Dave Duggal

2.1K posts

Dave Duggal banner
Dave Duggal

Dave Duggal

@Dave_Ideate

Founder, CEO @EnterpriseWeb - An AI-enabled no-code platform for intelligent business and infrastructure apps https://t.co/YXFWKXDZhb

New York Katılım Ocak 2010
1.7K Takip Edilen733 Takipçiler
Dave Duggal
Dave Duggal@Dave_Ideate·
@rao2z Given well documented weaknesses and costs of LLMs we reverse the relationship and ground LLM outputs with an ontology (domain knowledge and operational context) to verify deterministically. It eliminates training, minimize tuning and tokens to reduce costs & energy consumption.
English
0
0
0
12
Subbarao Kambhampati (కంభంపాటి సుబ్బారావు)
I find it hard to buy the claims on the supposed hallucination improvements for LLMs by RAG (Retrieval Augmented Generation). Seems to me that when RAG is claimed to work, it is because the external vector DB is giving almost the right answer to the prompt and LLM winds up summarizing this augmented prompt. But then this is LLM just being a front end to a search engine.. (In fact, if we were to do search engines from ground up now, we will probably store them as vector DBs anyways--since tf/idf ideas have outlived their usefulness, and using the learned embeddings over the web-scale corpora makes retrieval of related pages/context much more effective).
Subbarao Kambhampati (కంభంపాటి సుబ్బారావు)@rao2z

@tdietterich @GaryMarcus To the extent RAG helps, it is often by basically posting the prompt to an external vector DB (aka search engine) and having LLM summarize the augmented prompt, IMHO..

English
15
11
88
23.6K
Dave Duggal
Dave Duggal@Dave_Ideate·
@rao2z RAG and GraphRAG are primitive techniques. Throwing external context at an LLM in the hope the LLM will identify what's useful and improve the output has limited efficacy. The flimsy scaffolding also adds significant cost, complexity and overhead.
English
0
0
1
15
Jerry Chen
Jerry Chen@jerrychen·
The world is creating more data than ever, faster than ever. AI is the wedge we need to keep up.   Startups, this is your time. I see so much opportunity to build tools across the entire enterprise stack that will make businesses smarter and faster. I’m especially excited about: - AI data engineers: We need to make ETL jobs easier and build data pipelines faster - AI-powered business intelligence platforms: There’s opportunity for horizontal as well as vertical/industry-specific BI solutions. Let’s speed up the time to make data into ideas, and ideas into action. If you are a founder building in these areas, please reach out.
Jerry Chen tweet media
English
6
5
37
6.6K
Dave Duggal
Dave Duggal@Dave_Ideate·
@tprstly This config is under "More" > Settings > Privacy > Data Sharing and Personalization
English
0
0
0
22
Dave Duggal
Dave Duggal@Dave_Ideate·
@tprstly Done! Thanks for the tip. It's like play whack-a-mole with who is selling/exploiting your data now. When are we going to get privacy-first legislation that requires clear disclosures and "opt-in". Folks that do opt-in should be able to monetize their personal data.
English
0
0
0
29
Dave Duggal
Dave Duggal@Dave_Ideate·
@EvanKirstel @doctorow Crazy in its implications, but hopefully a wake up call on privacy in this age of centralized information and centralized power. Lesson: If it can happen, it will happen.
English
0
0
0
14
Dave Duggal
Dave Duggal@Dave_Ideate·
@karpathy Network architectures were designed for resilience and each independent node should be designed for resilience. It reflects too much centralization, opaque dependencies, lack of IT best practices.
English
0
0
0
9
Andrej Karpathy
Andrej Karpathy@karpathy·
What a case study of systemic risk with CrowdStrike outage... that a few bits in the wrong place can brick ~1 billion computers and all the 2nd, 3rd order effects of it. What other single points of instantaneous failure exist in the technosphere and how do we design against it.
English
494
712
8.3K
616.5K
Dave Duggal retweetledi
Evan Kirstel #B2B #TechFluencer
In this episode, we sit down with the CEO of EnterpriseWeb to explore the transformative power of their innovative technologies in driving digital transformation across various industries. We'll delve into the company's unique approach to addressing enterprise challenges, highlight key success stories, and discuss the strategic vision and partnerships that set EnterpriseWeb apart. Join us for an enlightening conversation that provides valuable insights into the future of enterprise solutions and the trends shaping the digital landscape.linkedin.com/video/live/urn… #AI #GenerativeAI #EnterpriseArchitecture #developer #appdev
Evan Kirstel #B2B #TechFluencer tweet media
English
0
5
5
1K
Dave Duggal retweetledi
JJ
JJ@JosephJacks_·
Utterly wild. Published in 1953. Predicts a leader literally called "Elon" will be first to colonize Mars. 🚀archive.org/details/Projec…
JJ tweet mediaJJ tweet media
English
5
13
58
9.1K
Dave Duggal retweetledi
Leopold Aschenbrenner
Leopold Aschenbrenner@leopoldasch·
Virtually nobody is pricing in what's coming in AI. I wrote an essay series on the AGI strategic picture: from the trendlines in deep learning and counting the OOMs, to the international situation and The Project. SITUATIONAL AWARENESS: The Decade Ahead
Leopold Aschenbrenner tweet mediaLeopold Aschenbrenner tweet media
English
331
1K
5.4K
2.3M
Dave Duggal
Dave Duggal@Dave_Ideate·
@rao2z It's Plato's Allegory of the Cave - no one wants to see the light. It's easier to conform and watch the shadows projected on the back of cave wall. This is across all media. Stay on message, dumb it down, repeat endlessly.
English
0
0
0
60
Dave Duggal retweetledi
Subbarao Kambhampati (కంభంపాటి సుబ్బారావు)
The reality of AI hype is that there is little market for podcasts that are actually skeptical of the hype.. you get viewers and clicks by leaning into the hype. There is no money in teaching people to be skeptical of their beliefs and oodles in the confirmation bias gig..
English
14
18
157
19.3K
Dave Duggal retweetledi
Yann LeCun
Yann LeCun@ylecun·
If you are a student interested in building the next generation of AI systems, don't work on LLMs
VivaTech@VivaTech

The Godfather of AI is at #VivaTech! Yann LeCun (@ylecun) advises students coming into the industry: "Don't work on LLM. This is in the hands of large companies, there's nothing you can bring to the table. You should work on next-gen AI systems that lift the limitations of LLMs.

English
343
1K
7K
1.7M
Dave Duggal
Dave Duggal@Dave_Ideate·
@furrier @CNBC Amen! It'd be nice if that applied to all news so we can escape echo chamber of vested interests.
English
0
0
0
9