Jay van Zyl

Interested in real-time predictions and experimentation

[Jay van Zyl @jayvanzyl]

Important factors to consider wrt cost of model training and serving: “SOTA models these days have about ~500B parameters and that represents at least ~1TB of GPU memory to operate with specialized infrastructure. That's a minimum of ~$60,000 - $100,000 p…https://t.co/7PyedGu7Lm

[Jay van Zyl @jayvanzyl]

StableLM is trained on a new experimental dataset built on The Pile, but three times larger with 1.5 trillion tokens of content. The richness of this dataset gives StableLM surprisingly high performance in conversational and coding…https://t.co/4XYLtbg9OQ https://t.co/iDshAxrqad

[Jay van Zyl @jayvanzyl]

The analogy between the syntax-semantics of natural languages and the sequence-function of proteins has revolutionized the way humans inves- tigate the language of life. https://t.co/szapPmr9yA

[Jay van Zyl @jayvanzyl]

With YouTube creators becoming increasingly empowered by versatile generative AI tools, it will only amplify the rising trend of audiences consuming more user-generated content on TVs, conducive to more YouTube advertising revenue,…https://t.co/CsRvgLRs0p https://t.co/aJ3Pzc4Ewu

[Jay van Zyl @jayvanzyl]

They say a good craftsman shouldn't blame his tools, but can a good tool [LLM] blame a shoddy craftsman?
But
Large language models specialize in generating human-like text. Correct answers are a bonus. https://t.co/ViqcLqPM9l

[Jay van Zyl @jayvanzyl]

Brilliant rendition of the human brain https://t.co/or82SGFpi0

[Jay van Zyl @jayvanzyl]

Another key concept to understand: Most of the AI-generated images currently produced rely on Diffusion Models as their foundation. https://t.co/rOUViLJRL2

[Jay van Zyl @jayvanzyl]

Together with https://t.co/VckAzyW9l6 real-time behavioral capabilities, generative models add a much needed angle to AI for business usefulness. Here is a another outline in summary for those who need a quick reference:

Generativ…https://t.co/s8WWxqxUp2 https://t.co/PBwsXmTJni

[Jay van Zyl @jayvanzyl]

Effective intervention :) https://t.co/3xAT9PfmX1

[Jay van Zyl @jayvanzyl]

Cape Town looks like a safe option while we're working on solving all of this :) https://t.co/fab0XN9vs4

[Jay van Zyl @jayvanzyl]

Excellent share @dxbrob. "It is perhaps uncontroversial to say that this claim
that one of us made eight years ago (Soman, 2015)
is now accepted as universal truth. Governments,
for-profit organizations, not for profits, startups,
consumer protect…https://t.co/BUyZaqvXbr

[Jay van Zyl @jayvanzyl]

FinGPT emphasizes the critical significance of data collecting, cleaning, and preprocessing in creating open-source FinLLMs using a data-centric approach. FinGPT seeks to advance financial research, cooperation, and innovation by p…https://t.co/xg1GQIXfRl https://t.co/8LxNu6S0bt

[Jay van Zyl @jayvanzyl]

Extreme actions for good results. https://t.co/2pRxavUa4G

[Jay van Zyl @jayvanzyl]

Great event yesterday in the city, San Francisco #mtpcon https://t.co/Dh5kYIoM1y

[Jay van Zyl @jayvanzyl]

Great paper on transformers: “Transformer large language models (LLMs) have sparked admiration for their exceptional performance on tasks that demand intricate multi-step reasoning. Yet, these models simultaneously show failures on…https://t.co/xPeVzIS8Ag https://t.co/LDXgdXAznr

[Jay van Zyl @jayvanzyl]

How will the generative AI world be affected? https://t.co/HhAwyHh2J8

[Jay van Zyl @jayvanzyl]

Gorilla is a major addition to the list of language models, as it even addresses the issue of writing API calls. Its capabilities enable the reduction of problems related to hallucination and reliability. https://t.co/ux06hP4qUl

[Jay van Zyl @jayvanzyl]

Another great set of models. Why use Falcon-40B?
1. It is the best open-source model currently available. Falcon-40B outperforms LLaMA, StableLM, RedPajama, MPT, etc. See the OpenLLM Leaderboard.
2. It features an architecture optimized for inference, wit…https://t.co/kpucrfPnDR

[Jay van Zyl @jayvanzyl]

Should recommender system be held liable? https://t.co/3CcNxJheHp

[Jay van Zyl @jayvanzyl]

As the commoditization of LLM models continue, here's a list to review. https://t.co/ME7shGG0Ut