Logan Kilpatrick

I talk a lot about "openness" as being key to including more key communities in the development of AI broadly and LLMs specifically. This week, I'm covering a specific way we can do this in the alignment space with existing fields of social science. I was on a recent paper "Social Choice for AI Alignment: Dealing with Diverse Human Feedback," which addresses many of the core questions I've been raising around reward models. Let us use this as an example that we don't need to reinvent all of human knowledge to solve alignment, we just need to work better, together! Read it here: https://lnkd.in/gq6xZ22c

42

1 Comment

Very solid lecture on the absolutely mind-blowing progression of LLMs, even though the transformer came out in 2017 the last couple years have been crazy for open source model research. Learn about how models are made useable. https://lnkd.in/gsfX7BCG

8

The latest OpenAI update and all the 'uncensored' models have me thinking about the three laws of robotics and Apple's HIG. Seems like it had Ivan Poupyrev, Dr. thinking the same. https://lnkd.in/egpsSFdf Please take a look at OpenAI's Model Spec https://lnkd.in/ev2rdWXD and consider the three laws when building AI applications. We owe it to the future. Daryl Thomas Mitko Vasilev Phillip Remaker #AI #GenerativeAI #DesignLeadership

1

Interesting new research on generative AI's impact on more specialized work, like coding...The Effects of Generative AI on High Skilled Work: Evidence from Three Field Experiments with Software Developers https://bit.ly/40DO8cP #govtech #AI #genAI

17

1 Comment

Do you have interests in discovering how other practitioners are leveraging #AI to address customers' needs? Then, you should attend the AI Tinkerers meetup on July 23rd in Chicago. This time will be extra special as it occurs during the #TechChicago week. #TeChicagoWeek

4

Great reflections on Google Cloud Next '24 by analyst and advisor Janakiram MSV: 💥 "A standout revelation from Google Cloud Next 2024 is the pivotal role that generative AI now plays within Alphabet Inc., with Gemini taking center stage. This shift underscores Google Cloud's critical position in spearheading the next wave of innovation for the tech giant." 💥 "As Gemini integrates deeply into Google's ecosystem, it not only enhances the strategic importance of Google Cloud but also positions it as a driving force behind Alphabet's future growth and technological advancements." 💥 "Compared to keynotes from other hyperscalers, generative AI led by Gemini stole the show." #GoogleCloud #GoogleCloudNext #GenerativeAI #Gemini

1

Check out the great conversation I had with Shaw Talebi, founder of the TDE podcast! We talked about AI, Education, being a YouTuber, Quantum Computing, and the implications of AI in society. https://lnkd.in/eTw-UK-f

46

3 Comments

To Master AI, You Need to Be Better Than AI: New Nature Study Reveals Why A new Nature Magazine paper [1] by Michelle Vaccaro, Abdullah Almaatouq, and Thomas Malone addresses a question I've been pondering: how do 10x engineers use AI to become 100x engineers, while regular engineers seem to struggle to fully harness AI's power? Through a meta-analysis of 106 studies on Human-AI combinations and their performance, the authors revealed three key findings: 1) Human-AI combinations generally yield worse performance than either humans or AI alone at their best 2) However, these combinations prove more effective in creative tasks than in decision-making tasks. I think research and exploratory analysis are prime examples where this synergy works well. 3) Most importantly, the combination shows positive synergy when humans outperform AI. In other words, 10x engineers using AI can become 100x engineers, but engineers performing at similar levels to AI may not benefit from this synergistic effect. These findings carry significant implications. How do we address the needs of those who aren't currently outperforming AI? Will this widen the gap between top performers and the average? Can we design AI to help reduce this disparity? I believe we can. We should focus on bridging the AI usability gap and ensuring AI's potential benefits are accessible to all. Moreover, we need to teach fundamental skills: how to ask good questions and think critically. My experience with AI has shown that even to effectively refine and iterate with AI, you need to correctly interpret its responses first! [1] Link to the paper: https://lnkd.in/dzk3kJ_M

31

3 Comments

Ever thought about becoming a robot? 🤖 No, I'm not talking sci-fi – this is happening now. MIT and UCSD just unveiled "Open-TeleVision," a mind-blowing system that lets you control a robot like it's your own body. But here's the kicker: This tech isn't just cool, it could completely change how we think about AI and automation in business. Why? Because it taps into our most powerful asset: the human brain. What if return to office meant you still work from home? Check out my latest article to learn: - Why human-robot teams might outperform pure AI - The surprising advantages for quick ROI - How this could solve your automation challenges faster than you think Read the full story and let me know your thoughts! #Robotics #AI #FutureOfWork #Innovation https://lnkd.in/eQHqb3Tv

4

2 Comments

LLama3 passed the super-proprietary, advanced test that we designed for modern LLMs! almost... You can read about that test here: https://lnkd.in/gPG8WndR When I first saw the response generating, it looked like I was going to have to come back here and apologize for slandering these machines when they were just one iteration away from solving very basic prompt injection issues. But then I looked a bit closer... can you see the mistake? #promptinjection #countingishard #LLMs #LLama3

13

4 Comments

“We need to bring awareness that real people are building these tools, that their decisions are meaningful and impactful, and their decisions end up being the ethos by which these tools are built. People don't feel empowered to talk about it because it’s beyond their expertise, but our goal is to educate and empower because self-advocacy can make a big difference.” #AI #artificialintelligence #chatgpt #equity #instructionaldesign #learninganddevelopment

5

Building a Compound AI System is like assembling an all-star NBA team. Reposting my blog. I explain "Compound AI Systems" using a basketball analogy. Original blog: https://lnkd.in/gJNFnJWd

13

1 Comment

Exciting news! Narratize, one of the portfolio companies in HWI Fund II, has been accepted into the NVIDIA Inception accelerator! Congratulations to Katie Trauth Taylor, PhD, and the team! Check out the details in their latest blog post. #GenerativeAI #AIinvesting #womenentrepreneurship #howwomeninvest Read more here: https://lnkd.in/g9tJaaHE

39

3 Comments

The future of vector embedding storage is integrated, not isolated 🙌 Learn how new solutions are addressing traditional vector database challenges by treating embeddings as indexes. [Timescale Weaviate Pinecone LanceDB] 📈 https://lnkd.in/gRQg7Z_8

19

1 Comment

🚨 We are starting in 15 minutes - it's not too late to join me!...The Center for Public Sector AI's Q2 2024 Generative AI Update for State and Local Government https://bit.ly/4eaVMA6 #govtech #AI #genAI #SLED

4

… of course.

1

📜 A new blog post explaining the fundamentals of RL for LLMs—the perfect place to intuitively understand PPO and RLHF from first principles. 👇👇 Pre-trained LLMs are often compared to Lovecraft's Shoggoth, a monstrous, shape-shifting beast. PPO is the algorithm that brought it under control. Take a look! 🔗 https://lnkd.in/eXqDvrUV

29

1 Comment