Aaron Rosenbaum

AI leaders have a crucial chance to address risks now and strengthen governance practices before playing catch-up to regulations, notes Kjell Carlsson, Ph.D. of Domino Data Lab. In a recent VentureBeat article, he shares his perspective on California's AI bill veto and why proactive governance is key to the future of AI innovation. Learn more about this addressing this opportunity: https://gag.gl/4WCTkP #AI #AIGovernance #AIInnovation #DataScience

1

Turns out #GenAI takes some basic common sense: (1) use the right Model for the right Use-Case (2) use Open models to leverage the Open eco-system (3) run models - where they should run - based on data availability, proximity, latency, operational efficiency, etc... On Tuesday, July 23, #Meta announced the launch of the Llama 3.1 collection of multilingual large language models (#LLMs). The instruction-tuned Llama 3.1-405B, which figures to be the largest and most powerful open-source language model available today and competitive with the best proprietary models on the market. So, if you have a use-case that benefits from this large scale OPEN model - it is now available on IBM® #watsonx.ai - where it can be deployed on the IBM cloud, in a hybrid cloud environment or on-premises. #watsonx provides a consistent common platform for use with your models, open models, and our models.... Read the announcement here: https://ibm.biz/BdKqta

18

Excellent advice from Kjell Carlsson if you want a quick takeaway from the article below: “ Enterprise leaders should seize this opportunity to proactively address AI risks and protect their AI initiatives now. Rather than wait for regulation to dictate safety measures, organizations should enact robust AI governance practices across the entire AI lifecycle: establishing controls over access to data, infrastructure and models, rigorous model testing and validation, and ensuring output auditability and reproducibility."

7

New research suggests AI-generated misinformation is worse than we thought, but is anyone surprised? Google's own fact-checkers seem to be struggling to keep up with the flood of fake images online. This study highlights the urgent need for better tools to combat the spread of misleading visuals. #AI #deepfake #misinformation #disinformation #generativeAI #LLM #cognitivesecurity #google https://lnkd.in/gbKhSMds

14

2 Comments

AI leaders have a crucial chance to address risks now and strengthen governance practices before playing catch-up to regulations, notes Kjell Carlsson, Ph.D. of Domino Data Lab. In a recent VentureBeat article, he shares his perspective on California's AI bill veto and why proactive governance is key to the future of AI innovation. Learn more about this addressing this opportunity: https://gag.gl/4WCTkP #AI #AIGovernance #AIInnovation #DataScience

1

AI next frontier is possibly developing and scaling the reasoning layer, where AI systems possess hard problem-solving and higher order cognitive intelligence. Came across this good article from Sequoia.

20

This thought piece shows a way to break out from the disincentives and inefficiencies in current government acquisitions, and details why the market for AI models and inferences is uniquely well positioned to lead that charge. I’m excited to be part of a company that’s turning these and other innovative ideas into realty, and extracting real operational value from ML models.

9

AI leaders have a crucial chance to address risks now and strengthen governance practices before playing catch-up to regulations, notes Kjell Carlsson, Ph.D. of Domino Data Lab. In a recent VentureBeat article, he shares his perspective on California's AI bill veto and why proactive governance is key to the future of AI innovation. Learn more about this addressing this opportunity: https://gag.gl/4WCTkP #AI #AIGovernance #AIInnovation #DataScience

3

Excellent report from General Dynamics Information Technology John P. Curran regarding adoption and trends of AI in the Govt. I used Meta to generate summary below and included link to original article. Main AI use cases are chat and creating/updating computer code. Note also the 23% failure rate of AI projects due to user expectation misalignment. This can be overcome with thorough protoyping and establishing use case metrics. - Successful AI project completion timelines are narrowing, with an average of 14 months from pilot to production. - Future demand for AI projects is building around chat and coding-related functions. - 54% of surveyed AI experts said projects currently under development involve improving computer code. - 59% of projects involve "research and understanding", which may relate to creating chat functions. - 23% of AI projects fail due to misalignment between AI solutions and user expectations. - Partnership efforts are required for successful AI projects. Data diversity is crucial, and incorporating diverse data sources can lead to better AI outcomes. Trends: - AI adoption is increasing across the Federal government, with a growing focus on chat and coding-related functions. - Agencies are leveraging cloud-native services, tooling, and other capabilities to accelerate AI project development. - There is a growing interest in translating older code to newer languages, such as Python or cloud-based analytic tools. - Domain-specific chat functionality is becoming a priority, with many agencies planning to implement chat services in the next two years. Statistics: - 325 AI experts surveyed across the Federal government - 14 months: average time from pilot to production for AI projects - 54%: percentage of projects currently under development that involve improving computer code - 59%: percentage of projects that involve "research and understanding" - 23%: percentage of AI projects that fail due to misalignment between AI solutions and user expectations. Bill Casola John Grim Leslie Anne McIndoe Kim Giffin Michael S. Freeman III Tobias Quaranta https://lnkd.in/eqvmAVFG

3

Beyond single-score measurement and rankings this article provides more comprehensive approach to LFM (Large Fundamental Model) evaluation from end user perspective. https://lnkd.in/gSZbrqEb

20

I had a talk at LDBC's TUC meeting last weekend, the main topic is around our recent studies around GPU vs. CPU on graph algorithm performance, system cost, and energy consumption... there are some "disturbing" findings that CPU-centric architecture can be 10X faster than GPU-centric arch... meanwhile, being 70-75% cheaper and greener... #Ultipa #XAI #GraphAnalytics

12

1 Comment

There's a lot to keep up with in the world of LLMs. But understanding a few key foundational elements can get you pretty far. One framework I find useful is relating them back to the greatest hits of Creed — the 90s/00s commercial rock band that we all know and love. "Higher" - Think of how LLMs learn to reach increasingly higher levels of understanding: - They start with basic pattern recognition - Progress to more complex language understanding - Eventually reach sophisticated reasoning capabilities - Each training iteration aims to achieve better performance "My Own Prison" - This relates to the concept of LLM hallucinations: - Like being trapped in a self-created reality - LLMs can sometimes generate convincing but incorrect information - They're confined by their training data - Need careful prompting to stay accurate "What's This Life For" - This connects to the purpose and goals of LLMs: - Designed to understand and generate human language - Assist with various tasks and problems - Help make information more accessible - Serve as tools for human enhancement, not replacement "With Arms Wide Open" - This represents how LLMs embrace and process information: - Accept diverse inputs and prompts - Process multiple languages - Handle various types of queries - Maintain an open architecture for different use cases "One Last Breath" - This relates to token limitations: - LLMs have finite context windows - Must make the most of limited tokens - Need to be efficient with information - Sometimes have to truncate responses Each concept builds on the previous one, just like Creed's discography evolved over time. Just remember that LLMs, like any tool, have their strengths and limitations - they're meant to uplift and assist, not to replace human judgment. Hope you find this as useful as I do.

26

4 Comments

Faster to AI value with AlloyDB AI and machine learning are all the rage right now. In my last year as an analyst at Gartner, there was nothing short of a stampede to leverage some of that magical GenAI capability. Never mind that there are skills required to really use GenAI and AI effectively, and those skills were not necessarily present in the desirous organizations. These types of folks – which would include the overwhelming majority of organizations – had a lot of learning to do to get to that value. Does your choice of database management system make any difference when you are looking to create AI value from your data? After all, AI is implemented with compute cycles. Does your choice of a DBMS even make a difference for AI use cases? It can, by providing a smoother path to starting to use AI, or add additional challenges and overhead to walking that path. Although all database management systems can claim to support AI and AI models by simply storing the models as objects in the database, a DBMS can offer real advantages to get you to AI-created value to address business needs. Since AI is done with compute cycles, it means that all databases, of all types, can store data for use as AI models. So what makes a difference?  What database management system can provide the smoothest and shortest path to AI value? Based on what I’ve seen in the market, AlloyDB is very well suited to support the use of AI, in a number of ways. For more on this topic, please see my blog post on this topic - https://lnkd.in/gD2N6tAH.

46

1 Comment

Interesting notes from Intuit's recent webinar on LLMs. What evaluation processes do you deploy? at tasq.ai our approach is to integrate the AI models with diverse, global human guidance, to support #responsibleai faster and near real time.

4

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