Jesse Wang (@aiwang) profile picture

AI=I: Automatic Information = Intelligence. Engineering Leader in AI, ML, NLP and knowledge systems. Building e2e chatbots platforms in last 6 years.

Greater New York area
Joined March 2009
Photo by Jesse Wang (@aiwang)Photo by Jesse Wang (@aiwang)Photo by Jesse Wang (@aiwang)Photo by Jesse Wang (@aiwang)Photo by Jesse Wang (@aiwang)Photo by Jesse Wang (@aiwang)Photo by Jesse Wang (@aiwang)Photo by Jesse Wang (@aiwang)Photo by Jesse Wang (@aiwang)
I just published The Best RAG 3: Effective KB Construction medium.com/p/the-best-rag-3-…
1
21
I just published Smart Ingestion for Knowledge Base Construction — The Best RAG (Part 2). This condenses many years of in-the-trench experiences building effective conversation AI applications. medium.com/p/smart-ingestion…
1
27
I just published Outstanding RAG (Part 1). It is the condensation from decades of in-the-trench knowledge management and conversation AI experiences from many domains. Hopefully you find it interesting and helpful! link.medium.com/JMAN1IFAUQb
49
Jesse Wang retweeted
These are not CGI. Reinforcement learning is so back. When operating on strings, it gives us o3. When operating on physical motors, it gives us a perfect humanoid backflip and a robot creature that out-maneuvers almost every animal on earth. RL is one of the only learning algorithms that can master both the world of bits and the world of atoms. Give me a reward function, and I shall move the world. 2025, Year of RL.
108
411
91
2,715
357,096
I fended off the pressure of using langchain a few times in the past 2 years because of exactly the same reason. For prototyping, maybe it's useful; for production, it's much better to just use the most suitable embedding models, vector database, etc. KISS.
I recommend everyone against using LangChain and this article explains well It uses abstractions on top of abstractions and actually makes your code needlessly complicated Just write API calls and add a vector database instead octomind.dev/blog/why-we-no-…
87
"The CODiEs are one of the top awards for software companies. We won in two categories, "Best Customer Service Solution" for the Conversational Cloud, and "Best AI-Enabled Content Solution" for KnowledgeAI. The application and evaluation period took about…lnkd.in/eAHcvqTM
45
We just released Transformers' boldest feature: Transformers Agents. This removes the barrier of entry to machine learning Control 100,000+ HF models by talking to Transformers and Diffusers Fully multimodal agent: text, images, video, audio, docs...🌎 huggingface.co/docs/transfor…
72
775
166
3,206
618,150
How Generative AI Tools Impact the Consumer Experience | LivePerson liveperson.com/blog/generati…
32
At #ElasticON NYC, lots of people packed into the tech session on vector search, trying to learn about it. We @LivePerson are already using it in production. #KnowledgeAI-Search
1
6
Large language models are inefficient, opaque, and static—but *retrieval* can help. New post highlights how retrieval-based systems ColBERT-QA and Baleen advance the state of the art for answering open-domain questions and verifying complex claims. https://teapawt.net/t.co/7rv1TRBYrG
1
10
52
Jesse Wang retweeted
Facebook applied for a patent for a technology built to spy on users' reactions of content through their webcams/smartphone cameras and microphones without their knowledge inc.com/minda-zetlin/faceboo…
7
1
15
Jesse Wang retweeted
WWW inventor & MIT prof @timberners_lee's big idea to save the web: "pods" that let you control your own data (sites visited, products bought). Orgs are then permitted to access part of the pod for specific tasks. NHS is doing a '21 pilot pod program: nyti.ms/2XwYFoE
7
52
4
177
Way to go!
So blessed with the outpouring of support for An Inconvenient Minority. We reached #1 in Asian-American History (not just new releases) in just two days. Pre-order your copy today! amazon.com/gp/bestsellers/di…
1
1
Jesse Wang retweeted
30 years ago today Tim Berners-Lee and Robert Cailliau emailed colleagues outlining a proposal for a hypertext project that they dubbed "WorldWideWeb". w3.org/Proposal.html (v/@w3c)
10
273
92
745
Jesse Wang retweeted
Why can’t programmers tell the difference between Halloween & Christmas? Because oct 31 = dec 25. #programmerhumor #HappyHalloween
8
175
46
858