Eric Horvitz’s Post

View profile for Eric Horvitz, graphic

Chief Scientific Officer of Microsoft

In response to an invite from the National Academy of Sciences, Tom Mitchell (CMU) and I put together this overview on "Scientific Progress in AI: History, Status, and Futures." The review serves to provide a "snapshot on AI" at a point in time and serves as a chapter in a forthcoming book, Kathleen Hall Jamieson, et al. on "Realizing the Promise and Minimizing the Perils of AI for Science and the Scientific Community." The review may be especially useful for people outside the AI research community. Direct link:   https://lnkd.in/gXFJFnaC

please post direct .pdf link in comments. this * insanely frustrating * "feature in Li to gain an extra click is one of the worst things ever to be added to this platform.

Milan Milenkovic

IoT System Architect, Author, Advisor, Speaker

2mo

Looks informative and insightful. Is there a pdf version?

Santosh Ananthraman

| AI/ML veteran | Deep experience in industrial, financial, and healthcare applications |

2mo

Excellent timeline of references! Maybe missed this key one by Hornik et al? https://www.cs.cmu.edu/~epxing/Class/10715/reading/Kornick_et_al.pdf

Like
Reply
Heidi Therese Dangelmaier

I run a global all-girl think tank driving the next wave of Intelligence, Innovation, technology and consumer growth. 0. 12.24 THE ASCENT BEGINS.

2mo

Eric Horvitz Is this science or just engineering.. i am not certain there is a lot of science happening here regarding the nature of intelligence

Like
Reply
Rick Gillespie

Rick Gillespie’s AI Methodology (IT IS FREE!!!)🕊️🕊️🕊️

2mo

🚀 Enhancing AI Governance for a Better Future 🚀 Just finished reviewing the AI Overview: History, Status, and Futures (Feb 2024) report, and while it's a great exploration of AI's trajectory, it lacks critical elements that could significantly improve the future of AI governance. 🧠🤖 Drawing from Rick Gillespie's AI Governance Methodology, I believe the future of AI can be drastically improved by incorporating: Philosophical Oversight: Integrating Aristotle’s virtue ethics ensures AI systems are aligned with fairness, honesty, and human well-being. 🌟 Game Theory: Using models like Nash Equilibrium ensures AI behaves optimally in both competitive and cooperative environments, addressing risks from adversarial agents. ♟️ Specialized LLMs with Control LLM: A modular approach enhances transparency and task-specific accuracy, making AI more explainable and scalable. 🔍 Ethical Feedback Loops: Continuous human-in-the-loop auditing ensures AI adapts to evolving ethical standards, preventing bias and improving trust. 🔄 These improvements would transform AI systems into ethical, reliable, and context-aware agents capable of navigating real-world complexities. 🌍

Patrick Hough

Senior Leader in Nuclear Operations, Maintenance & Refueling | Navy Veteran | Expert in Data & AI Innovation | Dedicated to Team Building & Talent Development | Bridging the Gap Between Managers & Cutting-Edge Technology

2mo

Thank you for the insightful AI exploration. I appreciate the depth with which you tackled both the potential and the challenges AI presents. Your focus on the history, current state, and ethics surrounding AI is especially important for those of us working on AI deployment today. The caution you raise about the need for ethical oversight and regulation is key. The risks of bias, lack of transparency, and the dangers of unruly generative models are very real. As we push innovation, we should ensure it benefits society as a whole. It reminds me of how, in 1996, the internet took the world by storm, with regulation only coming many years later. In 1996, early legislation like the Communications Decency Act was just beginning to address the challenges posed by the internet, but comprehensive regulation around privacy, security, and e-commerce didn’t emerge until years later. Looking forward to seeing how these conversations evolve as AI continues to develop.

Bibi Brahim, Msc Machine Learning, Diffusion Models

AI Architect - Full-stack Generative AI Scientist - QA production grade LLM app

1mo

In the future of AI, we will be able to communicate with everything around us. People will even be able to talk with their dogs and cats by implanting chips in their bodies. These chips will enable the animals’ brains to control their mouths and produce speech. Initially, animal welfare organizations might oppose this technology, but over time, they may come to accept it. To clarify, it isn’t actually the animals communicating, but AI mimicking their speech.

Like
Reply
Vincent Granville

AI/LLM Disruptive Leader | GenAI Tech Lab

1mo

And for game-changing stuff that you won't find in mainstream, see: https://mltblog.com/3zsnQ2g

Martin Chernacov

Tech support | IT Consulting | Problem Solver

2mo

What if a different AI could solve the shortfalls of LLM, ML, and it already exists? Where are the curious scientists that are eager to improve it. Please forgive my tone, but is so frustrating to have such an incredible system and no response from scientists, professors, corporations, leaders in the artificial intelligence community, ignoring my emails, connection requests, phone calls, and attempts to reach for someone to listen.

Like
Reply
Gregory H.

TOGAF | Microsoft COTS Enterprise | Cybersecurity

1mo

Eric Horvitz...I took your paper a step further and asked Co-Pilot to summarize and generate questions. My goal is to prove tests and homework are obsolete. I selected the "More Precise" configuration setting in Co-Pilot. Using Scientific Progress in AI - History, Status and, and Futures - National Academies Sure! Here is a question you might consider about the document: Machine Learning Advances: How have machine learning methods, particularly deep learning, contributed to recent AI advancements? Here are the answers Co-Pilot returns to the questions...using the web!! Machine Learning Advances: Machine learning methods, particularly deep learning, have contributed significantly to recent AI advancements. Deep learning involves neural networks with many layers that can learn complex patterns from large datasets. This has led to breakthroughs in areas such as speech recognition, natural language processing, and computer vision. The development of algorithms like backpropagation and the availability of large datasets and powerful computing resources have been key factors in these advancements. #NoMoreHomework #TestsAreObsolete

Like
Reply
See more comments

To view or add a comment, sign in

Explore topics