Abhishek Jha discuss how data quality impacts LLM performance. A reliable foundation that is well annotated and accessible to an LLM plays a major role in the value of its results. Using cancer research as an example, you’ll see examples of how LLM-powered artificial intelligence (AI) agents query across three versions of the same gene expression corpus with differing results. #medaff #datascience #webinar #AI https://buff.ly/4b4Jfvv

2