Bayes in the age of intelligent machines
Ғылым және технология
Tom Griffiths, Princeton University
Abstract: Recent rapid progress in the creation of artificial intelligence (AI) systems has been driven in large part by innovations in architectures and algorithms for developing large scale artificial neural networks. As a consequence, it’s natural to ask what role abstract principles of intelligence - such as Bayes’ rule - might play in developing intelligent machines. In this talk, I will argue that there is a new way in which Bayes can be used in the context of AI, more akin to how it is used in cognitive science: providing an abstract description of how agents should solve certain problems and hence a tool for understanding their behavior. This new role is motivated in large part by the fact that we have succeeded in creating intelligent systems that we do not fully understand, making the problem for the machine learning researcher more closely parallel that of the cognitive scientist. I will talk about how this perspective can help us think about making machines with better informed priors about the world and give us insight into their behavior by directly creating cognitive models of neural networks.
Bio: I am interested in developing mathematical models of higher level cognition, and understanding the formal principles that underlie our ability to solve the computational problems we face in everyday life. My current focus is on inductive problems, such as probabilistic reasoning, learning causal relationships, acquiring and using language, and inferring the structure of categories. I try to analyze these aspects of human cognition by comparing human behavior to optimal or "rational" solutions to the underlying computational problems. For inductive problems, this usually means exploring how ideas from artificial intelligence, machine learning, and statistics (particularly Bayesian statistics) connect to human cognition. These interests sometimes lead me into other areas of research such as nonparametric Bayesian statistics and formal models of cultural evolution.
I am the Director of the Computational Cognitive Science Lab at Princeton University. Here is a reasonably up-to-date curriculum vitae.
My friend Brian Christian and I recently wrote a book together about the parallels between the everyday problems that arise in human lives and the problems faced by computers. Algorithms to Live By outlines practical solutions to those problems as well as a different way to think about rational decision-making.
I am interested in how novel approaches to data collection and analysis - particularly "big data" - can change psychological research. Read my manifesto and check out the Center for Data on the Mind.
cbmm.mit.edu/video/bayes-age-...
Пікірлер: 26
This is a 360 video clip, very odd choice.
@rilmehakonen9688
27 күн бұрын
It's a Bayesian thing. You wouldn't understand. (just kidding)
Was this style of video intentional? I don't think anyone needs an immersive experience when watching a lecture 😅
Gonna use my VR headset to make the presentation more immersive
Looks like the presentation camera is not proper. IS there a better video?
@IanTKhoo
Ай бұрын
tilt your phone to see if it's better
It feels like I'm actually IN THE SLIDESHOW bruh
@rilmehakonen9688
27 күн бұрын
Yeah yeah, it's all about you.
why would you post this? does anyone check what they’re posting on the channel before making the videos public? this is unwatchable!
Why is this 360 🤣
How tf
Interesting, more on academic level. A machine self learning process requires processing of empirical data through a number of iterations, applying dynamic algorithms that is recreated for each iteration using the threshold values created by the previous iteration and input data properties.. The initial distribution of data could be deterministic or assumed based on sample captured from collected data set. Basically, it is an optimisation learning process where the methods from Operational Research could be applied in a iterative steps. Learning to walk is associated with falls where each fall adjust the previous behaviour and creates a new threshold values based on measurement of gap between last threshold value and desired outcome. All is driven by data input and patterns hidden within a data set. So, the dynamic algorithm is a feedback teacher that teach a system how to respond towards expected outcome. The most important part is input data set with structure driven by data ontology hierarchies, but thats another story.
One day we will invent an AI powerful enough to record a lecture that is watchable to the human eye!
How to get the information out of this video: Just below the vid, there is "Abstract: ...more". Click that and you can read the story: just the abstract, or the whole script. 😁
I thought it was the problem of my settings😂😂
Better one, pls. 😢
Proceeds to explain how the transformer works in terms of probability distributions somehow leaving out attention but insists on using AGI arbitrarily. We are doomed with researchers like this…
@GerardSans
Ай бұрын
This is not social media but a university lecture. I’m embarrassed by the frivolous use of AGI and lack of technical rigour.
@username2630
Ай бұрын
When talking about inference in LLMs, which is in the end _sampling_ from an autoregressive model and hence _probabilistic_ , you dont have to mention attention at all.
Nó không hiểu được tiếng nào
good theme, bad video quality
What is that 😂
Bad imagine
AI playing practical joke
This video is a joke 😂