Looking back at the announcements yesterday, I think the guard system and the automated evaluation framework we launched address a real challenge in the generative AI space right now. I'm shocked there are not better options for this yet, it's been such a clear signal from our customer base implementing genAI use-cases. People want to use the most capable models, but they have very specific guardrails they need in order to actually trust it and to get the desired behavior they need. After that, they need to measure how well it actually works. It's popular to talk about LLM benchmarks with all the new models coming out, but this is fundamentally flawed for businesses to use. LLMs can be cajoled to output almost anything. Even the best are a prompt away from hilarious, embarrassing, or serious undesired behavior in different use-cases. Instead, I think businesses should focus on benchmarking the set of guardrails they put in place to control the generative outputs. These are real implementations of the use-case requirements, and they allow you to swap in new models rapidly while maintaining the same guarantees needed. I think this release provides the best options to do this well today. See the articles below for more info: * datanami.com/2024/05/02/data… * siliconangle.com/2024/05/02/… @DataRobot #DataRobotSpringLaunch

Filming for the @DataRobot April Launch

That’s a wrap on an amazing week at the @Gartner_inc Data & Analytics Summit! 🤖 We had so much fun meeting with analysts, customers, and broader #AI community. One of the highlights was our value-packed session featuring our CEO @Saha_Deban and CTO Michael Schmidt as well as AI leaders from our customers, Santiago Hernandez from Mercury Financial and David Coffey from CAA North & East Ontario, sharing how we’re working together to close the AI confidence gap and deliver real-world results for their organizations. #GartnerDA

Excited about the amazing field of Symbolic Regression (data -> equation) with the release of our Φ-SO approach ? Check out these resources for more on SR: (By the way, if you're in astronomy can you guess the equation being recovered by Φ-SO here ? 👇)

Following on work by @DoctorJosh @eureqa @tegmark @eigensteve on discovering laws from data, AI is learning do the critical precursor step: Discover the variables from which to build the laws in the first place creativemachineslab.com/hidd…

I'm giving a keynote on innovation at DataRobot on applied AI. Feel free to register if interested: aix.datarobot.com/?utm_sourc…

These solutions bring together nearly a decade of innovation at DataRobot to help solve critical problems like predicting supply chain disruptions and volatility, forecasting consumer demand, preventing fraud, and transforming patient care to help save lives-all faster than ever.

@DataRobot AI Cloud is here, 10 years and 1.5 million engineering hours in the making. Join us today to learn more about the next generation of AI: draicloudlaunchevent.com

I wrote a blog telling about some of the Eureqa story today, check it out here if interested: datarobot.com/blog/reimagini…

We are excited to announce our 2nd major release of the year! 🚀 @DataRobot 7.1 is packed with new features and enhancements across the board to take your #AI projects to the next level. Explore DataRobot 7.1: bit.ly/3vrqWLU #AugmentedIntelligence #datascience

My most recent @GoogleAI residency project is out: "Hamiltonian Neural Networks" Blog: greydanus.github.io/2019/05/… Paper: arxiv.org/abs/1906.01563 Starting from noisy (pixel) data, we can learn _exact_ conservation of energy-like quantities.

Mark and I gave workshop at #ODSC today on time series forecasting .

We announced a huge project I've worked on at DataRobot for time series modeling and forecasting. Check it out below: bit.ly/2OUq0ve

Michael Schmidt @DataRobot discussing automation using hybrid ML models #mlse2018: Effective for physical systems.

"Incorporating prior knowledge and constraints are important" - Michael Schmidt #mlse2018

Great talk by @eureqa about feature engineering for time series at @ODSC_Ukraine.

Jeremy Achin, @DataRobot CEO, #ODSCKyiv #AI #DataScience #MachineLearning

@pprett gave a great talk on Data Shift @ #ODSCUkraine Check out the sides here github.com/pprett/dataset-sh… #ODSC #Datascidence #ODSCKyiv

Mark Steadman - Time series modeling with Open Source Tools #ODSCKyiv #ODSCUkraine #Data @odsc

I wrote an article for kdnuggets on time series feature engineering: kdnuggets.com/2017/11/automa…