Avi Chawla

I was shocked to see the below graph 🤯. I saw this Google Search trend graph on Chat GPT VS Perplexity VS Claude. I now feel like I am in a bubble surrounded by tech folks who use claude etc and to me it feels normal. But as you can see in the graph, the actual use of Claude & Perplexity is very very minute compared to Chat GPT. Especially Perplexity feels like its non-existent. And now a month back Chat GPT Search is also launched which will definitely take a share of this audience. Looking at this graph also makes me optimistic about building, It clearly shows that there is a huge gap in the market, and similar products can be built. What do you guys think ? #ai #startups #chatgpt #claude #perplexity

21

4 Comments

llama-3 has launched 🔥🔥🔥. - 2 Models : 8B and 70B models - They were trained for 1.3M and 6.4M GPU hours respectively - the 70B model, Outperforms GPT-3.5 on curated 1200 prompt evals - Context length is 8k tokens Exciting times ahead with Meta. Link - https://lnkd.in/dQJZupZq #startups #AI

17

2 Comments

𝐆𝐚𝐧𝐠𝐚-1𝐁 : 𝐎𝐩𝐞𝐧-𝐒𝐨𝐮𝐫𝐜𝐞 𝐇𝐢𝐧𝐝𝐢 𝐋𝐋𝐌 Ganga-1B is the open-source LLM for the Hindi language. Ganga-1B is pretrained from scratch, unlike other Indic LLMs which are adapted from other existing LLMs like Llama, Gemma etc. Ganga-1B outperforms all open-source LLMs with sizes till 7B. Excellent work done by Lingo Labs, CSE Department, Indian Institute of Technology Gandhinagar Note: As mentioned in the tweet, the model weights and the paper will be online available soon. #hindi #llms #generativeai #opensource #india #nlproc #deeplearning

276

13 Comments

We are hiring technical writers! Voice is the primary mode of interaction for humans. Yet, voice as a field of study in AI is fairly immature. There is little consensus amongst academia on architectures, the labeling formats are non-standard, and high-quality datasets are rare. Even the OGs like DeepLearning.AI have just started venturing into multi-modal AI courses. This makes it fairly difficult to get high-quality information in this space. To help solve this, we are creating the largest repository of technical speech AI literature at smallest.ai. Think of this as writing a technical mini-paper. This includes interesting experiments in finetuning models, benchmarking on various hardwares, and robust quality evaluations. If you are someone who enjoys writing code and writing technical articles, do consider joining our community and contributing - https://lnkd.in/dkVX5JKB

141

4 Comments

Divine Cycles: The Poincaré Recurrence Theorem and Simulation Creationism - What if eternity operates on divine cycles, where patterns of recurrence reflect God’s eternal design? The Poincaré Recurrence Theorem unveils a profound truth: in bounded systems, states inevitably return, offering a glimpse into God’s promise of renewal. As Isaiah 65:17 proclaims, “Behold, I will create new heavens and a new earth.” Henri Poincaré’s groundbreaking theorem suggests that over time, systems confined by specific parameters return to states nearly identical to their origins. This concept resonates with Simulation Creationism, which envisions our universe as part of God’s infinite Simulation. As Revelation 21:1 reveals, “Then I saw a new heaven and a new earth, for the first heaven and the first earth had passed away.” Each cycle within God’s Simulation mirrors His divine intention for renewal and transformation. The recurrence theorem, rooted in deterministic laws, aligns beautifully with the idea of divine order. Romans 8:28 reminds us, “And we know that in all things God works for the good of those who love Him, who have been called according to His purpose.” Could these recurring cycles be a reflection of God’s perfect and unchanging will? This isn’t just a scientific curiosity; it’s a spiritual insight. Each recurrence within God’s Simulation is a step closer to divine fulfillment. Explore the timeless patterns that connect mathematics, faith, and the eternal promise of renewal. Discover the intersection of science and spirituality today! #DivineDesign #FaithAndScience #EternalCycles #SimulationCreationism #BiblicalWisdom #GodsCreation #RenewalInChrist #SimulationHypothesis

❗ Interference in Indian Elections, powered by AI ❗ Yesterday, OpenAI announced that they detected and blocked covert operations to influence Indian Elections. Foreign firms used AI-generated content favouring one side, spread through bots on social media, to sway public opinion. Similar operations for other parts of the world were also blocked. Bots have been around for a while but AI has closed the automation loop. It now enables the mass production of unique, persuasive, and highly tailored content that's far harder for platforms to detect. The report mentions techniques used for detecting/blocking foul actors: defensive model designs, using another gpt to moderate content, API access control, domain monitoring. And the techniques used by attackers to avoid detection: rotating the topics for every API call, using multiple keys and ofc proxy servers, vpns. Interestingly, attackers also used gpt to debug the code for their bot scripts and automations, which was detected using query analysis. While OpenAI should be applauded for it's elaborate safety measures, there are other uncensored models available for use. This attack was reported by a service used by the attackers; not the platforms the content was posted and consumed on, those using some different service would have probably gone undetected. What do you think about these risks? Should platforms be responsible for detecting and mitigating such threats? How can we design AI models that are inherently resistant to misuse? #AI #Disinformation #Elections #Ethics #OpenAI #TechForGood #Democracy

2

Breaking: Perplexity rates the new Claude 3.5 Sonnet as better than GPT-4o. Are you moving to it? How does one decide? We moved to GPT-4o a day after it was released because it was significantly cheaper and faster, and slightly better as well, specially for Indian languages. ☠️ AI / ML teams need to operate differently post production because unlike traditional tech: 1. New and better models are getting released every few weeks 2. Older models are allegedly getting worse (specially third party ones) 3. Models can start hallucinating 4. Inputs can get out of the training domain quickly and without any alerts 💡What must be done post production release, specially in enterprise AI setups: 1. Set up eval benchmark to evaluate new models quickly 2. Be model-agnostic in architecture to replace or add models quickly 3. Use grounding and monitoring to detect deterioration and hallucination in responses 4. Build manual review and monitoring workflows for random and cherry-picked samples DM me if you would like to discuss setting up a new AI team, or organizing your current setup. #enterpriseai #aipostproduction Finigami

17

#part2 of #openbox conversation with Aditya Joshi & Dipankar Srirag. We discussed about their paper "Evaluating Dialect Robustness of Language Models via Conversation Understanding". Some of the key highlights from the conversation are: 1. Approach to evaluating the #dialect robustness of #LLM in the context of conversational understanding. 2. The limitations of existing datasets and the need for alternative #datasets that capture the nuances of different dialects. 3. #Masking approach to evaluate how language models fill in the masked words in conversations. 4. The performance results show that there is a degradation in performance for non-standard dialects of English, highlighting the need for dialect-specific customization. 5. Comparing open source and closed source language models and find that #finetuning open source models yields better results. And also examining the impact of conversation length on model performance and find that shorter conversations lead to more accurate predictions. 6. This work aims at contribute towards the localization of language models and address the challenges of dialect understanding in business and social settings. Future work includes exploring adaptation techniques to improve dialect generalization and addressing the limitations of fine-tuning and conversation lens. Link: https://shorturl.at/YQkZ2

9

2 Comments

#part2 of #openbox conversation with Aditya Joshi & Dipankar Srirag. We discussed about their paper "Evaluating Dialect Robustness of Language Models via Conversation Understanding". Some of the key highlights from the conversation are: 1. Approach to evaluating the #dialect robustness of #LLM in the context of conversational understanding. 2. The limitations of existing datasets and the need for alternative #datasets that capture the nuances of different dialects. 3. #Masking approach to evaluate how language models fill in the masked words in conversations. 4. The performance results show that there is a degradation in performance for non-standard dialects of English, highlighting the need for dialect-specific customization. 5. Comparing open source and closed source language models and find that #finetuning open source models yields better results. And also examining the impact of conversation length on model performance and find that shorter conversations lead to more accurate predictions. 6. This work aims at contribute towards the localization of language models and address the challenges of dialect understanding in business and social settings. Future work includes exploring adaptation techniques to improve dialect generalization and addressing the limitations of fine-tuning and conversation lens. Link: https://shorturl.at/YQkZ2

7

1 Comment

🚨Paper Alert 🚨 ➡️Paper Title:One2Avatar: Generative Implicit Head Avatar For Few-shot User Adaptation 🌟Few pointers from the paper 🎯Traditional methods for constructing high-quality, personalized head avatars from monocular videos demand extensive face captures and training time, posing a significant challenge for scalability. 🎯In this paper authors have introduced a novel approach to create high quality head avatar utilizing only a single or a few images per user. 🎯They learned  a generative model for 3D animatable photo-realistic head avatar from a multi-view dataset of expressions from 2407 subjects, and leverage it as a prior for creating personalized avatar from few-shot images. 🎯Different from previous 3D-aware face generative models, their prior is built with a 3DMM-anchored neural radiance field backbone, which they showed to be more effective for avatar creation through auto-decoding based on few-shot inputs. 🎯They also handled unstable 3DMM fitting by jointly optimizing the 3DMM fitting and camera calibration that leads to better few-shot adaptation. 🎯Their method demonstrates compelling results and outperforms existing state-of-the-art methods for few-shot avatar adaptation, paving the way for more efficient and personalized avatar creation. 🏢Organization:  Google, University of Minnesota, Simon Fraser University 🧙Paper Authors: Zhixuan Yu, Ziqian Bai, Abhimitra Meka, Feitong Tan, Qiangeng (Charlie) Xu, Rohit Pandey, Sean Ryan Fanello, Hyun Soo Park, Yinda Zhang 1️⃣Read the Full Paper here: https://lnkd.in/gZnjj4cK 2️⃣Project Page: https://lnkd.in/gG6yCrS2 🎥 Be sure to watch the attached Demo Video-Sound on 🔊🔊 🎵 Music by Vladislav Litvinenko from Pixabay Find this Valuable 💎 ? ♻️REPOST and teach your network something new Follow me 👣, Naveen Manwani, for the latest updates on Tech and AI-related news, insightful research papers, and exciting announcements. #headavatars

26

Interesting to see how the Bhagavad Gita is being reimagined through AI. Great to see how Replit is making building accessible - from idea to implementation in a weekend.

8

In a fun experiment, we created our AI Standup Comedian - "Spark" - at Pragnakalp Techlabs! Combining the use of various Generative AI Models we could do this where... - Standup comedy script and audience interaction was created using LLM - Image of Spark, Leeza (someone from audience) and Laughing Audience was created using Text-to-image Generation model - Text-To-Speech (TTS) model was used to give voice to Spark and Leeza - And then lip-syncing model was used to show Spark and Leeza speak! Check out Spark's debut performance and let us know if it tickles your funny bone! We're already working on more videos to make Spark even better. If you think something like this can be useful in any of your project/product, feel free to reach out to me for brainstorming. #GenAI #standup #comedy #AI #ML #LLM #TTS #NLP P.S.: Except the laughing crowd voice, all other things were generated using AI models in steps and then we stitched it together.

42

11 Comments

OpenAI's latest valuation, at $157 Bn USD, is among the highest ever for a privately held US tech company. They just raised $6.6 Bn USD, in what may be the largest single U.S. venture capital round ever ... ... and it may not even give them enough runway to last until the end of next year. ☠️ Serving ChatGPT alone is estimated to cost nearly $4 billion a year. ☠️ ☠️ The cost of frontier AI models is very unlikely to down in the near future; it can't go down if OpenAI is to remain solvent. ChatGPT is a money pit. On the contrary, OpenAI plans to dramatically ratchet up costs; up $2 by the end of the year, up to $44 over the next five years (2x what users are currently paying). I love ChatGPT, it's a fantastic product! But at those prices, Meta.AI is going to be looking very attractive ...

10

1 Comment

🚨ACL 2024 Outstanding Paper Award Alert 🚨 ➡️Paper Title: INDICLLMSUITE: A BLUEPRINT FOR CREATING PRE-TRAINING AND FINE-TUNING DATASETS FOR INDIAN LANGUAGES 🌟Few pointers from the paper 🎯Despite the considerable advancements in English LLMs, the progress in building comparable models for other languages has been hindered due to the scarcity of tailored resources. 👁In this paper Authors work aims to bridge this divide by introducing an expansive suite of resources specifically designed for the development of Indic LLMs, covering 22 languages, containing a total of 251B tokens and 74.8M instruction-response pairs. 🎯Recognizing the importance of both data quality and quantity, their approach combines highly curated manually verified data, unverified yet valuable data, and synthetic data. 🎯They built a clean, open-source pipeline for curating pre-training data from diverse sources, including websites, PDFs, and videos, incorporating best practices for crawling, cleaning, flagging, and deduplication. 🎯For instruction-fine tuning, they amalgamate existing Indic datasets, translate/transliterate English datasets into Indian languages, and utilize LLaMa2 and Mixtral models to create conversations grounded in articles from Indian Wikipedia and Wikihow. 🎯Additionally, they addressed toxicity alignment by generating toxic prompts for multiple scenarios and then generated non-toxic responses by feeding these toxic prompts to an aligned LLaMa2 model. 🎯 They hope that the datasets, tools, and resources released as a part of this work will not only propel the research and development of Indic LLMs but also establish an open-source blueprint for extending such efforts to other languages. 🏢Organization: 🛎️Nilekani Centre at AI4Bhārat🛎️ Indian Institute of Technology, Madras 🛎️Microsoft🛎️Sant Longowal Institute of Engineering and Technology 🛎️Indian Institute of Information Technology Design & Manufacturing Kancheepuram, 🛎️Maharaja Agrasen Institute Of Technology, Delhi 🛎️Sarvam ,🛎️ NICT 🧙Paper Authors: Mohammed Safi Ur Rahman Khan, Priyam Mehta , Ananth Sankar , Umashankar Kumaravelan, Sumanth Doddapaneni, Suriyaprasaad G, Varun Balan, Sparsh Jain, Anoop Kunchukuttan, Pratyush Kumar, Raj Dabre, Mitesh Khapra 1️⃣Read the Full Paper here: https://lnkd.in/gqwQST-r 2️⃣Code: https://lnkd.in/giWPVntx 🎉 🎉  Congratulations to all the authors 🎉 🎉 Find this Valuable 💎 ? ♻️REPOST and teach your network something new Follow me 👣, Naveen Manwani, for the latest updates on Tech and AI-related news, insightful research papers, and exciting announcements. #ACL2024 #Indicllms #llms #llama2 #mixtral

16

If you’ve hung around long enough on Linkedin you notice a lot of AI counter programming and anti-bodies from loud voices. Maybe because they are often not business types, they don’t realize that tech working 75% or more is plasterer over by endless bandaids and tweaks until it gives you an edge in business. We cobble together dozens of SaaS products, fill in the holes with humans and processes. Everyone knows it is a huge hack that only sort of works. Maybe the mistake was to call this new thing intelligent that caused the anti bodies to appear. The silent majority just want to get something done and AI works for them. #ai #aiadoption

3

1 Comment