3/ TextGrad – a new framework for automatic differentiation through backpropagation on textual feedback provided by an LLM; this improves individual components and the natural language helps to optimize the computation graph…
RESEARCH
-
Algorithm Cannot Distinguish Positive From Negative Engagement
By
–
fwiw: the problem isnt this post the core problem is that the algorithm cant tell if people engage positively or negatively – negative engagement is easier – hence more done
-

Multimodal LLMs: Analysis and Applications from Recent Talks
By
–
[Slides] Multimodal LLMs Recently, I gave a talk about multimodal LLMs at IndabaX Rwanda and 18-661 Introduction to Machine Learning for Engineers in Student Research Lecture at CMU. In the talk, we dissect multimodal LLMs, multimodal tasks, and a general multimodal LLM
-
Learning AI Through Building: Hands-On Experience Matters Most
By
–
I totally agree with you there Santiago. This line really resonates with me: nothing will teach you more than building things yourself.
-
GPLVM superiority in low data regime versus modern generative AI
By
–
That's why GPLVM works so well on low data … low dim inputs. I don't think there's a "modern" Gen AI approach that competes in (very) low data regime. My favourite demonstration was @baxissimo's Latent Doodle Space paper.
-
Expert Review: Balanced Perspective on AI Pros and Cons
By
–
For people interested in the pros and cons of artificial intelligence, I highly recommend reading or listening to this authoritative and superbly thought out book.
— Nando de Freitas (@NandoDF) 23 juin 2024
Admittedly, as an “AI expert” (apologies for the over-used term), I was a little bored at the beginning, but as I… https://t.co/3M2ClRTqGeFor people interested in the pros and cons of artificial intelligence, I highly recommend reading or listening to this authoritative and superbly thought out book. Admittedly, as an “AI expert” (apologies for the over-used term), I was a little bored at the beginning, but as I
-

Beyond Rational Thinking: Multiple Preconditions for Winning
By
–
yud made the win criteria monocausal (a la @eigenrobot
) when in reality rational thinking is just one of many necessary preconditions to win. just riffing on whats missing: 1) strategic direction: choosing the best objective function given clear perception of reality now, and -

Biological Intelligence: Nature’s Greatest Wonder Explored
By
–
This is a wonderful weekend read. Biological intelligence is one of the true miracles of nature.
-

HyperZ Operator Connects Slow-Fast Networks Context
By
–
HyperZ⋅Z⋅W Operator Connects Slow-Fast Networks for Full Context Interaction Harvie Zhang : https://
arxiv.org/abs/2401.17948 #ArtificialIntelligence #DeepLearning #MachineLearning -

R&D Digitization Becomes Essential for Cost Innovation
By
–
As companies look for ways to reduce costs, speed time to market, and make better products, investing in digitizing the R&D function is less of a choice than a necessity. Source @BainandCompany Link https://
bit.ly/4b4yZmN via @antgrasso #innovation
