Going from the State of the Art (SOTA) to the Future of LLMs and Generative AI
Greg Makowski
ABSTRACT
What is the future of AI? There has been a buzz lately about ChatGPT and Large Language Models (LLMs). It helps to understand the past, current state of the art, before discussing the future trends, concerns and excitement. This presentation intended to broadly cover many topics, to get a broad sense of what is going on.
AI Progress in the PAST
Types of AI Algorithms
Growth in Complexity over Time
Time Series, From Regression to Large Language Models (LLMs)
How LLM is a Time Series
Caution: a model is no better than it's training data
Emergent Properties
AI State of the Art NOW
Emergent Properties
One Place to find State of the Art (SOTA)* AI's rapid growth
Generative AI: Text to Image DALL*E2
Microsoft's New AI Can Simulate Anyone's Voice From a 3-Second Sample
I Challenged my AI Clone to Replace me for 24 Hours - WSJ
Intel Introduces Real-TIme Deep Fake Detector
ChatGPT 4.2 Test Taking, Languages
Chain of Thought - Size Matters for Reasoning
ChatGPT + Tree of Thoughts Reasoning
Constitutional AI (for ethics and rules)
There is a lot of Generative AI Evolution in a Short Time
Portugal Startup Makes ChatGPT its CEO
ChatGPT 5 coming in 2024
Meta's ImageBind
Looking to the FUTURE of AI
LLM Short term impact to the economy, McKinsey report
Supporting Tech to drive AI
AI in 2-5 years
AI in 10+ years
3 Levels of Future Impossibilities, Michio Kaku
AI, Class I-II Impossibilities
SPEAKER BIOGRAPHY
Greg Makowski has been deploying data mining models since 1992, has worked for 6 startups and been acquired 4 times. He has been building Data Science teams since 2010. He is currently working for Johnson Controls, Inc, with enterprise apps for Internet of Things (IoT), vision applications and LLMs. For more details, see / gregmakowski
0:00 Chapter Intro
1:35 Speaker Intro
4:20 Presentation
1:19:34 Q&A