Newton Jain

The Economic Times's article on GENAI features my conversation with Himanshi Lohchab on why Cost (especially inference cost) and high quality Conversational Data are the key barriers to make LLMs ready for India Scale. TWO.AI's mission with #ChatSUTRA and #SUTRA models is to bring energy-efficient language models that are truly multilingual. https://lnkd.in/gYshMMNz

81

2 Comments

NEW: Ex-TuSimple CEO and cofounder launches a new robot truck startup Xiaodi Hou, the cofounder and CTO behind the failed startup TuSimple, is making another attempt to master robotic trucking. His new company just raised $20 million from venture investors and aims to launch commercial operations next year. Houston-based Bot Auto plans to operate autonomous semis on highway routes in Texas, hauling freight between warehouses and logistics facilities in Houston and San Antonio. Unlike San Diego-based TuSimple, which delisted shares early this year after a major implosion, the new company will own and operate its trucks directly rather than provide autonomous cargo-hauling technology as a service, Hou said. #selfdriving #trucking

78

2 Comments

At the NVIDIA AI Summit '24, I got a fresh perspective on the evolution of AI in real-world applications. The Levels of AI model is a fascinating framework to understand where products stand today: Level 1: Manual Ops – No AI, purely human-powered. Level 2: Human-Centric AIOps – AI assists with insights but requires human decision-making. Example: Amazon's review summaries help us make buying decisions. Level 3: Machine-Centric AIOps – AI makes recommendations based on user input with minimal human intervention. Think Netflix suggesting titles based on our preferences. Level 4: Fully-Automated AIOps – AI handles it all. A prime example: driverless cars. Interestingly, most products today sit comfortably at Level 2. Moving from Level 2 to Level 3, however, is often more of a product decision than a leap in ML or accuracy. It requires rethinking the user experience and determining how much control to give AI vs. the user. The journey from assistive AI to more autonomous systems is one of trust, user experience, and strategic vision. Exciting times ahead! iMocha #AI #ProductManagement #AIOps #NvidiaAISummit

38

We are launching FreightIntel AI today. It's a tool that digests large sets of transportation data and creates actionable intelligence from that, allowing for chat-like interaction with the insights. It's a standalone product but it's also already natively integrated with our ShipperGuide TMS. You can see the 3-minute demo here. This tool is right for you if: - You are a large shipper with extensive logistics data that finds it difficult to find opportunities for optimization without month-long projects or extensive analyst resources - You run optimization or cost-saving projects only once a year and would like to be nimbler and find savings on an ongoing basis during the calendar year - You use a legacy TMS system with data reporting that contains too much noise and has an unfriendly UI, making it difficult to derive intelligence from raw data - You rely on a managed transportation or 3PL/4PL provider to handle your logistics and want to hold them accountable and ensure they are exploring all optimization opportunities - You have an M&A strategy that relies on identifying cost-savings opportunities of target companies and would like an easy and fast tool to execute this analysis - You are a CFO who is looking to understand potential optimization opportunities in your transportation budget to drive shareholder value Super proud of the Loadies who made this happen - huge effort!

87

3 Comments

The Product Roadmap edition #4: We speak to Sahil Gupta (Mckinsey, Blackstone, HBS, Amazon) who has had a starry career over the last ~15 years. He's worked for over a decade in the US, working at Amazon across multiple different teams, and most recently moving back to India. Key Takeaways from an insightful conversation: 👉 The importance of choosing large and growing trends when picking a career direction. Sahil saw the impact tech was going to have on the world and made a deliberate choice to work in that space. 👉 Amazon is an organization that gives someone in product a large canvas to work on across multiple types of roles - from core consumer facing product, to robots, AI, and even chips. Sahil’s experience at Amazon spans robotics, Alexa, and AI product. 👉 Sahil stresses the importance of calibrating problem statements to PM skills and setting expectations clearly. This is a key part of a PM leader’s job, and PMs should sometimes take initiative to drive these expectation setting conversations. 👉 A framework to thinking about career growth as a PM. Sahil Gupta has just begun a new role at Murf AI. He is hiring PMs (3-5 years of experience) to build the next generation of Voice AI products. Reach out to him if interested. Read the fully story here!

92

2 Comments

I think NVIDIA's strategy is genius. Invest in OpenAI. OpenAI is cooking up more, and larger models. To keep pace, competitors will have to do the same. More compute in the form of GPUs is needed. There is no near-term viable alternative to the current MLP architecture. (Sorry hype men for neuromorphic approaches, and graph mind models, etc. - you'll have to actually produce a demo people can use before I'm buying it can be done near term at scale). I mean sure, small language models, and quantizing models are great - but the benchmarks are clear. The larger models perform better. There's no escaping the scaling laws here. They'll perform even better again when optimized using everything the industry has learned from pruning, quantizing, etc., while still having more parameters and more data. This creates a huge demand for GPUs well beyond OpenAI, but in the entire ecosystem around it. Even if OpenAI makes absolutely no money at all, ever, NVIDIA's investment will be returned to them 100-fold by the stimulated demand for GPUs. That's the definition of a blue-chip investment. #AI #AGI #Datacenters #Energy #OpenAI #NVIDIA

5

At Corgnit Research Lab, India, we are excited to share our latest findings on the convergence of Reinforcement Learning and Federated Learning and their implications for advancing AGI. Key Insights: - 𝐒𝐜𝐚𝐥𝐚𝐛𝐢𝐥𝐢𝐭𝐲 𝐚𝐧𝐝 𝐄𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐜𝐲: The distributed nature of FL enables scalable and efficient training of RL models, significantly reducing computational overhead and accelerating convergence. - 𝐏𝐫𝐢𝐯𝐚𝐜𝐲-𝐏𝐫𝐞𝐬𝐞𝐫𝐯𝐢𝐧𝐠 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠: By ensuring that sensitive data remains local, FL facilitates the development of AGI in privacy-sensitve domains, such as healthcare, without compromising data security. - 𝐂𝐨𝐥𝐥𝐚𝐛𝐨𝐫𝐚𝐭𝐢𝐯𝐞 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠: Multiple RL agents can share knowledge through FL, enhancing overall learning efficiency and robustness without the need to share raw data. - 𝐏𝐞𝐫𝐬𝐨𝐧𝐚𝐥𝐢𝐳𝐞𝐝 𝐀𝐝𝐚𝐩𝐭𝐚𝐭𝐢𝐨𝐧: The integration of FL allows AGI models to be tailored to individual users or environments, combining global intelligence with local customization. - 𝐑𝐨𝐛𝐮𝐬𝐭𝐧𝐞𝐬𝐬 𝐚𝐧𝐝 𝐑𝐞𝐬𝐢𝐥𝐢𝐞𝐧𝐜𝐞: Aggregating diverse experiences through FL results in more generalized and resilient AGI models, mitigating the risk of overfitting. Benefits for Researchers: - 𝐑𝐞𝐬𝐨𝐮𝐫𝐜𝐞 𝐎𝐩𝐭𝐢𝐦𝐢𝐳𝐚𝐭𝐢𝐨𝐧: Leveraging distributed computational resources lowers costs and enhances the feasibility of large-scale AGI experiments. - 𝐄𝐧𝐡𝐚𝐧𝐜𝐞𝐝 𝐂𝐨𝐥𝐥𝐚𝐛𝐨𝐫𝐚𝐭𝐢𝐨𝐧: Model updates can be shared instead of raw data, promoting cross-institutional and interdisciplinary research. - 𝐀𝐜𝐜𝐞𝐥𝐞𝐫𝐚𝐭𝐞𝐝 𝐃𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭: Faster training times and rapid iteration lead to quicker advancements in AGI. - 𝐄𝐭𝐡𝐢𝐜𝐚𝐥 𝐀𝐈 𝐃𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭: Prioritizing user privacy and data security aligns with ethical standards and regulations, fostering responsible AI research. - 𝐑𝐞𝐚𝐥-𝐖𝐨𝐫𝐥𝐝 𝐀𝐩𝐩𝐥𝐢𝐜𝐚𝐭𝐢𝐨𝐧𝐬: Immediate feedback and improvements from real-world data enable practical deployment in sectors like smart cities, healthcare, and IoT. Our research at Corgnit underscores the transformative potential of merging RL and FL in the pursuit of AGI. We believe these insights will catalyze further innovations and collaborations in the AI community. #ArtificialGeneralIntelligence #ReinforcementLearning #FederatedLearning #AIResearch #CorgnitResearchLab #PrivacyPreservingAI #ScalableAI #EthicalAI Corgnit, Corgnit Research India, Reinforcement Learning, Federated Learning Consortium

20

🤔How can an engineer become a master #GTM strategist? I was intrigued by his track record: ✅IIT graduate ✅ #Engineeringleadership roles at iconic companies including Sandisk ✅ Left a senior leadership role at a category-creating Unicorn Nutanix after a successful IPO ✅ Becoming a cofounder and leading GTM at DevRev ✅ Raised $100M at $1.1 B valuation & expanding fast globally I had only heard about him from his CEO & co- founder Dheeraj Pandey previously but today had a chance to meet and listen to Manoj Agarwal in-person. As Aditi Aggarwal engaged Manoj in a riveting fireside chat followed by extended audience Q&A for over 3 hours, it all became clear. His philosophy & practices, how he approaches global go-to-market strategies and fosters a first principles based engineering mindset and a thriving company culture with risk taking and experimentation. In fact, his innate "bania mindset" #DhandeKiBaat and experience as a child helping his parents' mom and pop business became his hidden superpower. His middle class business upbringing helped take DevRev 0 to 1 in the initial founding years from his home garage to now a swanky office in Palo Alto. Not too shabby for someone who developed fluency in English conversation only after coming to the US in his twenties. Clearly Manoj did not let his Hindi medium background & upbringing become a hindrance in his career! Great reconnecting with friends old and new Mani Iyer Divye Kapoor Rajvinder Kaur Anna Vu Sandeep Todi Jimmy Padia Soumitra Sharma Aditya Sharma Manik Mehta I had a blast asking questions and learning during and after the event. What would you want to ask such a unicorn founder if they were to appear on #TheDigitalAgenda show?

55

11 Comments

Improvements include better understanding of context, increased accuracy in generating responses, reduced bias, and enhanced ability to generate coherent and contextually appropriate text. It also has better fine-tuning capabilities for specific tasks and domains, as well as improved handling of nuanced language and ambiguous queries.

1

Fascinating story by Rebecca and Rita at TechCrunch on Mo Chen and TuSimple’s nine-year history that could see the autonomous trucking company in the US pivot to AI and gaming in China - not so much of a stretch as could be first imagined given the power of Chen's guanxi. https://lnkd.in/eZMfT2nh "[Chen's] connections helped the company land a critical investment in 2021 from Sina Corp., which runs China’s biggest social media platform Sina Weibo. Charles Chao and Bonnie Yi Zhang, respectively the CEO and CFO of Weibo, were both members of TuSimple’s board. Sun Dream, an affiliate of Sina, was TuSimple’s largest shareholder for a time, although that investment did lead to an investigation by the Committee on Foreign Investment in the United States. Chen also has a personal connection with Yunli Liu, who was Sina’s head of investment and backed the founder’s web gaming company, Beijing Blue Brothers, which Sina later acquired. Sina also controls the IP of 'The Three-Body Problem' [that TuSimple is now trying to turn into a game]." China is in a desperate search for cash for entrepreneurs more broadly. Eleanor and Xueqiao Wang at the FT point to the almost total collapse in new company formation and venture funding as state ownership and control throttle entrepreneurship. They have even used redemption clauses to try to claw back assets from their insolvent investee companies through the courts. "Today, VC firms have to explain to the state why their companies failed and why they have lost the country’s money. Chinese business publication Caixin reported in August that the country’s leading state-owned VC, Shenzhen Capital Group, has filed 41 lawsuits since 2023, of which 35 were against companies that had largely failed to go public by a set date and had not repurchased shares." https://lnkd.in/eW5BsJnR However, Alibaba and Tencent have been increasing their focus on CVC again and China's size and importance mean they remain bellwethers not just in China but Asia - meet Tencent's managing partner, Jeffrey, and his peers at the www.GCVAsia.com congress in November H/T Jim, Kaz, Yukiko, Carey, David, Yoko, AYANO, Mia, Markus, Sophia, Tim, 🌱🤝🌍 Nicolas, Victoria, Todd, Noriya, Gen, Sumiyo, Che Wang , Min, Peter, (you can register for the delegation here https://lnkd.in/eap8Q32K And please help us prepare for the event with our GCV Keystone annual survey of CVCs: https://lnkd.in/eZV3rpAv

7

The keynote speech by Trevor Cai from #OpenAI at #HotChips 2024 was interesting. Trevor's presentation reiterated the ongoing hypothesis at OpenAI—that the AI models must continue to grow to achieve superior performance, with no apparent end in sight... He bolstered this argument using a log-log power law plot, which essentially plots the relationship between compute (in FLOPs) and model performance (i.e., loss) on a logarithmic scale. This approach has enabled OpenAI to accurately predict the performance of models like GPT-4 by curve-fitting the smaller models on the plot, thus demonstrating the scalability of this method. Trevor then drew a fascinating parallel to Moore's Law, which famously predicted the doubling of transistors on chips every two years. This prediction held true for decades, fueling massive advancements in semiconductor technology. He suggested that similar underlying principles might sustain the exponential growth of AI models for a significant time, though the mechanisms might evolve. However, Trevor notes that such growth isn't without its challenges. He highlighted several critical issues, including the need for advanced data centers, enhanced reliability, availability, and serviceability (RAS) systems, and more sophisticated power management strategies. The massive scale of synchronized training introduces vulnerabilities like power draw jitters and single-point failures that can cause widespread disruptions. If synchronized training causes so many challenges, I wonder if the research community is focusing on alternative algorithms for LLMs that don't require gradient synchronization after each iteration.? 🤔 💡 Trevor points out that as AI models became more intelligent, the demand for inference has grown significantly. He notes that while training infrastructure can be planned and scaled, inference capacity must be more responsive to actual usage and demands, making planning more challenging and dynamic. The technology and economics of inference will continue to evolve. While the conference is about hot chips, he has no specific recommendations on GPU/accelerator architectures, whether inference could benefit from custom hardware, etc. Overall, it is evident that openAI will continue to grow the models and datasets it trained on for each new release of GPT and will continue to challenge the AI/GPU and networking hardware industry to support this with technological advancements. At some point in the future, more than half the data on the Internet will have originated from the LLM models. At that point, how would the log-log plot look? 🤔 Wouldn’t models trained on their self-generated data collapse, and the losses start to grow? He acknowledges the problem in Q&A and says they are "investigating" techniques to overcome this…. Any thoughts? #AI #LLMs #datacenters #junipernetworks

83

8 Comments

https://lnkd.in/esZgE24d Seven years ago I gave this talk at IIT Kanpur. Back then it seemed inevitable that we would have self-driving cars by now, and some predictions I made were wrong. Overall, though, I think this talk has aged reasonably well. Little did I know back then that the reference to the 2016 MaxSAT I gave would spark the creation of a new company, InsideOpt. The Seeker Solver that we sell is based on the technology outlined, and it is changing the optimization landscape. On the social implications, we since edited a special issue of the AI Magazine on the social disruption that AI brings. More work is needed. We need to have a wide societal discourse on how we are going to share the increasing wealth that AI brings to the world. #insideopt.com

8

By end of 2028 this can be as high as $80-100B #plischeme benefits lasts till 2028 . Winning #apple over was key for #manufactuing being 25% of #GDP from current 16% . There needs to be faster reforms to hit $1T merchandize exports faster . whenever you are happy with current growth watch #china $4.5T exports which leaves space open to grab more market share Apple’s India business surpasses $23 billion as manufacturing growth reaches 50-year high

1

This is worth a quick listen. Watch beyond the no-task-beneath-me part to how he shares his logic and reason.

7

Luma AI recently launched #DreamMachine a competitor of OpenAI's #Sora which was announced a while back but yet to be launched. I tried the model and generated 2 videos. Given my interest in Indian Itihasas 1st prompt was - "Lord Hanuman and Lord Ganesh playing football with vanarasena as their teams" It generated a 5sec video, and I didn't find the output to be good. Then being a superhero fan the 2nd prompt was - "Superman and batman playing cricket with batman batting and Superman bowling" It generated a 5sec video, and I didn't find the output to be good but was better than the 1 video. I think the bar is set so high that we expect the world at our fingertips. Must say great to see the pace of innovation! Do checkout lumalabs.ai and try DreamMachine and share your views in the comments. #GenAI #AI #VideoGen #TexttoVideo

28

2 Comments

✅ "Harnessing genAI for enhanced developer productivity at scale" - Below are the Details of my talk - expanding commercial offerings - Fine tuning LLMs - leveraging internal networks #genAI #speaker #ml #ai #slb

309

25 Comments

Meta is facing BIG restructuring toward innovation and growth. CTO, Andrew “Boz” Bosworth, has unveiled the most extensive restructuring of Reality Labs to date, splitting our hardware division into two focused groups: Metaverse and Wearables. 🔷 Metaverse: This team now includes our cutting-edge Quest headset line and is dedicated to pushing the boundaries of immersive experiences. 🔷 Wearables: Encompassing all other hardware efforts, including the innovative smart glasses collaboration with Ray-Ban. This strategic move aims to streamline and sharpen our focus on delivering groundbreaking products in these two domains.  #Meta #Innovation #Wearables #Metaverse #TechTransformation https://lnkd.in/gQ5RdydW

5

Quick trip to DC for the AI-focused session at Semafor’s World Economy Summit 2024. Three takeaways so far: 1. My dishwasher is officially a robot. 2. It’s going to take a while before Amazon Alexa can seamlessly order a Pizza. 3. Productive teams and companies will become more productive. Insightful to discuss AI applications with experts across various industries. At Tincre we first and foremost leverage AI to enhance client results, but also to streamline our team’s workflows. #AI #Semafor #Adtech

10

2 Comments